The U.S. Department of Energy’s responsibilities run the gamut from managing the nuclear stockpile and the strategic petroleum reserve to running the power grid in 36 states to performing basic and applied research to protect national security, ensure stable power sector operations and accelerate the clean energy transition. Leveraging the power of DOE’s computing infrastructure, including the world’s fastest supercomputer, simulation models are used to accelerate advancements in nearly every field of research across DOE. Through a series of examples highlighting grid management, cybersecurity, cavern modeling and fundamental physical phenomena, this keynote will illuminate how DOE applies modeling and simulation to both research and operations.

Supply chain resilience refers to the capacity of a supply chain to proactively prepare for unforeseen events, effectively address disruptions, and bounce back from them while ensuring the sustained smooth operation of the supply chain at the preferred level of connectivity and management of its structure and functions. Recent disruptive events including the Covid-19 pandemic and the Russian invasion of Ukraine have caused an increased emphasis on supply chain resilience. In this presentation, we discuss strategies to prepare for, address, and bounce back from (potential) disruptions and the role that simulation can play in enhancing supply chain resilience.

Climate change and the resulting increased frequency of unpredictable extreme weather events create new operational challenges for the commercial seed industry, which is a key pillar of a sustainable and secure global food supply. More specifically, extreme weather events translate into two main effects on agricultural production: Higher yield variability and lower expected yields. In recent years, extreme weather events already caused reductions in the yields of cereals, maize, and other staple crops. It is also projected that a warming of +2C (+4C) would increase the coefficient of variation of corn yield by 62% (192%) in six countries that collectively account for 73% of global production. In this presentation, we first examine how the increased likelihood of extreme weather events affects agricultural supply chains in terms of R&D, production planning, contracting, allocation, and storage decisions. We then discuss the key challenges associated with each stage and highlight how simulation can help address them under increased volatility.

Screening procedures for ranking and selection have received less attention than selection procedures, yet they serve as a cheap and powerful tool for decision making under uncertainty. Research on screening procedures has been less active in recent years, just as the advent of parallel computing has dramatically reshaped how selection procedures are designed and implemented. As a result, screening procedures used in modern practice continue to largely operate offline on fixed data. In this tutorial, we provide an overview of screening procedures with the goal of clarifying the current state of research and laying out opportunities for future development. We discuss several guarantees delivered by screening procedures and their role in different decision-making settings and investigate their impact on screening power and sampling efficiency in numerical experiments. We also study the implementation of screening procedures in parallel computing environments and how they can be combined with selection procedures.

This tutorial discusses principles and strategies for the interplay between applied work with organizations and an academic research agenda. I emphasize lessons I have learned through my own work and my own mistakes, with special focus on some high-stakes settings, including advising Cornell University’s response to the COVID-19 pandemic and work with the emergency services, among other applications.

The goal of this paper is to develop a methodology for the systematic analysis of asymptotic statistical properties of data-driven DRO formulations based on their corresponding non-DRO counterparts. We illustrate our approach in various settings, including both phi-divergence and Wasserstein uncertainty sets. Different types of asymptotic behaviors are obtained depending on the rate at which the uncertainty radius decreases to zero as a function of the sample size and the geometry of the uncertainty sets.

This work provides an overview of digital twins, digital replicas of real entities conceived to support analysis, improvements, and optimal decisions. Specifically, it aims to better clarify what digital twins are by pointing out their main features, what they can do to support their related physical twins, and which models they use. An illustrative example together with a few selected application examples is used to better describe digital twins. A discussion on the actual challenges and research opportunities is also reported.

We review bootstrapping and batching as devices for statistical inference in simulation output analysis. Bootstrapping, discovered in the late 1970s and developed over the ensuing three decades, is widely held as being among the important scientific discoveries of the previous century due primarily to its facility for general statistical inference. By contrast, batching was introduced in the 1960s but was developed within the simulation community (in the 1980s) for the narrower contexts of variance parameter estimation and confidence interval construction. In recent years, however, there has been increasing realization that batching, much like bootstrapping, can be used also for general statistical inference, and that batching often compares favorably with bootstrapping in dependent data contexts. Bootstrapping and batching have tremendous applicability for uncertainty quantification in simulation, and are prime candidates for adoption in simulation software. We describe the general principles underlying bootstrapping and batching, outline guarantees, and discuss implementation.

We present a tutorial on setting-up the oxDNA coarse-grained model for simulations of DNA and RNA nanotechnology. The model is a popular tool used both by theorists and experimentalists to simulate nucleic acid systems both in biology and nanotechnology settings. The tutorial is aimed at new users asking "Where should I start if I want to use oxDNA". We assume no prior background in using the model. This tutorial shows basic examples that can get a novice user started with the model, and points the prospective user towards additional reading and online resources depending on which aspect of the model they are interested in pursuing.

Large deviations theory has a long history of providing powerful machinery for designing efficient rare-event simulation techniques. However, traditional large deviations theory fails to provide useful bounds in heavy-tailed contexts, and designing efficient rare-event simulation algorithms for heavy-tailed systems has been considered challenging. Recent developments in the theory of heavy-tailed large deviations enable designing a strongly efficient importance sampling scheme that is universally applicable to a wide range of rare events. This tutorial aims to provide an accessible overview of the recent developments in the large deviations theory for heavy-tailed stochastic processes, which is followed by a detailed account of the design principle behind the strongly efficient importance sampling scheme for such processes. The implementations of the general principle are demonstrated through a few specific heavy-tailed rare events that arise in stochastic approximation, finance, and queueing theory contexts.

Verification and validation of complex agent-based human behavior simulation models is a challenging endeavor, particularly since a dearth of real-world data makes it impossible to use most traditional validation methods. Data farming techniques have stepped up to the challenge, proving to be a valuable tool for verification and validation of complex models. In this paper we demonstrate how data farming and analysis aids in the verification and validation of complex models by presenting specific examples pertaining to WRENCH, an agent-based simulation model that represents complex interactions between security forces and civilians during civil security stability operations. We first provide an overview of data farming and its relevance for verification and validation of military agent-based simulation models, then give an overview of WRENCH, and finally demonstrate with examples how we have used data farming to aid in the verification and validation of WRENCH.

Machine Learning has become increasingly popular in developing Intrusion Detection Systems (IDS) for cybersecurity. However, the focus has mainly been on achieving high detection accuracy rather than evaluating the impact on cybersecurity resiliency. In this paper, we use agent-based simulation to investigate the impact of different IDS algorithms on the cybersecurity resiliency of organizations under DoS attacks. Our simulation includes a server agent equipped with either Naive Bayes or SMO-based IDS, and a cybercriminal agent capable of launching different types of Denial of Service attacks. Our results suggest that the choice of IDS algorithm can significantly affect an organization’s cybersecurity resiliency against DoS attacks. Specifically, while SMO shows better overall accuracy on the KDD Cup 1999 dataset, Naive Bayes-based IDS proves more effective in practice due to its better-balanced detection rates across different types of DoS attacks. Our findings have important implications for improving organizations’ cybersecurity posture.

Agent-based models can be a powerful tool for evaluating the impact of policy decisions on a population. However, analyses are traditionally beholden to one set of rules hypothesized at the conception of the model. Modelers make assumptions of agent behavior that are not necessarily governed by data and the actual behavior of the true population can vary. Evolutionary Model Discovery provides a solution to this problem by leveraging genetic algorithms and genetic programming to explore the plausible set of rules that can explain agent behavior. Here we describe an initial use of the EMD system to develop robust policies in a resource constrained environment. In this instance, we extend the NetLogo implementation of the Epstein Rebellion model model of civil violence as a sample problem. We use the EMD framework to generate plausible populations and then develop policy responses for the government that are robust across the plausible populations.

Agent-based modelling (ABM) is becoming a popular policy tool by modelling the reasoning processes and interactive behaviors of individuals against external environments. However, the presence of heterogeneous agents, non-linear interactions and complex emergent patterns raised by even simple behavior rules pose challenges in the model explanation process. In this work, we propose a novel iterative analysis method that leverages causal discovery algorithms to facilitate policy formulation and evaluation based on a causal understanding of the model. It strengthens the explanation power of ABM by elucidating causal relations among modelled components. We applied the method to an agent-based simulator that models passengers' routing behaviors in a virtual airport terminal. By discovering the causal relations among passengers' goals, actions, and an airport terminal environment under different COVID-19 regulations, we showed that the method can inform more effective indirect-control policies leading to positive passenger experiences, compared with a conventional ABM analysis method.

This study provides a systematic review of existing studies that used agent-based modeling (ABM) to inform substance abuse policies and identifies future research directions. The detailed review included 20 articles, among which, tobacco, alcohol, cannabis, opioids, and heroin substance abuse were studied. These studies examined substance abuse interventions and the associations between substance use and social behavior, such as peer interaction and selection. Effective interventions included retailer density reduction policies, restriction of trading hours of licensed venues, ecstasy pill-testing and passive-alert detection dogs by police at public venues, and a mass-media drug prevention education policy. ABM can capture the dynamic interactions among and between agents and environments, making it appropriate to model complex substance abuse behaviors. Limitations in current studies include a lack of ABM validation efforts and generalizable data. Future studies should use generalizable and abundant information to inform their ABM, as well as have an explicit validation method.

For emergency departments (EDs) to maintain sustainable care of patients, hospital management must continually explore potential interventions to clinical practice. Agent-based modelling (ABM) can be a valuable tool to support this planning in a controlled environment. Existing approaches to ABM development are best suited for one-off models. However, conditions in EDs can change frequently, making the use of one-off models infeasible. Decision-makers must be able to trust simulations appropriately for them to be effective in intervention exploration. Domain-specific modelling languages (DSMLs) can address these challenges by offering a reusable library of appropriately-abstract, domain-familiar, modelling concepts across case studies and automatic translation of these concepts into executable models. In this paper we present a DSML to support repeated modelling exercises in the ED domain and illustrate the use and reuse of this DSML across two concrete case studies in London-based NHS emergency departments.

Microscopic traffic simulation tools are able to evaluate possible impacts induced by automated shuttles under various conditions. However, automated shuttles operate more and more often in shared space areas and few microscopic traffic simulation tools are able to handle networks with shared space infrastructure. Interaction behaviors between road users and automated shuttles are addressed only seldom as well. In this paper, we propose the concept of bi-directional edges in the open source microscopic traffic simulation suite SUMO to simulate road users’ interactions in a bi-directional shared-space corridor. A case study, where automated shuttles and cyclists share the bike path, and the related data collection were conducted to examine the performance of the proposed concept and understand the usage of the shared corridor. The simulation results are promising. Further refinement of the proposed concept is planned for properly reflecting complex interaction behaviors among diverse road users, and their surrounding environment.

Integrated, demand-responsive passenger and freight transport (IDRT) potentially provides flexibility and higher service frequency in areas of low demand due to economies of scale, while reducing negative traffic-related externalities such as pollutant emissions, noise emissions or accidents. However, to allow for efficient operations in terms of minimum travel distances, short customer waiting times, and high vehicle utilization rates, IDRT requires effective rebalancing strategies that balance supply and demand capacities by strategically positioning vehicle resources in the operational area. Therefore, we propose a rebalancing strategy for IDRT and measure its effectiveness through an agent-based simulation model. To evaluate our approach, we compare the rebalanced IDRT with a static scenario with backhauls to a central depot. Our results indicate that the proposed rebalancing approach can outperform a system without rebalancing by up to 15.1% in terms of total fleet kilometers and 30% in terms of passenger waiting time.

This paper presents an open-source agent-based simulation model to study bio-inspired charging policies for local sustainable energy systems in an industrial setting where electric vehicles (EVs) perform transportation jobs. Within this context, we focus on a system that allows to control the charging-schemes of individual EVs. To this end, we develop an agent-based simulation model in NetLogo. We present and implement a bio-inspired approach based on the foraging behavior of honeybees and our approach results in simple, yet effective decision-making logic. Our approach provides the necessary parameters to control and balance sustainable energy systems in terms of EV productivity and the consumption of locally generated energy. Our simulation results look promising: the balance between EV productivity and the use of sustainable energy can be efficiently tweaked in a predictable manner using the parameters and thresholds of the model, yielding close-to-optimal performance.

Intelligent game agents are a vital part of modern games as they add life, story, and immersion to the game environment. The requests in the gaming industry for more realism have made intelligent agents more important than ever before. Modeling and simulation of game agents and their surrounding environment provide an alternate setting to study dynamic agent behavior before integration into the game engine. The Cell-DEVS formalism, an extension of Cellular Automata, allows modeling such behaviors using the rigorously formalized Discrete Event Systems Specification (DEVS) formalism. In this paper, we explain how to model and test reactive game agents using the Cell-DEVS formalism and the CD++ toolkit. To analyze the dynamic behavior of such agents, we perform several experiments in varying system configurations. Our experimental results confirm the versatility of Cell-DEVS and the functionalities in the CD++ toolkit to model comfort-driven, exploratory, and desire-driven game agents.

Experiments that are games played among a network of players are widely used to study human behavior. Furthermore, bots or intelligent systems can be used in these games to produce contexts that elicit particular types of human responses. Bot behaviors could be specified solely based on experimental data. In this work, we take a different perspective, called the Probability Calibration (PC) approach, to simulate networked group anagram games with certain players having bot-like behaviors. The proposed method starts with data-driven models and calibrates in principled ways the parameters that alter player behaviors. It can alter the performance of each type of agent (e.g., bot) in group anagram games. Further, statistical methods are used to test whether the PC models produce results that are statistically different from those of the original models. Case studies demonstrate the merits of the proposed method.

Simulation models are subject to uncertainty and sensitivity, meaning that even small variations of input can cause considerable fluctuations in the output results. Consequently, this can amplify the uncertainty associated with the simulation, thereby limiting the confidence one can have in its outcomes. To mitigate these effects, this paper suggests using a cooperative game theory-based feature importance method, which can identify uncertainty in a dataset, and provide additional insights that could be used in the development or analysis of a simulation model. A predator-prey scenario was considered, demonstrating its usefulness in identifying important parameters or features. By identifying the most influential parameters or features, this approach can help improve the accuracy, explainability, and reliability of simulation models as well as other models with highly variable input parameters.

This paper develops a simulation-based quantitative method to investigate the joint impact of multiple risks on the supply chain system during the pandemic. A hybrid simulation method that combines the susceptible-infected-recovered (SIR) model and the agent-based simulation method is proposed to simulate the risk propagation along the supply chain and the interactions between distribution centers and retailers. By analyzing the results of scenarios with different interventions under COVID-19, results show that the impact of interventions is diminishing along the supply chain. For intervention deployment, adding testing capacity is of great importance. For stakeholder management strategies, diversifying the upstream partners is helpful. Against the backdrop of a multi-wave global pandemic, this paper takes the COVID-19 pandemic as an example to provide a paradigm for modeling the risk propagation in supply chain systems. Also, the study demonstrates how to estimate possible time-varying risk scenarios in face of the data shortage challenge.

To continue operations of the inland waterway transportation system (IWTS), the interconnected infrastructure, such as locks and dam systems, must remain in good operating condition. However, as the IWTS ages, unexpected disruptions increase, causing significant transportation delays and economic losses. To evaluate the impacts of IWTS disruptions, a Python-enhanced NetLogo simulation tool is developed, where the extreme natural events are considered and represented by a spatiotemporal model. Utilizing this tool, optimal maintenance strategies that maximize cargo throughput on the IWTS are determined via deep reinforcement learning. A case study of the lower Mississippi River system and the McClellan-Kerr Arkansas River Navigation System is conducted to illustrate the capability of the developed simulation and machine learning-based method for IWTS maintenance optimization.

With steadily increasing performance of computers, agent-based modeling has evolved from an analysis method for qualitative phenomena to strategy for quantitative decision support. With this orientation, however, the modeler faces new challenges during implementation. In particular, an appropriate simulation tool must feature the combination of data and model flexibility, process reproducibility, performance and portability. While existing simulators often do not sufficiently cover these features, it is also not sustainable to generally implement models from scratch. In this work, we want to present the idea of simulation templates as a compromise between the two strategies. We show, on the example of our Agent-Based Template and two use cases, the importance of the described challenges and how the simulation template concept supports solving them. We aim to generally promote the idea of developing a customized template, which, as a layer between simulator and from-the-scratch implementation, combines the advantages of both approaches.

Is transparency always beneficial in complex systems such as traffic networks and stock markets? How is transparency defined in multi-agent systems, and what is its optimal degree at which social welfare is highest? We take an agent-based view to define transparency (or its lacking) as delay in agent observability of environment states, and utilize simulations to analyze the impact of delay on social welfare. To model the adaptation of agent strategies with varying delays, we model agents as learners maximizing the same objectives under different delays in a simulated environment. Focusing on two agent types - constrained and unconstrained, we use multi-agent reinforcement learning to evaluate the impact of delay on agent outcomes and social welfare. Empirical demonstration of our framework in simulated financial markets shows opposing trends in outcomes of the constrained and unconstrained agents with delay, with an optimal partial transparency regime at which social welfare is maximal.

We study how experience with asset price bubbles changes the trading strategies of reinforcement learning (RL) traders and ask whether the change in trading strategies helps to prevent future bubbles. We train the RL traders in a multi-agent market simulation platform, ABIDES, and compare the strategies of traders trained with and without bubble experience. We find that RL traders without bubble experience behave like short-term momentum traders, whereas traders with bubble experience behave like value traders. Therefore, RL traders without bubble experience amplify bubbles, whereas RL traders with bubble experience tend to suppress and sometimes prevent them. This finding suggests that learning from experience is a mechanism for a boom and bust cycle where the experience of a collapsing bubble makes future bubbles less likely for a period of time until the memory fades and bubbles become more likely to form again.

A critical design decision for crowdsourcing platforms is the degree to which the platform mediator controls participant interactions. Platforms having a centralized model of mediation optimize for convenience, speed, and security in participant interactions, while platforms operating under decentralized control require greater user effort but offer them greater control and agency. The research described in this paper is a preliminary study using agent-based modeling to evaluate and compare the performance of crowd-shipping platforms with centralized/decentralized control over matchmaking of carriers and senders. Results indicate that centralized matchmaking protects the platform from premature failure when initial carrier/sender participation is low. Furthermore, when the platform’s assignment algorithm is designed to maximize platform revenue, subject to meeting carriers’ profit expectations, centralized matchmaking will tend to outperform decentralized matchmaking for both the mediator and the carriers.

Customers’ waiting times are the most commonly used performance data to measure the quality of service in service systems such as call centers and healthcare. Unlike stationary queueing models where customers’ waiting times are statistically similar, the prediction of waiting times is far less straightforward in time-varying queues having nonstationary demand (i.e., arrival rate) and supply (i.e., number of servers). In this paper, we develop a novel methodology for more accurately computing the wait time distribution in a time-varying queueing system. We design extensive simulation experiments to evaluate our prediction methods. In addition, we discover that the waiting-time prediction is highly sensitive to the work-releasing policy of the staffing plan, i.e., the rule under which the number of servers changes in time.

In this paper, we develop a new staffing algorithm for achieving stable service-level targets in queues with time-varying arrivals. Specifically, we aim to stabilize the tail probability of delay, which is the probability that the waiting time exceeds a designated target τ > 0. We integrate reinforcement learning into the decision making in queueing models; our new method recursively evolve the staffing decision by alternating between two phases: (i) we generate simulated queueing data by operating the system under the present staffing function (exploration), and (ii) we utilize the newly generated data to devise improved staffing decision (exploitation). We demonstrate the effectiveness of our new method using various numerical examples.

This paper estimates the intensity function of a nonhomogeneous Poisson process (NHPP) using a spline-based method with constrained quadratic programming (CQP). Based on the property of B-splines, we transform the estimation problem into an optimization problem and apply CQP to obtain the estimated intensity function with low computational expense. Numerical experiments are conducted to verify the performance of our method. In addition, the impacts of the number of intervals from event-count data and the number of knots in B-splines are also discussed to explore the properties of spline-based models.

When estimating a function of means, where some but not necessarily all of them correspond to rare events, we provide conditions under which having efficient estimators of each individual mean leads to an efficient estimator of the function of the means. We illustrate this setting through several examples, and numerical results complement the theory.

This paper studies the efficient estimation of expectations defined on convex rare-event sets using importance sampling. Classical importance sampling methods often neglect the geometry of the target set, resulting in a significant number of samples falling outside the target set. This can lead to an increase in the relative error of the estimator as the target event becomes rarer. To address this issue, we develop a conditional importance sampling scheme that achieves bounded relative error by changing the sampling distribution to ensure that a majority of samples lie inside the target set. The proposed method is easy to implement and significantly outperforms the existing approaches in various numerical experiments.

In rare-event simulation, importance sampling (IS) is widely used to improve the efficiency of probability estimation. Asymptotic optimality is a common efficiency criterion, which requires that the relative error of the estimator only grows subexponentially in the rarity parameter. Most studies, however, consider low-dimensional problems and the effect of dimensionality is seldom analyzed. Motivated by recent AI-related applications, we take a first step towards high-dimensional rare-event simulation and demonstrate that for very simple examples, IS proposals that utilize exponential tilting, arguably the most common IS approach, can suffer from the "curse of dimensionality". That is, while the growth rate of the relative error is polynomial in the rarity parameter thus leading to asymptotic optimality, the degree of the polynomial depends on the problem dimensionality. Therefore, when the dimension is high, the relative error can be huge even in the rarity parameter regime where IS is conventionally believed to work well.

The initial bias in steady-state simulation can be characterized as the bias of a ratio estimator if the simulation model has a regenerative structure. This work tackles input uncertainty quantification for a regenerative simulation model when its input distributions are estimated from finite data. Our aim is to construct a bootstrap-based confidence interval (CI) for the true simulation output mean performance that provides a correct coverage with significantly less computational cost than the traditional methods. Exploiting the regenerative structure, we propose a $k$-nearest neighbor ($k$NN) ratio estimator for the steady-state performance measure at each set of bootstrapped input models and construct a bootstrap CI from the computed estimators. Asymptotically optimal choices for $k$ and bootstrap sample size are discussed. We further improve the CI by combining the $k$NN and likelihood ratio methods. We empirically compare the efficiency of the proposed estimators with the standard estimator using queueing examples.

In stochastic simulations, input model uncertainty may significantly impact output estimation accuracy. Although variance reduction techniques alleviate the computational burden, input model uncertainty remains unaddressed. Among several variance reduction techniques, we propose a robust version of the importance sampling method. We formulate a min-max optimization problem for finding a robust sampling density for simulation inputs considering a parametric uncertainty set that represents candidates of the true input distribution. We utilize the Bayesian optimization framework for solving the outer problem and the barrier method for tackling the inner problem. By incorporating input model uncertainty in the sampling stage, our method effectively allocates simulation effort to improve estimation robustness. Numerical experiments demonstrate the advantages of the proposed method over a benchmark model assuming a precisely known input model. Our approach produces more accurate output estimation (i.e., an estimator with lower variance), highlighting its robustness and potential applicability in a variety of situations.

Importance sampling (IS) is a classical variance reduction technique. Under mild conditions, an IS estimator is unbiased, so one often seeks variance-minimizing optimal sampling distribution. IS has remarkable success in many applications such as engineering, operations research, and finance. In some applications such as enterprise risk management and input uncertainty quantification, complex simulation designs such as nested simulation arises naturally: The outer-level simulation generates a set of risk factors, i.e., the scenarios, which are used as inputs for inner-level simulations. Nested simulation leads to wasteful use of computations as inner simulation outputs in each scenario are isolated from other scenarios. In this study, we propose, analyze, and test a generalized importance sampling technique for nested simulation. Our generalized IS approach reuses one set of inner simulation outputs across different outer scenarios. Numerical experiments show that our proposal is orders of magnitudes more efficient than the standard procedure.

Bootstrapping has been used to characterize the impact on discrete-event simulation output arising from input model uncertainty for thirty years. The distribution of simulation output statistics can be very non-normal, especially in simulation of heavily loaded queueing systems, and systems operating at a near optimal value of the output measure. This paper presents issues facing simulationists in using bootstrapping to provide confidence intervals for parameters related to the distribution of simulation output statistics, and identifies appropriate alternatives to the basic and percentile bootstrap methods. Both input uncertainty and ordinary output analysis settings are included.

Batching methods operate by dividing data into batches and conducting inference by aggregating estimates from batched data. These methods have been used extensively in simulation output analysis and, among other strengths, an advantage is the light computation cost when using a small number of batches. However, under computation budget constraints, it is open to our knowledge which batching approach among the range of alternatives is statistically optimal, which is important in guiding procedural configuration. We show that standard batching, but also certain carefully designed schemes using uneven-size batches or overlapping batches, are large-sample optimal in the sense of so-called uniformly most accurate unbiasedness from a dual view of hypothesis testing.

Randomized Quasi-Monte Carlo (RQMC) methods provide unbiased estimators whose variance often converges at a faster rate than standard Monte Carlo as a function of the sample size. However, computing valid confidence intervals is challenging because the observations from a single randomization are dependent and the central limit theorem does not ordinarily apply. A natural solution is to replicate the RQMC process independently a small number of times to estimate the variance and use a standard confidence interval based on a normal or Student t distribution. We investigate the standard Student t approach and two bootstrap methods for getting nonparametic confidence intervals for the mean using a modest number of replicates. Our main conclusion is that intervals based on the Student t distribution are more reliable than even the bootstrap t method on the integration problems arising from RQMC.

We propose FQUEST, a fully automated fixed-sample-size procedure for computing confidence intervals (CIs) for steady-state quantiles. The user provides a (simulation-generated) dataset of arbitrary size and specifies the required quantile and nominal coverage probability of the anticipated CI. FQUEST incorporates the simulation analysis methods of batching, standardized time series (STS), and sectioning. Preliminary experimentation with the waiting-time process in a congested M/M/1 queueing system showed that FQUEST performed well by delivering CIs with estimated coverage probability close to the nominal level, even in unfavorable circumstances where the sample sizes were inadequate. In the latter cases and for very small samples for steady-state quantile estimation, the close conformance of the CI coverage probability typically came at the expense of loose CI precision.

We extend the COSIMLA approach (short for "COmbined SIMulation and Linear Algebra'') recently developed in Zheng, Infanger, and Glynn (2022) to compute stationary expectations for Markov chains with large or infinite discrete state space. Our work follows the idea of combing the best of linear algebra and simulation---using linear algebra to compute the "center'' of the state space and using simulation to compute the contributions from outside of the "center''. Different from Zheng, Infanger, and Glynn (2022) that needed to fix a single regeneration state, our work develops a new method that allows the use of a flexible regeneration set with a finite number of states. We show that this new method allows more efficient computation for the COSIMLA approach.

Simulation of queueing networks is generally carried out by discrete-event simulation (DES), in which the simulation time is driven by the occurrence of the next event. However, for large-scale queueing networks, especially when the network is very busy, keeping track of all events is computationally inefficient. Moreover, as the traditional DES is inherently sequential, it is difficult to harness the capability of parallel computing. In this paper, we propose a parallel fast simulation approximation framework for large-scale Markovian queueing networks, where the simulation horizon is discretized into small time intervals and the system state is updated according to the events happening in each time interval. The computational complexity analysis demonstrates that our method is more efficient for large-scale networks compared with traditional DES. We also show its relative error converges to zero. The experimental results show that our framework can be much faster than the state-of-the-art DES tools.

RNA structure and functional dynamics play fundamental roles in controlling biological systems. Molecular dynamics simulation, which can characterize interactions at an atomistic level, can advance the understanding on new drug discovery, manufacturing, and delivery mechanisms. However, it is computationally unattainable to support the development of a digital twin for enzymatic reaction network mechanism learning, and end-to-end bioprocess design and control. Thus, we create a hybrid ("mechanistic + machine learning") model characterizing the interdependence of RNA structure and functional dynamics from atomistic to macroscopic levels. To assess the proposed modeling strategy, we consider RNA degradation which is a critical process in cellular biology that affects gene expression. The empirical study on RNA lifetime prediction demonstrates the promising performance of the proposed multi-scale bioprocess hybrid modeling strategy.

Digital Twins (DTs) have immense promise for exploiting the power of computer simulation to control large-scale real-world systems. The key idea is to evaluate or optimize decisions using the DT, and then implement them in the real-world system. Even with best practices, the DT and the real-world system may become misaligned over time. In this paper we provide a statistical method to detect such misalignment even though both the simulation and the real-world system are inherently stochastic. An empirical evaluation and a realistic illustration are provided.

We consider a stopping problem and its application to the decision-making process regarding the optimal timing of organ transplantation for individual patients. At each decision period, the patient state is inspected and a decision is made whether to transplant. If the organ is transplanted, the process terminates; otherwise, the process continues until a transplant happens or the patient dies. Under suitable conditions, we show that there exists a control limit optimal policy. We propose a smoothed perturbation analysis (SPA) estimator for the gradient of the total expected discounted reward with respect to the control limit. Moreover, we show that the SPA estimator is asymptotically unbiased.

Space-filling designs (SFDs) underpin many large-scale simulation studies. The algorithms that construct SFDs are mostly stochastic and cannot guarantee that optimal solutions can be found within a practical amount of time. This paper uses massive experimentation to find the empirical distributions of a diverse set of design-quality measures in highly-used classes of SFDs constructed by leading software packages. The objective is to provide simulation practitioners with a better understanding of what they can expect from different SFD choices. The results show substantial variability in measures of correlation and space-fillingness in the design classes and dimensions investigated. Therefore, computer experimenters should generate and assess several candidate designs using different random-number-generator seeds to reduce the risk of using a poor design simply due to random chance. We also find that in the largest designs investigated, the uniform designs generally perform best for both our correlation and uniformity measures.

We propose a novel screening framework (abbreviated to TopmSB) to identify the top m key factors affecting the system performance. Our framework builds on the standard SB screening mechanism but incorporates an adaptive multi-armed bandit (MAB) procedure in each stage to prioritize the largest group. Compared to SB, TopmSB avoids specifying perplexing (un)importance threshold parameters, while providing desired computational efficiency and statistical precision guarantee. Numerical experiments demonstrate the efficiency and effectiveness of the proposed method.

We formulate, analyze and solve the problem of best arm identification with fairness constraints on subpopulations (BAICS). Standard best arm identification problems aim at selecting an arm that has the largest expected reward where the expectation is taken over the entire population. The BAICS problem requires that a selected arm must be fair to all subpopulations (e.g., different ethnic groups or different types of customers) by satisfying constraints that the expected reward conditional on every subpopulation needs to be larger than some thresholds. The BAICS problem aims at correctly identify, with high confidence, the arm with the largest expected reward from all arms that satisfy subpopulation constraints. We analyze the complexity of the BAICS problem by proving a best achievable lower bound on the sample complexity with closed-form representation. We then design an algorithm and prove the sample complexity to match with the lower bound in terms of order.

We consider bandwidth selection for kernel density estimation. The performance of kernel density estimator heavily relies on the quality of the bandwidth. In this paper, we propose an efficient plug-in kernel density estimator which first perturbs the bandwidth to estimate the optimal bandwidth, followed by applying a kernel density estimator with the estimated optimal bandwidth. The proposed method utilizes the zeroth-order information of kernel function and has a faster convergence rate than other plug-in methods in existing literature. Simulation results demonstrate superior finite sample performance and robustness of the proposed method.

In black-box optimization problems, Bayesian optimization algorithms are often applied by generating inputs and measure values to discover hidden structure and determine where to sample sequentially. However, information about system properties can be available. In different learning tasks, we may know that the objective is the minimum of functions, or a network. In this paper we consider the case where the structure of the objective function can be encoded as a tree. We propose the new Conditional Gaussian Process tree (CGPT) model for "tree functions'' to embed the function structure and improving the prediction power of the Gaussian process. We utilize the intermediate information at the tree nodes, to formulate a novel likelihood for the estimation of the CGPT parameters. We formulate the learning and investigate the performance of the proposed approach. Our study shows that CGPT always outperforms a single Gaussian process model.

In this paper, we propose a simulation approach to mean-variance optimization for portfolios comprised of derivative securities. The key of the proposed method is on the development of an unbiased and consistent estimator of the covariance matrix of asset returns which do not admit closed-form formulas but require Monte Carlo estimation, leading to a sample-based optimization problem that is easy to solve. We characterize the asymptotic properties of the proposed covariance estimator, and the solution to and the objective value of the sample-based optimization problem. Performance of the proposed approach is demonstrated via numerical experiments.

Minimization of the carbon footprint of aviation is an active area of interest to the industry and policy makers alike. Optimization of the individual flight phases is an important step in that direction. This paper considers the holding phase, wherein aircraft hold in the terminal airspace of airports prior to approach and landing during times of busy operation or when the arrival capacity is reduced due to factors such as bad weather. We propose a tactical method to allocate landing slots while minimizing the environmental impact of holds. An environmentally-driven policy can be perceived as unfair, particularly by airlines whose environmentally friendly aircraft which might need to hold longer than they would under a fair first-come-first-served policy. To alleviate this challenge, we propose a number of economic reward schemes, including one based on a linear programming problem obtained by applying complementary slackness to the dual of the assignment problem.

This paper describes the use of variable-sized entities within the framework of the InControl simulation software product Pedestrian Dynamics to rapidly model passenger flow and congestion for a series of check-in hall lobby designs for a US domestic airline terminal. Note that the airline and airport will remain anonymous for this presentation due to confidentiality.

This paper presents a reinforcement learning (RL) algorithm prototype to solve the aircraft line maintenance scheduling problem. The Line Maintenance Scheduling Problem (LMSP) is concerned with scheduling a set of maintenance tasks during an aircraft's ground time. To address this problem, we introduce a novel LMSP method combining a hybrid simulation model and reinforcement learning to schedule maintenance tasks at multiple airports. Initially, this paper briefly reviews the existing literature on optimization-based and AI-enhanced aircraft maintenance scheduling. Secondly, the novel reinforcement learning LMSP method is introduced, evaluated using industry data, and compared with optimization-based LMSP solutions. Our experiments demonstrate that the LMSP method using reinforcement learning is capable of identifying near-optimal policies for scheduling line maintenance jobs when compared to the exact and heuristics-based methods. The proposed model provides an excellent foundation for future studies on AI-enhanced scheduling problems.

Accurate navigation and control of Aerial Vehicles requires precise estimations of their position and attitude. Measuring an aircraft's rotation involves comparing two vectors in different reference frames, such as inertial and body axes. Typically, a GNSS sensor-based matrix with at least three sensors is utilized for this purpose, taking advantage of the carrier phase measurements. However, factors such as multipath, frequency lock loss, cycle slips, and severe clock drifts can impede accurate integer ambiguity resolution. To address these challenges, a new neural network-based technique has been developed to optimize the management of large amounts of data and increase carrier phase ambiguity resolution reliability. By using carrier phase difference and pseudorange information, various neural network configurations can be trained to solve the ambiguity and estimate the precise attitude of the GNSS sensor matrix. The provided solution can be used alone or hybridized with other attitude sensor such as gyroscope information.

Modern enterprise infrastructures (EIs) including those of industrial control systems (ICSs) are becoming increasingly crucial to businesses in a wide range of sectors spanning multiple end-user verticals (e.g., energy, chemical, manufacturing, biotechnology). These EIs improve the (real-time) decision support, productivity, and efficiency of business processes, but necessarily reliant upon the cyber-resilience of complex infrastructures for sustainable business continuity. We are interested in the long-standing open question in the cyber-resilience domain: how can managers formally quantify cyber-resilience for any complex networked EI (sub-)system in the event of a cyber-attack affecting its multiple (inter-dependent) components? We propose a simulation-backed framework derived from probabilistic graph theory to answer this question. We pioneer the derivation and analysis of a quantifiable, closed-form manager friendly expression exhibiting the degree of cyber-resilience (dependent upon individual EI component functionality quality and the varying extents of functional dependencies across networked components) within the (sub-)system post cyber-attack(s) affecting an EI.

The rapid growth of distributed energy resources (DERs), such as renewable energy sources, generators, consumers, and prosumers in the smart grid infrastructure, poses significant cybersecurity and trust challenges to the grid controller. Consequently, it is crucial to identify adversarial tactics and measure the strength of the attacker’s DER. To enable a trustworthy smart grid controller, this work investigates a trustworthy artificial intelligence (AI) mechanism for proactive identification and explanation of the cyber risk caused by the control/status message of DERs. Thus, proposing and developing a trustworthy AI framework to facilitate the deployment of any AI algorithms for detecting potential cyber threats and analyzing root causes based on Shapley value interpretation while dynamically quantifying the risk of an attack based on Ward’s minimum variance formula. The experiment with a state-of-the-art dataset establishes the proposed framework as a trustworthy AI by fulfilling the capabilities of reliability, fairness, explainability, transparency, reproducibility, and accountability.

The density of enterprise cyber (re-)insurance markets to manage (aggregate) enterprise cyber-risk has been low enough to realize their potential to significantly improve cyber-security and consequently the cyber-reliability of (ICS) enterprise ecosystems. In this paper, we propose the use of catastrophic (CAT) bonds as a radical and alternative residual cyber-risk management methodology to alleviate the big supply demand gap in the current cyber (re-)insurance industry, by boosting capital injection in the latter industry. Two important follow up questions arise: (i) when is it feasible for cyber (re-)insurers to invest in CAT bonds? and (ii) how can we price cyber-CAT bonds conditioned on the feasibility condition(s)? We focus on answering the second question pivoted upon an existential answer to the first. We propose a novel practically motivated information asymmetry (IA) driven cyber-CAT bond pricing model, built upon theories of financial stochastic processes and Monte Carlo simulations, in realistic arbitraged incomplete markets.

Socio-Cyber-Physical Systems are ubiquitous in today’s world. They are inherently complex systems built out of many large-scale systems that encompass different perspectives and numerous stakeholders. This leads to several challenges in managing their complexity and emergent behavior. In addition, these systems tend to include many adaptive and autonomous systems with different goals and different adaptations to environment changes or failures. The design, analysis, and testing of such systems is inherently challenging but is becoming critical due to their wide adoption. In this panel, we aim to discuss some of these challenges and potential solutions.

We assert that simulation should be an integral part of technology development and acquisition. Its use to iteratively evaluate new technology across the development timeline can help ensure technologies contribute to resilience in work operations. This, in turn, benefits trust and likelihood of adoption. Potential hindrances to simulation in technology development are the time and complexity simulation can introduce. Time may be needed to model entities and dynamics to be simulated, plan and conduct simulation-based tests and experiments, and translate the results into requirements, user stories, or other inputs to the technology’s design and implementation plan. Complexity is increased when simulation results suggest new or changed requirements, identify technology design and implementation improvements, or produce conflicting feedback from potential users. We will discuss these challenges, methods and tools that minimize their disruptive effects, varieties of simulation we have used to support technology development, and benefits of using simulation in development.

Enterprises are increasingly facing the need to be resilient in the face of uncertainty and dynamism. Simulatable digital twins have become critical aids for analyzing and adapting complex systems. Design thinking and service design methodologies, in contrast, are gaining momentum for ideation, subjective evaluation, and innovation. A systematic application of these methodologies to explore innovative ideas and a faithful virtual environment to test and fine-tune those ideas without impacting real systems could be transformational. This paper presents an approach that establishes a symbiotic relationship between these two approaches to introduce precision and innovativeness to make enterprises resilient. We describe the key characteristics of resilient enterprises, present our approach, and illustrate its effectiveness with a case study focusing on a transformation toward a new normal to address the Covid-19 pandemic induced disruptions in the IT industry.

In multi-tasked services such as in messaging-based contact centers, parallel service interactions share a mutual dependence through the agent's concurrency. Here, we introduce Markov process simulation methods for bivariate Hawkes cluster service models that are not Markovian by default due to their concurrency dependence. To do so, we propose an alternate construction that maintains extra "shadow" variables for how the process would be under other concurrency levels. We prove that this construction yields an equivalent Markov process, and we show through numerical experiments that its corresponding simulation algorithm is significantly more efficient than the non-Markovian alternatives.

Power grid planners and power portfolio managers are increasingly concerned with anticipating “net demand” risks, which is defined as customer demand minus renewables for a particular time period. Net demand is a better predictor of grid stress than peak demand in a grid with significant renewables penetration. For Holy Cross Energy, Sunairio simulated 1,000 probabilistic outcomes of hourly weather across a geographic region that encompassed the locations of customers and renewable energy resources (wind, solar), for 15 years. The hourly weather simulations were transformed to hourly energy simulations of customer demand, wind generation, and solar generation via machine learning models, creating a broad, climate-change-aware, coincident data set from which to quantify concurrent risks to net demand. Net demand paths of particular interest for grid planning were curated via statistical processing.

Disruptions in the supply chain network can have significant and far-reaching consequences, especially in pharmaceutical supply chains that affect health and financial outcomes and raise equity concerns. To inform strategies that can address this critical global problem, we study disruptions in pharmaceutical supply chains using multiagent simulations. These simulations include decision-theoretic agents with a theory of mind reasoning that allows them to reason about the other agents in the supply chain, including their trustworthiness. The simulations reveal how supplier-buyer interactions have non-local effects which can exacerbate and extend disruption impacts. In addition, a distributor’s focus on its own short-term profit can lower its long-term profit and damage equity in healthcenters. We also demonstrate how agents adapt to changes in the environment and changes in other agents’ behavior and how in the absence of explicit communication and coordination, the timing of these adaptations inhibits disruption mitigation efforts from transpiring.

Non-pharmaceutical interventions (NPI) have been proven vital in the fight against the COVID-19 pandemic before the massive rollout of vaccinations. Considering the inherent epistemic-aleatoric uncertainty of parameters, accurate simulation and modeling of the interplay between the NPI and contagion dynamics are critical to the optimal design of intervention policies. We propose a modified SIRD-MPC model that combines a modified stochastic Susceptible-Infected-Recovered-Deceased (SIRD) compartment model with mixed epistemic-aleatoric parameters and Model Predictive Control (MPC), to develop robust NPI control policies to contain the infection of the COVID-19 pandemic with minimum economic impact. The simulation result indicates that our proposed model can significantly decrease the infection rate compared to the practical results under the same initial conditions.

This paper develops a probabilistic model to assess the cascading failure of transformers in an electric power grid experiencing geomagnetic disturbances caused by a solar storm. We propose a model in which the probability of failure is a function of the intensity of the solar storm, the physical properties of the transformer, the geographical location of the transformer, and the flow of electrical power. We demonstrate the proposed model using the IEEE 14-bus system and several notional solar storms. The model quickly computes the initial and cascading failure probabilities of the transformers in the system as a first step towards quantifying the risks posed by future solar storms.

Heat exchangers in the steam generation system (SGS) of concentrated solar power (CSP) plants are unique in their functionality. Consequently, equipment replacements have long lead times. A typical CSP plant using an organic Rankine cycle has one or two salt-to-steam trains (SSTs) within the SGS. When one heat exchanger in the SGS fails, the individual SGS fails. We use an existing framework that combines simulation and optimization models to assess the impacts of irrecoverable failures on long-term production. The methodology provides an optimized dispatch with the integration of unplanned simulated failures over a thirty-year period. Our work shows a system of two trains provides resiliency and reduces downtime of a plant by six to eight times compared to a single train. The gross revenue increases by 31% and 11% for single and two trains, respectively, when the expected lifetime increases from five to 10 years.

A traditional metamodel for a discrete-event simulation approximates a real-valued performance measure as a function of the input-parameter values. We introduce a novel class of metamodels based on modular dynamic Bayesian networks (MDBNs), a subclass of probabilistic graphical models which can be used to efficiently answer a rich class of probabilistic and causal queries (PCQs). Such queries represent the joint probability distribution of the system state at multiple time points, given observations of, and interventions on, other state variables and input parameters. This paper is a first demonstration of how the extensive theory and technology of causal graphical models can be used to enhance simulation metamodeling. We demonstrate this potential by showing how a single MDBN for an M/M/1 queue can be learned from simulation data and then be used to quickly and accurately answer a variety of PCQs, most of which are out-of-scope for existing metamodels.

Simulation built upon partial and ordinary differential equations has been a classic approach to modeling cardiac electrophysiological dynamics. However, mitigating the computational burden of differential equations is still a challenging problem. This paper provides a novel alternative utilizing data-driven recurrent neural networks for cardiac electrophysiological dynamic simulation. Specifically, we develop a long short-term memory (LSTM)-assisted simulation to capture the underlying dynamics of cardiac electrophysiology while preserving computational efficiency. Experimental results demonstrate the efficiency and effectiveness of the proposed method, which outperforms the differential equation-based simulation approach while significantly reducing the computational cost. The proposed method offers a promising alternative to traditional simulation and may contribute to the development of more efficient and accurate approaches for simulating cardiac electrophysiology.

Computational models help decision makers understand epidemic dynamics to optimize public health interventions. Agent-based simulation of disease spread in synthetic populations allows us to compare and contrast different effects across identical populations or to investigate the effect of interventions keeping every other factor constant between "digital twins." FRED (A Framework for Reconstructing Epidemiological Dynamics) is an agent-based modeling system with a geo-spatial perspective using a synthetic population that is constructed based on the U.S. Census data. In this paper, we show how Gaussian process regression can be used on FRED-synthesized data to infer the differing spatial dispersion of the epidemic dynamics for two disease conditions that start from the same initial conditions and spread among identical populations. Our results showcase the utility of agent-based simulation frameworks such as FRED for inferring differences between conditions where controlling for all confounding factors for such comparisons is next to impossible without synthetic data.

Modern parallel/distributed simulations can produce large amounts of data. The historical approach of performing analyses at the end of the simulation is unlikely to cope with modern, extremely large-scale analytics jobs. Indeed, the I/O subsystem can quickly become the global bottleneck. Similarly, processing on-the-fly the data produced by simulations can significantly impair the performance in terms of computational capacity and network load. We present a methodology and reference architecture for constructing an autonomic control system to determine at runtime the best placement for data processing (on simulation nodes or a set of external nodes). This allows for a good tradeoff between the load on the simulation's critical path and the data communication system. Our preliminary experimentation shows that autonomic orchestration is crucial to improve the global performance of a data analysis system, especially when the simulation node's rate of data production varies during simulation.

Simulation and optimization of procurements might employ clustering dataset elements to exclude possible duplicates and improve processing resiliency. This study presents a case of applying scaling methods to reduce computation time of clustering between a smaller and a larger dataset. In this example (of selecting close supplier names), computation time scales as square of N (the number of elements), and the presented approach in effect brings computing time to be linear in N. As a result, computation time in our case is reduced by over an order of magnitude.

Adopting data-driven decision-making approaches can significantly enhance profitability and foster growth in economic situations through quantitative analysis of market dynamics. One intriguing market that warrants examination is the price competition observed within the Danish pharmaceutical sector, where numerous companies are vying for a larger market share through the offering of diverse pharmaceutical products. This paper aims to shed light on this market by employing subsequence time series clustering techniques to identify pricing patterns among the players involved in the Danish pharmaceutical industry. The data analysis pipeline performed in this study allows for the identification of price patterns for clustering and discovering different agent groups, as well as providing a foundation for expanding the current agent-based model of the European pharmaceutical parallel trade market by analyzing the pricing behavior and patterns of players, facilitating the utilization of historical data to model agent behavior and advancing research in this area.

One approach to construct or calibrate simulators, when representative real data exist, is to ensure that the synthetic data generated by the simulated match the empirical distribution of the real data. However, such approach to construct simulators does not take into consideration where the constructed simulators will be used. For some applications, there are clear tasks (such as performance evaluation of different decisions) in users’ mind where the simulated data will serve as input to the tasks. In this work, we propose an approach to use the knowledge of these tasks to guide the construction of simulators, in addition to the distribution match of simulated data and real data by regularizing the objective function with a task related penalty. We conduct a preliminary numerical study of this approach to illustrate the effectiveness compared to not taking into consideration the specific tasks of the simulators.

In a high-tech production environment, capacity investment and production planning are often based on the demand information from manufacturers within a supply chain. A supplier solicits forecast information from a manufacturer, and the manufacturer provides demand forecasts that are updated on a rolling horizon basis. Problems arise with this setup if the manufacturer provides volatile forecast quantities due to the market's fluctuating demand or internal bias. As a result, suppliers' mistrust regarding forecast quantities grows, leading to adjusted production plans based on planners' anecdotal experience. The paper presents a decision model to determine the reliability of forecasts provided by manufacturers to facilitate better production planning. The study also suggests alternate forecasting techniques in case of low reliability. To evaluate the effectiveness of the proposed approach, a simulation study is conducted for different manufacturers and scenarios. Our experiments showed an average cost reduction of 14% across all instances.

The world is increasingly reliant on energy systems, making them a critical infrastructure for essential services. This also makes them vulnerable to attacks, which can result in significant disruptions and damage. Microgrid (MG) monitoring systems play a crucial role in ensuring the safety and reliability of energy systems. However, traditional fault diagnosis techniques are limited to already established faults due to the use of only historical data, making it challenging to keep up with the increasing demand for safety and reliability. This paper proposes a digital twin based machine learning (DTML) framework for fault diagnosis in MG monitoring systems, with a focus on assessing the resilience of MG end-to-end systems to potential disruptions from adversaries. The proposed framework utilizes digital twin based random forest (RF) and support vector machine (SVM) and logistic regression (LR) model and shows that the RF based model outperforms other models with an accuracy of 95%.

IT/OT driven industrial control systems (ICSs) such as water/power/transportation networks are increasingly meeting the daily functional needs of civilian society around the globe. This, alongside making societal businesses more automated, efficient, productive, and profitable. However, often poorly configured IoT security settings increase the chances of occurrence of (nation-sponsored) stealthy spread-based APT malware attacks in ICSs that might go undetected over a considerable period of time. The ICS enterprise management is often keen to get apriori statistical estimates of cyber-loss impact post any cyber-attack event such that it can plan ahead on its cyber-resilience budget. In this paper, we propose the first mathematical theory, based upon stochastic processes and concentration inequalities, to (a) statistically quantify apriori the cyber-loss impact (distribution) on an ICS infrastructure network post an APT cyber-attack event, and subsequently (b) bound the tail of such a cyber-risk distribution, for arbitrary impact distributions.

Natural disasters disrupt systems, leading to critical infrastructure vulnerabilities prone to cyber-attacks. The MITRE ATT&CK Enterprise Matrix is a knowledge base for threat analyses in the cybersecurity community. Existing processes to derive possible attack methodologies from this Matrix are largely manual and time-consuming. It is essential to automate the information retrieval process to reduce human errors, improve efficiency, and free up resources for identifying unrevealed cyber-attacks. We propose a framework that incorporates Natural Language Processing (NLP) and Text Mining to automatically generate sets of attack paths from the technique descriptions in the Matrix. The framework generates similarity between techniques based on their descriptions and creates an output showing potential pathways an adversary can take to infiltrate a system. The outputs are compared against an annotated approach and attack report. The results of this study provide an approach to more quickly and effectively assess potential cyber-attacks towards protecting critical infrastructure.

In a circular economy, recycling of materials at the end of a product's life cycle is a key issue. This paper discusses the sustainability impacts of different recycling strategies for CFPR-made aircraft and how they weigh up against alternative measures such as waste reduction and lower material consumption in the manufacture of the product. The analysis includes environmental and cost impacts for different strategies and market scenarios. A quantitative system dynamic simulation of the life cycle of an aircraft program is used. The subject of the life cycle simulation model is the CFRP mass flow, CO2 emissions and associated costs. In addition, the effects of R&T investments in new technologies for recycling and waste prevention as well as the reduction of material consumption were investigated.

Due to the vitality of semiconductor products for other industries, the production of semiconductors and impact of external disruptions on the semiconductor supply chain should be well understood. As semiconductor manufacturing is accompanied with intrinsic long manufacturing cycle times ranging from 50 to 100 days where operations run 24/7, 365 days per year, correct understanding of potential disturbances should be considered. Examples of these disturbances include pandemics, extreme weather events, geopolitical tensions and war. These hazards pose various risks for supply chains, for example, the bullwhip and ripple effect. To simulate the result of such risks, a simplified system dynamics model of a typical semiconductor manufacturing supply chain was constructed using the Anylogic Software. The model serves as a what-if scenario foundation to evaluate certain external circumstances dependent on current global situations to enhance supply chain resilience

Local food systems, in which consumers source food from nearby farmers, offer a sustainable alternative to the modern industrial food supply system. However, scaling up local food production to meet consumer demand will require farmers to allocate more land to this purpose. This paper describes an agent-based model that represents commodity-producing Iowa farmers and their decisions about converting some of their acreage to specialty crop production for local consumption. Farmer agents’ land-use decisions are informed by messages passed to them via their social connections with other farmers in their communities and messages from agricultural extension agents. Preliminary experimentation revealed that leveraging extension agents to increase the frequency and strength of messages to farmers in support of local food production has a modest positive impact on adoption. By itself, however, this intervention is unlikely to yield significant improvements to food system sustainability.

The fashion industry is widely known as one of the most environmentally impacting. To address the overconsumption issue, the fashion renting business model allows renting clothes or accessories instead of buying them, extending the useful life of products. However, concerns about the sustainability of fashion renting supply chains are arisen, especially due to reverse logistics. In this context, a hybrid simulation model is developed to support fashion companies in the design and evaluation of renting supply chain configurations. Through Discrete Event Simulation (DES) logistics flows are represented, while Agent-Based Modeling (ABM) integrated with Geographic Information System (GIS) allow to represent supply chain’s nodes in the real environment. GIS concurs to estimate the sustainability of the supply chain importing effective data related to the covered distances. The proposed parametric model will enable performing scenario analyses to assess the best configuration in terms of environmental impact.

The current energy crisis and high fossil fuel costs are challenging energy intensive industries such as non-ferrous foundries. It is therefore important to promote the transition to renewable energy sources with the electrification of melting units. This pilot study is the first to simulate the transition of conventional foundries to sustainable technologies. For this purpose, a simulation model based on a selected example company is developed. It takes into account the energy consumption and the logistical effects of a converted operation. The simulation model is implemented as a hybrid simulation combining a discrete event simulation at the plant level and a process simulation within the furnaces. The study shows how a sustainable energy supply can be achieved in foundries. The effects of efficiency as well as energy costs and emissions are also taken into account.

Buildings contribute to about 40% of global energy-related CO2 emissions, and reducing energy demand in buildings has become one of the vital components of the current climate change mitigation strategies. Optimizing energy for the urban building stock by energy-efficient retrofits is becoming increasingly popular in developed countries where the functional and construction elements of the stock are uniform, along with the updated stock database already built in desirable standard formats for energy simulation exchange. However, a decision support system to arrive at energy-efficient retrofits for developing countries where the building stock is highly diverse, with varying construction and operational philosophies, and has no readily available datasets of existing stock is highly challenging. To close this gap, this study suggests an adaptable decentralized community building energy simulation and modeling schema using free and open-source tools for retrofit decision-making.

Harmful Algae and Cyanobacteria Blooms (HACBs) are dangerous dynamic processes for the users/inhabitants of the hydric resources. Their development and contingency plans can be anticipated by using Autonomous Surface Vehicles (ASVs) equipped with a self-driven system capable of deciding how to displace the ASV and its multi-parametric probe to take measurements in the 3D locations of the water body where the HACB is likely to occur. This paper presents a new self-driven system for that purpose, consistent on 1) an offline trajectory planner for the ASV that exploits the information provided by a commercial HACBs simulator to optimize, in turn, the ASV horizontal and probe vertical displacements; and 2) a guidance and control system specially designed for making the ASV follow the planned trajectories. The paper also presents a comprehensive set of simulations to evaluate our proposal's performance and adjust its parameters.

The present study investigates the extension of an existing simulation model combining system dynamics and discrete event simulation by linear optimization for an electric vehicle charging system. The existing simulation framework is extended by a smart charging strategy based on linear programming in order to exploit the flexibility of real charging processes at a workplace parking lot for a better integration of solar photovoltaic electricity generation. Therefore, different smart charging strategies are evaluated. In multiple simulation runs, the strategies are compared with immediate charging using a stationary battery energy storage system for intermediate storage of electricity generated by solar photovoltaic. Results show that smart charging strategies can achieve similarly good results with respect to the self-sufficiency rate and self-consumption rate. In the context of a 100kWp PV system the combination of optimizing charging rates and stationary battery energy storage resulted in self-sufficiency rates of more than 90% in the simulation.

We develop a simulation-based tool for supporting direct current fast charger (DCFC) installation decisions. Our simulation captures details of the DCFC network configuration, non-stationary arrival patterns of the electric vehicles to the fast charging DCFC stations, various DCFC attributes, charging time distributions, and customer behavior. The statistical analysis of the simulation generated output data produces various key performance indicators (KPIs) including DCFC utilizations, number of electric vehicles charged and left uncharged, and queueing experience of the customers. One of the key challenges of developing this simulation is its validation: we have validated the simulation with the historical DCFC charging session data and past observations of the DCFC utilizations. The resulting data-driven simulation is used for supporting DCFC planning through its capability to conduct scenario analysis and predict various KPIs.

The global climate crisis and population growth restrict the availability of essential environmental resources such as water and energy and this situation continues to deteriorate. If and when conditions become extreme, only the well-offs will have access to these valuable resources. With that in mind, we look for solutions to achieve equity within societies while preserving, the degree possible, natural resources. We suggest a method for setting differential pricing for each population stratum, so that each spends a relatively similar percentage of their income on these basic commodities, without depleting valuable resources. Our method optimizes the prices while simultaneously estimating the unknown consumption–price relation. We show the effectiveness of our method based on data from Israel and through extensive simulation experiments reflecting different levels of income inequality within societies, different consumption–price relations, and resource availability. Our study shows that equity and resource preservation can go hand-in-hand.

The coastal zone of the Ganges-Brahmaputra-Meghna (GBM) Delta is widely recognized as one of the most vulnerable places to sea-level rise (SLR), with around 57 million people living within 5 m of sea level. Sediment transported by the Ganges, Brahmaputra, and Meghna rivers has the potential to raise the land and offset SLR. There is significant uncertainty in future sediment supply and SLR, which raises questions about the sustainability of the delta. We present a simple model, driven by basic physics, to estimate the evolution of the landscape under different conditions at low computational cost. Using a single tuning parameter, the model can match observed rates of land aggradation. We find a strong negative feedback, which robustly brings land elevation into equilibrium with changing sea level. We discuss how this model can be used to investigate the dynamics of sediment transport and the sustainability of the GBM Delta.

Climate change presents a significant challenge for many coastal communities as sea level rise is expected to cause widespread and chronic flood inundation. This study examines the case of a coastal watershed of ecological importance, which is threatened by sea level rise and land subsidence, as well as seasonal severe storms. The health of the watershed and flood inundation protection to the community depends on water outflow; something which sea level rise will further restrict. Infrastructure planning for an active water management solution resilient to severe storms and electrical grid disruptions is needed. A dynamic simulation is used to evaluate microgrid energy system design performance and effectiveness in powering a critical infrastructure pumping station during storm-related electrical grid outage and restoration scenarios.

This paper provides an introductory overview of how one may employ importance sampling (IS) effectively as a tool for solving stochastic optimization formulations incorporating tail risk measures such as Conditional Value-at-Risk. Approximating the tail risk measure by its sample average approximation, while appealing due to its simplicity and universality in use, requires a large number of samples to be able to arrive at risk-minimizing decisions with high confidence. In simulation, IS is among the most prominent methods for substantially reducing the sample requirement while estimating probabilities of rare tail events. Can IS be similarly effective for optimization as well? This tutorial aims to provide an overview of the two key ingredients in this regard, namely, (i) how one may arrive at an effective importance sampling change of measure prescription at every decision, and (ii) the prominent techniques available for integrating such a prescription within a solution paradigm for stochastic optimization.

This tutorial aims to introduce Event Graphs (EGs), invented 40 years ago by Lee Schruben to allow event-based modeling of discrete dynamic systems. Their simplicity and naturalness in causality modelling and simulation modelling made EGs popular in research and practice. In a simulation, an event causes state changes in a system as well as other events to happen in the future. EGs provide a parsimonious diagram representation for the Event Scheduling paradigm of Discrete Event Simulation. We first introduce their visual syntax and informal semantics, and then present a recent extension by adding objects to EGs. Our tutorial also includes an introduction to the formal semantics of EGs and a Python implementation for executing EGs.

This tutorial defines what a digital twin is and outlines its four required characteristics. Digital twins are developed to derive insights to control entities and processes in the digital world with simulation as one of the key technologies lying at the heart of this development. The resulting insights are used to prescribe actions in the physical world to fix future problems before they happen. This tutorial describes the key digital twin development functions together with the digital twin enabling technologies with focus on the use of simulation for process twin development. The corresponding functions and technologies are displayed on several different digital twin development frameworks with the potential to serve as guides for practitioners interested in developing digital twin solutions. We conclude with an example of a supply chain digital twin use case and the role of simulation and AI in the twin development.

How can you make your projects successful? Modeling can certainly be fun, but it can also be quite challenging. With the new demands of Smart Factories, Digital Twins, and Digital Transformation, the challenges multiply. You want your first and every project to be successful, so you can justify continued work. Unfortunately, a simulation project is much more than simply building a model -- the skills required for success go well beyond knowing a particular simulation tool.

Output-data analysis is arguably the most-researched topic in the field of simulation modeling, with more than 1000 technical papers having been written. However, many of the published papers are highly mathematical in nature, making them difficult to understand for many simulation practitioners. In this tutorial, we discuss the replication and replication/deletion approaches which can address most analysis problems using three simple formulas (or expressions) from a first undergraduate statistics course. Although the replication approaches discussed above are widely used for estimating the mean of a single simulated system, we show that the same three formulas can also be used to compare any number of simulated systems, to handle multiple system performance measures simultaneously, and also to estimate performance measures such as probabilities and percentiles rather than just means. We also discuss a relatively simple graphical methodology for determining a warmup period if steady-state characteristics are of interest.

This tutorial reviews methodologies for quantifying statistical uncertainty in computationally expensive black-box models, which arise frequently in data-driven simulation analyses under input uncertainty. When facing these models, it can be difficult to run repeated evaluations due to computation cost, and also to obtain auxiliary information such as gradients due to analytical intractability, thus rendering many traditional statistical approaches challenging to apply. We describe several lines of approaches to resolve these challenges, including data-splitting methods based on batching variants, a recent so-called cheap bootstrap approach, and subsampling schemes. We discuss the applications of these approaches to simulation, including problems suffering from both aleatory error exhibited via Monte Carlo noises and epistemic error stemming from the input uncertainty.

Metamodels are fast-to-compute mathematical models that are designed to mimic the input-output behavior of discrete-event or other complex simulation models. Linear regression metamodels have the longest history, but other model forms include Gaussian process regression and neural networks. This introductory tutorial highlights basic issues in choosing a metamodel type and specific form, and making simulation runs to fit the metamodel. The tutorial ends with a warning on potential pitfalls, and suggestions on further reading to expand your knowledge of metamodeling.

The Discrete-Event System Specification (DEVS) is a formalism devised by Bernard Zeigler in the late 1970s for modeling complex dynamical systems using a discrete-event abstraction. At this abstraction level, a timed sequence of pertinent "events'' input to a system causes instantaneous changes to the state of the system. The main advantages of DEVS are its precise, implementation independent specification, and its support for modular, hierarchical composition. This tutorial introduces the Classic DEVS formalism in a bottom-up fashion, using a simple traffic light example. The syntax and operational semantics of Atomic (i.e., non-hierarchical) and of Coupled (i.e., hierarchical, connecting interacting components) models are introduced. Finally, a simplified DEVS model for performance analysis of vessel movements in the Port of Antwerp is presented. All examples in the paper use PythonPDEVS, though other DEVS tools could equally well be used. We conclude with suggestions for further reading on DEVS theory, variants, and tools.

The COVID-19 pandemic has had a profound impact on the lives of working parents, who are struggling to balance their responsibilities at work and at home, as well as childcare providers who are working hard to keep their doors open. In this paper, we examine the effect of childcare policies on the availability of childcare. Specifically, we investigate how classroom size, the likelihood of COVID-19 infection, and the number of days a classroom may need to close affect the amount of time parents will need to stay at home with their children. Our results show that even low probabilities of infection combined with stringent policies can have a large impact on the duration of a child's exclusion from childcare services.

The COVID-19 pandemic has emphasized the importance of preparedness and response plans for healthcare providers and rational responses from society to effectively manage infectious disease outbreaks. Strategic guidelines should be created to ensure the availability of required resources while considering the rational response of individuals under different policy scenarios. This study uses a simulation-optimization-game theory approach to first determine the daily number of infected people in response to social distancing policies in a game theoretical setup. Second, this daily number of infected people is used in a simulation to determine an optimal replenishment policy for restocking personal protective equipment (PPE) items. The model incorporates a combination of mean field games modeling and a simulation model in Simio to perform optimization tasks. This approach aims to guarantee the availability of required resources by taking into account the rational response of individuals under different policy scenarios.

Resource planning during pandemics presents many challenges and equitable decisions about resource allocation must be made. There is no standard definition of equity. Robust mathematical formulations can require a lot of data. In a novel pandemic there is limited historical information available to inform decisions. Decision makers can look to define equity through population proportions (pro-rata). This notion of equity is readily implementable. We present a practical framework for an equitable allocation of scarce resources using population proportions, disease demographics, and resource utilization. We assess our framework using a stochastic simulation model, calibrated to COVID-19 case data, in a case study for convalescent plasma distribution in the context of the clinical trial CONCOR-1. We show that pro-rata resource allocation can be inequitable and that decision makers can consider readily available information, such as resource utilization and case data, to inform equity and proactively manage scarce resources during a pandemic.

During the COVID-19 pandemic, India has seen some of the highest number of cases and deaths. Quality of data, continuously changing policy, and public health response made forecasting extremely difficult. Given the challenges in real-time forecasting, several countries had started a multi-team collaborative effort. Inspired by these works, academic partners from India and the United States setup a repository for aggregating India-specific forecasts from multiple teams. In this paper, we describe the effort and the challenges in setting up the repository. We discuss the development of simulations of compartmental models to model specific waves of the pandemic and show that the simulation model designed specifically for the Omicron wave was able to predict the onset and peak sizes accurately. We employed a median-based ensemble model to aggregate the individual forecasts. We observed that median-based ensemble was relatively stable compared to the constituent models and was one of better performing models.

Evidence has shown that random screening tests are effective in reducing COVID-19 infections in schools. However, test administration may be hindered due to a limited budget or low participation caused by pandemic fatigue. Thus, we seek to balance the number of tests administered with end-of-semester infections. To do this we use an SEIR model to simulate SARS-CoV-2 transmissions within K-12 schools, design a multi-objective simulation optimization problem, and tune an NSGA-II algorithm to find the best testing schedules. We find the Pareto front of optimal schedules of screening tests, which can be used by stakeholders to inform test administration strategies. We discuss insights about the characteristics of optimal strategies, for example, when there are limited number of tests available or a desire to use few tests, the optimal plan is to perform the tests earlier in the semester and at higher intensity.

While mathematical models of disease have been important drivers of public policy since the eighteenth century, the incorporation of endogenous behavior driven by risk perception is a relatively recent phenomenon (Klein et al., 2007). Models incorporating behavior as endogenous variables may enhance their usefulness by providing an explicit mechanism for how behavior varies in response to public health measures and epidemic dynamics, resulting in a more nuanced understanding of disease transmission. We conducted a systematic scoping review to understand the extent to which endogenous behavior was incorporated into models of COVID-19 transmission.

This research explores the resilience components in emergency departments (EDs) during surges through discrete-event simulation (DES). By focusing on the resistance and recoverability components, the resilience of the ED is analyzed, as well as the flow of the patient and the resources required at each step. A simulation is developed to model an ED in the UAE and validated through collected timestamps. The results demonstrate the ordinary conditions of the ED and its calculated resilience, recoverability, and resistance, as well as its strength under conditions of surge demand. To investigate the impact of resources on the ED’s resilience, the resilience triangle is analyzed, and different interventions are applied by adding physicians, nurses, and beds and their effects. The methodology and simulation model provides significant insights to ED managers to evaluate and improve their department’s resilience during surges and emergencies.

The care of multiple victims such as natural disasters in an Emergency Department is critical. This differs from ordinary care by the number of patients that arrive, their severity and the insufficient staff for these events. Designing and simulating this real life scenario will be useful for disaster management decision makers. The objective of this simulation is to model a system with resilience to critical situations. To model the input of this research, we worked with the percentage of patients received by the Cauquenes Hospital during the Chilean Earthquake February 27, 2010. A comparison of two situations is made: the admission of patients before an earthquake with a normal daily attention versus the admission of patients before an earthquake and the activation of the relief chain. The latter situation allows the system to be resilient and adapt quickly to its new reality.

We describe the design of a generalizable simulation model of an emergency department (ED) that forms part of a symbiotic simulation tool designed to improve short-term decision-making. While the paper will give an overview of the planned symbiotic simulation tool, our focus here is on the generalizability of the simulation model. The model is coded such that the routing logic of patient pathways are not explicitly defined but are instead included as an input parameter. By structuring the model this way, the pathways can instead be discovered through process mining methods on standard healthcare transactions data. This enables the simulation model to be applied to other EDs without redesigning all of the logical flows within the model. As symbiotic simulation tools are designed for ongoing use within the system they model, utilizing process mining also allows for automating recalibration of the patient pathways if changes occur in the physical system.

We construct a simulation model of a homeless care system to determine the amount of new housing and emergency shelter needed to support the growing unsheltered population in Alameda County, California. To quantify the performance of the system, we assess the number of people having unmet need via an estimate of the quantile field using a recently developed batching method. This approach helps right-size the amount of housing and shelter resources needed to quickly provide services to the unsheltered population. We find that with a large investment in housing to help the system reach steady state, current levels of emergency shelter may be sufficient to serve those with unmet need.

Despite high levels of spending on prenatal care, the U.S. has the worst maternal mortality outcomes amongst peer high-income nations. In response to a growing need for modernized prenatal care policies, national prenatal care stakeholders have developed a new model of prenatal care, which moves away from a “one-size-fits-all” model of prenatal care delivery, and instead tailors care to patients’ specific needs. In this article, we develop a data-driven discrete event simulation model to quantify the operational impacts of adopting this new care paradigm. We consider a case study of a large academic health center, and derive input parameters for the model from historical data. Our results suggest that when compared with the “one-size-fits-all” model of care, the new tailored care policy leads to reduced patient delays, as well as a reduction in overbooking, implying increased flexibility in the system.

The increase in elective surgical waiting lists as a result of the COVID-19 pandemic is creating significant consequences for health services worldwide. In the UK, the allocation of capital funds to increase capacity for managing elective waits has created planning and operational challenges for health services. This paper reports on the development and deployment of an interactive web-based discrete-event simulation model for supporting capacity planning of surgical activity and ward stay in a proposed new ring-fenced orthopedic facility in a UK health service. The model is free and open-source and developed to be generic and applicable for new capacity planning of elective recovery in orthopedics in other regions. With minor adaptations it can also be readily modified for application to other specialties. Given the current relevance of managing record elective waiting lists, there is potential widespread applicability of the simulation model which is supported by our open approach to modeling.

Individual-based models (IBMs) of infectious disease dynamics with full-country populations often suffer from high runtimes. While there are approaches to parallelize simulations, many prominent epidemic models exhibit single-core implementations, suggesting a lack of consensus among the research community on whether parallelization is desirable or achievable. Rising demands in model scope and complexity, however, imply that performance will continue to be a bottleneck. In this paper, we discuss the requirements and challenges of parallel IBMs in general and the German Epidemic Micro-Simulation System (GEMS) in particular. While the exploitation of unique model characteristics can yield significant performance improvement potential, parallelization strategies generally necessitate trade-offs in either hardware requirements, model fidelity, or implementation complexity. Therefore, the selection of parallelization strategies requires a comprehensive assessment. We present a point-based evaluation scheme to assess the potential of parallelization strategies as our main contribution and exemplify its application in the context of GEMS.

Ensuring safety and public health is a paramount concern in mass vaccination against contagious respiratory infections. This study examines the effects of layout methods and path routing decisions on average patient travel distance (TD) and time-in-system (TIS) within the context of a theoretical mass vaccination clinic. Two distinct layout methods, Perimeter and Serpentine, are evaluated in conjunction with two path routing conditions, Cyclical and Unidirectional. Employing discrete-event simulation, the study investigates multiple patient turnouts and clinic operational hours. The results reveal the significant impact of layout on average TD, underscoring the heightened efficiency of the Perimeter layout and Unidirectional path. Furthermore, the findings highlight the significant effect of layout method on TIS when considering optimal staffing configurations. Conversely, the analysis indicates that path directionality does not exert a statistically significant effect. This study emphasizes the critical role of layout design in optimizing vaccination clinics for efficiency and effectiveness.

High-resolution network-based contagion models are being increasingly used to study complex disease scenarios. Due to network-induced heterogeneity and sophisticated disease and intervention models, even simple simulation exercises can lead to large volumes of complex simulation outcomes. New approaches are required to analyze them. Simulations of such network spread processes can be viewed as attributed temporal graphs. We describe a network-based analytics framework that enables a user to leverage this graphical viewpoint and apply graph mining methods to perform fine-grained analysis of the simulation outcomes and the underlying network. The framework is based on a microservices-oriented architecture, and is designed to be general, adaptable, and scalable. We demonstrate its utility through a case study motivated by the COVID-19 pandemic involving the spread of two variants on a large realistic population network with multiple interventions. We study the transmissions within and between age-groups, importance of non-essential interactions, and efficacy of interventions.

Due to the aging global population, the science of heart recovery is an essential area for research to improve patient health, reduce time-to-discharge, and delay overall mortality. New medical device technology is needed to advance these goals. For the medical community to gain trust in and use these technologies in their hospital environments, optimal study design and proper execution of randomized controlled trials is necessary. Such RCTs will result in the collection of valid scientific evidence for establishing the new device’s risk and benefit profile in targeted patient populations. Continuous time-to-event survival models are commonly used to determine the amount of data needed to demonstrate an improvement in these profiles over current standard-of-care therapies. This paper will compare simulated power functions and sample size requirements for a variety of survival methods in a two-sample RCT setting. Simulation scenarios will encompass various effect sizes, survival distribution forms, and time-to-event density functions.

The Organ Procurement and Transplantation Network (OPTN) in the US allocates kidneys for transplantation, but nearly one fifth of kidneys from deceased donors are not utilized due to the avoidance of transplantation for kidneys that have been removed from a donor for too long. To be able to provide clinically relevant recommendations to the OPTN contractor, we updated the KSIM discrete event simulation of kidney allocation in the academic literature using actual OPTN individual-level records for patients and donors. As a case study, we simulated offering kidneys at high risk of discard to the first accepting transplant center after 10 hours of accumulated cold time and found increased utilization. The updated model allows for greater clinical fidelity and can be embedded in medical decision support systems.

Understanding why patients' viral loads vary dramatically across individuals is a critical challenge in addressing respiratory infections, especially the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The spatial-temporal dynamics of viral infection in the respiratory system and the immune system's response remain difficult to study. Using modelling and simulation (M&S) techniques may address this problem. In this paper, we present a novel modelling approach using the Cell-DEVS formalism (a combination of Cellular Automata and DEVS), to simulate the spatial-temporal dynamics of viral spread in the lungs. Using a two-dimensional cellular space that mimics a lung, the proposed approach focuses also on the immune system response, viral infection spread, state of lung epithelial tissue damage, and immune cells' state. We demonstrate the pertinence of our proposal on three different scenarios representing three types of patients. Qualitative evaluation by expert biologists confirms that the produced simulations match the observations made on patients.

This study introduces an innovative strategy for addressing the Home Healthcare and Dial-a-Ride Problem (HHCDAP) concerning the transportation of medical staff and patients, taking into account the stochastic nature of service and travel times. The problem involves assigning suitable medical staff to patients and clients, determining the order of visits, and identifying opportunities for medical staff and patients to share trips. We propose two objective functions to minimize travel time for drivers and medical staff. This problem adheres to numerous constraints, including maximum work duration, maximum waiting time, professional qualifications, and vehicle capacity limitations. We test our approach on a small-scale instance to understand the trade-offs between minimizing drivers' travel time and minimizing the travel and waiting times of medical staff and patients. Our results indicate that the proposed strategy enhances the efficiency of transporting medical staff and patients.

Total hip and knee arthroplasty (THA/TKA) surgeries are frequently performed on elderly individuals and consist of preoperative, operative, and rehabilitation stages. Despite efforts to improve patient satisfaction,there is a lack of personalized studies that optimize the THA/TKA pathway. Our aim is to address this gap by proposing a predictive simulation model that considers patient-specific factors to enhance patient satisfaction and organizational efficiency. To achieve this, we propose using process mining techniques to analyze the French national healthcare database and distinguish between standard care phases and patient-dependent phases. We then apply machine learning algorithms to predict specific stages of care. The insights gained from these analyses are used to compare and test predicted patient pathways and their performances using our simulation model.

Emergency Departments (EDs) are the primary access points for millions of patients seeking medical care. The increasing patient demand and lack of long-term dynamic planning strain the EDs in providing timely patient care, leading to crowding. While a well-recognized problem, ED crowding is still prevalent, where suboptimal resource allocation is one significant contributing factor. In this research, we developed an end-to-end solution that first forecasted the patient arrivals to the partner ED and then used an optimization model to develop an optimal physician staffing schedule to minimize the combined cost of patient wait times, handoffs, and physician shifts. Finally, the new schedule was tested using the validated simulation model to evaluate the ED performance. By generating shift schedules based on forecasts and testing them in the validated simulation model, we observed that patient time in the ED and handoffs could be reduced by 5.6% and 9.2% compared to current practices.

Conventional simulation models used in Operations Research and Management Science (OR/MS) use historical data. With the increasing availability of real-time data, technologies commonly associated with applied computing, such as Data Acquisition Systems (DAS), may need to be integrated with conventional OR/MS models to develop Hybrid Models (HMs). We distinguish between HMs that use only real-time data – we refer to them as Digital Twins (DTs) – and those using a combination of historical and real-time data – called Real-time Simulation (RtS). Our previous contribution focused on the challenges of such integration, a concept referred to as information fusion, and presented a conceptualization of DT/RtS. This paper focuses on DT/RtS data synchronization and methods that could be employed from Parallel and Distributed Simulation (PADS). The conceptualizations and discussions reflect on the authors' experience implementing an RtS of a network of Emergency Departments and Urgent Care Centers in the UK.

This paper focuses on the healthcare application field of Genomic Sequencing and addresses the challenge of efficient organization and ramp-up of sequencing platforms. High-throughput sequencing platforms are currently in an industrial prototyping phase in France for large national deployment afterwards. In the current state of our knowledge, there is no scientifically established generic model nor decision-making support at the operational level which could guide the medical authorities in designing organizational rules, then managing the deployment of such platforms at the national level. After analyzing the state of the art, a simulation model of a genome sequencing platform is presented, then used as a decision-making support to manage a ramp-up situation for an application case of a French sequencing platform. These first results are discussed, together with the perspective to develop a generic model and decision-aid approach.

Compartmental models have been utilized in the study and understanding of the COVID-19 pandemic. Traditional models have been expanded to include geographical level transmission dynamics and new states. Here, we present a model based on Cell-DEVS specifications that can be used to define and study the effects of basic human behavior. We include mask wearing and lockdown fatigue, and an adaptable framework allowing for the rapid prototyping of different diseases and behaviors. We exemplify how to build the model and adapt the attributes using the provinces of Canada as a case study. The results show the effect mask mandates, mask wearing, and lockdown fatigue have on case counts over time.

Poliomyelitis (polio) is an infectious disease that paralyzed millions of people worldwide before polio vaccines were available. Despite the successes of the Global Polio Eradication Initiative, there are circulating vaccine-derived poliovirus outbreaks that require improved interventions. We built a compartmental model to simulate the spread of polio that considers mutation of the live-attenuated virus (in the oral polio vaccine) to evaluate the effectiveness of interventions. We validated the model in a case study of northern Nigeria and tested the impact of interventions that varied in the number of vaccination rounds and the target regions. Results indicated that the model captures polio dynamics by matching the case counts and their spatiotemporal and age distributions in the data. To stop the outbreaks, stakeholders should conduct aggressive interventions with more rounds and broader coverage, especially in the under-vaccinated regions, compared to the current practice.

The Conceptual Modeling (CM) stage of an M&S study focuses on developing an abstraction of the real world for subsequent implementation as a computer model. Several studies have acknowledged the importance of CM in the success of simulation projects. Yet, there is a lack of literature on applying CM frameworks to real-world case studies, which arguably impedes the translation of CM research into practice. In this paper, we present the development of a conceptual model, using Robinson’s CM framework, for our case study investigating the perishable product of human milk within the milk banking supply chain. We present the application of the various stages of the framework, reporting on stakeholder engagement, which has allowed us to develop a shared view of the CM. The paper adds to the literature on CM in practice, providing a detailed narrative on developing a conceptual model for perishable inventory management.

Modelling clinical pathways from Electronic Health Records (EHRs) can optimize resources and improve patient care, but current methods for generating pathway models using clustering have limitations including scalability and fidelity of the clusters. We propose a novel pathway modelling approach using Maximum Likelihood (ML) data clustering on Markov chain representations of clinical pathways. Our method is calibrated to produce clusters with low inter-cluster variability across the pathways. We use machine learning with Stochastic Radial Basis Functions (SRBF) kernels for surrogate optimization to handle non-convexity and propose an incremental optimization method to improve scalability. We also define a methodology based on novel replayability scores to help analysts compare the fidelity of alternative clustering results. Results show that our ML method produces clusters that have higher fidelity in terms of replayability scores than k-means based clustering and in capturing queueing contention, which is important for bottleneck identification in healthcare.

The Covid-19 pandemic has placed extraordinary amounts of stress upon public hospitals globally. This paper describes a simulation model for estimating hospital bed demand based on generated scenarios. Statistical tools were also developed for generating these scenarios, in particular, for fitting distributions to patients' lengths-of-stay and for predicting the number of daily arrivals of Covid-19 patients. A web dashboard has been created for ease of use. The simulation model and statistical tools have been used to estimate Covid-related bed demand at an NHS hospital in the East of England.

Epidemiological models must be calibrated to ground truth for downstream tasks such as producing forward projections or running what-if scenarios. The meaning of calibration changes in case of a stochastic model since output from such a model is generally described via an ensemble or a distribution. Each member of the ensemble is usually mapped to a random number seed (explicitly or implicitly). With the goal of finding not only the input parameter settings but also the random seeds that are consistent with the ground truth, we propose a class of Gaussian process (GP) surrogates along with an optimization strategy based on Thompson sampling. This Trajectory Oriented Optimization (TOO) approach produces actual trajectories close to the empirical observations instead of a set of parameter settings where only the mean simulation behavior matches with the ground truth.

Buruli ulcer (BU) is a debilitating disease affecting the skin, soft tissue, and bone. It is the third most common mycobacterial disease in humans. The mode of transmission is not fully understood, posing challenges in prevention, and delayed diagnosis. One effective approach to promote early diagnosis and treatment is the utilization of community health volunteers (CHVs) for active case-finding. In this study, we developed an agent-based model to investigate the impact of CHVs in referring BU patients for treatment. We compared the effects of two strategies: offering self-referral alone versus self-referral combined with CHVs, on the early diagnosis and treatment of BU. Our findings confirm previous knowledge that integrating CHVs in active case-finding leads to earlier detection of BU cases, decreasing the number of individuals recovering with major disabilities.

Business-to-business distributors deem it necessary to inspect the quality of inbound deliveries to their distribution centers. This paper observes a company that experiences an inefficient quality inspection and wishes to improve the process. The broader product-receiving process is under-researched in warehousing literature but possesses similarities with manufacturing quality control. The paper aims to extend prediction-based quality inspection to the warehousing field. It applies a hybrid model, combining discrete-event simulation and machine learning multi-label classification to decrease the required inspection volume and evaluate its effects on the ability of the inspection and the workload and costs of distribution center operations. The results show that the inspection volume can drastically be decreased, reducing the workload and costs at the expense of the inspection capability of infrequently occurring delivery quality flaws in training data. The configuration of the classification model determines the degree of inspection volume reduction and wrongly predicted delivery quality flaws.

In today’s globally interconnected economy, transportation delays that impact a specific industry’s supply chain can quickly propagate to other industries, dramatically impacting inventory levels and economic production on the local, state, national, and global levels. This research proposes a hybrid System Dynamics and Input-Output simulation model that represents the impact of transportation delays on the flow of goods across industries and between geographic regions. The model is applied to a case study involving the port of Los Angeles to quantify the direct and indirect effects of a 30- and 60-day delay in container movement on gross output across the 55 major industries in the United States. The capability to predict the scope and scale of the economic impact resulting from various transportation delays provides decision makers the opportunity to conduct preliminary what-if analyses which can support the development of potential mitigation strategies before the actual shock occurs.

Counterfeiting, the production of imitation goods, is a critical threat in the Information and Communication Technology (ICT) manufacturing supply chain (SC). Countermeasures (CMs) are strategies to mitigate disruptions and enhance a SC. We present a novel hybrid approach for assessing and selecting CMs in ICT SCs. Our model incorporates insights from subject matter experts (SME), via Delphi elicitation, into the simulation. This technique is used to study SC resilience against disruptions caused by counterfeiting. ICT is an integral part of our daily lives and life-supporting systems, making resilience against such threats vital. Using performance criteria including system service levels, delivery time, and product quality, our findings show the importance of integrating expert knowledge in simulation and the effectiveness of certain CMs.

Conventional methods used to elicit risk-taking preferences have demonstrated significant disparities with real-world behaviours, compromising the validity of the data collected. Serious gaming (SG) provides a high potential to bridge this gap. This paper presents a serious game as a novel approach to elicit risk-preference in an industrial manufacturing context, focusing on the game-design and implementation using hybrid simulation modelling. The developed SG serves as a tool for conducting incentivized experiments aimed at assessing human behaviour towards risk, to inform policy recommendations. The game incorporates two influential factors in shaping risk-taking behaviour in a manufacturing environment, namely the social learning and production pressure, and use a variety of game mechanics to promote the players’ motivation and engagement. A usability study was conducted with 10 participants using the Usability Scale System (SUS), to identify problems in the usability of the game. Results have shown that our game has a good usability.

To gain economic competitive advantage from the closed loop supply chain (CLSC), firms must ensure that the cost of reconditioning products does not exceed the cost of purchasing new products. The uncertainties associated with product returns (i.e., product condition, quantity etc.) make it difficult for managers to efficiently allocate resources. This study develops and employs a hybrid simulation (HS) model as a decision support tool in a case study from industry. We demonstrate via our HS that the company could save significant money each quarter by converting their existing schedules from two shifts to single shifts and redistributing resources. Furthermore, we found maximizing the subprocess output doesn't necessarily reduce costs. The company's focus on output-oriented subprocess evaluation could impede cost-saving efforts. Future research will explore how the mix of new and returned items affects process yield, different resource configurations, prioritization of product types, and processing time disparities.

Factory simulation can guide leaner production operations and resilient supply chains by informing capital allocation and real-time decision-making. This is especially true for emerging production methods, like additive manufacturing (AM), where a lack of expertise and relative technological novelty make it difficult to quantitatively assess technology economics across applications. While reported cost models provide detailed analysis on the AM printing process, accurate modeling requires specific evaluation of process-level and production-level considerations that significantly impact factory dynamics and cost. Advances in factory simulation modeling therefore promise the development of comprehensive and actionable cost models. This paper reviews progress in simulation-based costing, hybrid simulation, and automated model generation, and proposes an integrated approach for cost modeling using an AM-based factory. We demonstrate the feasibility of this approach by simulating the production of two common AM part geometries, and evaluate the associated cost and time performances of different factory configurations.

In discrete-event simulation models, the way we establish the relationship between a real-world object and the model entity (a single indivisible object flowing through the model) is crucial to some classes of problems due to possible computational unfeasibility. In addition, the entity size also relates to results accuracy and simulation running time - a subject barely explored in the literature. In this paper, these questions were investigated through case studies which supported our initial hypothesis about the general relationships involved. Then, a simple algorithm was developed for correctly choosing the best entity size to provide the desired accuracy, measured as a discretization error, with promising results. The limitations of the algorithm are addressed and some directions for future research are pointed.

Hybrid modeling and simulation studies combine well-defined methods from other disciplines with a simulation technique. Especially in the area of output data analysis of simulation studies, there is great potential for hybrid approaches that incorporate methods from machine learning and AI. For their successful application, the analytical capabilities of machine learning and AI must be combined with the interpretive capabilities of humans. In most cases, this connection is achieved through visualizations. As methods become more complicated, the demands on visualizations are increasing. In this paper, we conduct a data farming study and delve into the analysis of the result data. In doing so, we uncover typical errors in visualizations making the interpretation and evaluation of the data difficult or misleading. We then apply the concepts of visual analytics to these visualizations and derive general guidelines to help simulation users to analyze their simulation studies and present results unambiguously and clearly.

Supervisory systems are used to and act when certain events are detected. Studying supervisory using formal Discrete Event Modelling & Simulation allows analyzing an application and then using the model to build the controllers. Supervisors can lead to a state space explosion if the model size increases, thus, reducing the state space complexity can expand the practicality of the model. We present a method based on Discrete Event System Specifications using an approximate method that reduces the state space complexity. The plant models and synthesized controllers can then be deployed on embedded hardware providing model continuity. We discuss the method and present a case study of a supervisory system.

Increased data availability has stimulated the interest in studying sports prediction problems via analytical approaches; in particular, with machine learning and simulation. We characterize several models that have been proposed in the literature, all of which suffer from the same drawback: they cannot incorporate rational decision-making and strategies from teams/players effectively. We tackle this issue by proposing hybrid simulation logic that incorporates teams as agents, generalizing the models/methodologies that have been proposed in the past. We perform a case study on the NBA with two goals: i) study the quality of predictions when using only one predictive variable, and ii) study how much historical data should be kept to maximize prediction accuracy. Results indicate that there is an optimal range of data quantity and that studying what data and variables to include is of extreme importance.

The ageing of the population has forced changes in many areas of social policy, including pension systems. Countries are reforming their retirement policies in such a way that the size of pension benefits depends on the total period of employment, contributions made, and life expectancy. Due to the fact that in these types of system, employment plays a significant role in the accumulation of pension capital, a gender pay gap translates into a gender pension gap. In this article, we propose a hybrid simulation model to analyze the impact of long-term economic and demographic changes on the level of pension benefits when a worker retires, with a special focus on gender wealth pension gaps. The model combines demographic simulation conducted using a systems dynamics approach with discrete stochastic simulation by means of which we model the employment history of men and women. The model uses data from Polish statistical databases.

Hybrid Simulation (HS) is the application of multiple simulation techniques, for example, Discrete-event, Agent-based and System Dynamics, in the context of a single simulation study. HS is a growing area of research; numerous papers have delved into conceptualizations, frameworks, and case studies applied to specific application domains. The focus of our paper is on the construction domain. Through a systematic methodology for literature assessment, it presents a synthesis of the existing literature, providing insights on the choice of simulation technique, the context of its application, and the level of implementation, among others. Through an in-depth review of 36 relevant papers published over the past two decades, we contribute to a comprehensive understanding of the current state-of-the-art in HS as applied to Construction. The results of our investigation underscore the immense potential of HS in construction, with broad applicability spanning diverse areas such as structural analysis and building performance evaluation.

Military workforce modelling is typically limited to either a series of analytic equations, or a simulation model. However, developing two such models in tandem has the benefit of cross-validation as well as the opportunity to explore problem space not easily accessed by a single approach. In particular, business rules for force employment are not easily described by closed-form equations while simulation models require exceedingly large computational resources to reach the asymptotic behaviour provided by analytic equations. This work leverages the benefits of both approaches to describe the population and career trends of technician individuals. As this career tends to have well defined training requirements, hence clear delineation between semi-functional apprentices and fully-functional journeymen, it is well suited to population modelling. Notional distributions for career parameters are assumed and the results for career progression and fleet readiness are compared.

Worldviews play a central role in M&S providing the basic constructs to describe simulation models. Three main worldviews have been defined: event scheduling, activity scanning, and process interaction (PI). The latter has been described in two flavors, one centered in the network of resources and other in the transitory transactions that flow in the network. In this paper we present a new M&S approach based on the πHYFLOW formalism that combines network and transaction PI, while keeping the support for modular and hierarchical models. We demonstrate πHYFLOW expressiveness by representing a hybrid production unit with a variable number of machines subjected to breakdowns. The hybrid model combines a fluid queue describing the work-in-progress, with discrete events modeling machines arrivals, departures, and breakdowns. Arrivals and departures of machines are achieved through modular communication, enabling model composition with other πHYFLOW components.

The reliability of just-in-sequence supply systems depends to a large extent on the efficiency of a supplier’s internal logistics distribution system. Thus, improving the logistics efficiency is a major objective for many suppliers in the automotive industry. In this paper, a discrete event simulation model is developed to evaluate the operational implications of different logistics strategies in just-in-sequence supply systems. Building upon the case of a major automotive supplier from Germany, the implications of various transportation resources and routing approaches are investigated and analyzed when it comes to the supply of components from an internal warehouse to the assembly lines. Experimental results show that the combined, load-carrier-specific use of forklifts, pallet trucks and tugger trains holds a high potential to achieve more efficient supply operations and meet different operational performance criteria such as downsizing the vehicle fleet, improving supply reliability and punctuality at the assembly lines, or minimizing warehouse traffic.

Warehouse systems are progressively shifting towards mixed fleet models where automated and manually operated vehicles work together sharing the same floorspace. This is posing communication and co-ordination challenges from both a design and an operational perspective. Mixed fleet co-ordination is particularly challenging from a traffic control viewpoint due to the erratic behavior that human drivers may exhibit. In this work, an optimisation framework that aims at selecting the optimal route among candidate ones in a mixed fleet warehouse environment is developed. More specifically, the foundational deterministic components of the framework are described and an interactive dashboard used for verification purposes is presented. The development work of the stochastic component and the simulator is still ongoing. Initial feedback based on virtual testing conducted by an industrial partner suggests that a static optimisation approach based on historical traffic information may not lead to optimal choices when the human behavior is neglected.

Advances in Autonomous Vehicle (AV) technology has fueled industry and research fields to dedicate significant effort to the study of the integration of AVs into the traffic network. This study focuses on the transition phase between all Human Driven Vehicles (HDVs) in the network to all AVs, where these different vehicle types coexist in a mixed traffic environment. This paper investigates the potential impacts of aggressive merging behaviors by human drivers on traffic performance in a mixed environment. For this, three vehicle types – AVs, HDVs, and Aggressive HDVs (AHDVs) are modeled in an open-source microscopic traffic simulation model, SUMO. In the developed simulation, the AHDVs are modeled to emulate aggressive merging behaviors in front of AVs at a merge section of a freeway exit ramp. Several experiments are used to study the impact of such behavior. Results show travel-time gains by AHDVs at the expense of AVs and HDVs.

Nowadays, vast amounts of data can be collected by sensors and used for data-driven approaches. Digital twins provide a framework to exploit these data for solving various issues. For many companies in the industrial sector, machine maintenance is a significant issue. Maintenance is essential for high overall equipment efficiency, but it can also be costly. Therefore, it should only be performed when necessary, based on the machine’s condition. Condition monitoring is used to assess a machine’s condition periodically, allowing for condition-based maintenance. In this paper, a simulation-based approach for parameter estimation is presented that contributes to condition-based maintenance. It introduces condition indicators for certain features of machines and demonstrates how to evaluate them using data farming, which employs simulation models as data generators. Additionally, the implementation of this approach in digital twins is discussed.

Simulation is a proven method in industry and research to constitute the basis for further decisions. Therefore, the credibility of its results is of major importance. Generally, simulation studies are guided by procedure models comprised of several phases with specific results. To assess the credibility, verification and validation (V&V) is used by applying V&V techniques to these phase results, which requires significant effort. Additionally, the amount of processed data increases and there is a growing desire for real-time-adjustable models, increasing the effort required for V&V while reducing the time available. One way to address these challenges is to automate V&V. For this purpose, the notions of automation and associated automation levels have to be transferred to the domain of V&V in order to assess and classify the automatability of individual V&V techniques. The effort for application of V&V techniques can be reduced while keeping or increasing the credibility of simulation.

Effective management of supply chain costing is crucial for decision-making during times of disruption. It provides accurate cost indicators, enabling organizations to adapt to the risks of disruptions and mitigate their adverse effects. Supply chain costing literature has shown that traditional cost accounting approaches are inadequate in addressing the dynamic and complex nature of supply chain performance and the nonlinear behavior of the involved processes. Consequently, this paper presents a simulation-based supply chain costing framework that integrates discrete event simulation and time-driven activity-based costing to explore the dynamics of management accounting tools in a real context with all their complexities and interdependencies. The framework will be applied to the logistics function of an automotive supply chain to demonstrate the applicability of a static versus a dynamic time-driven activity-based costing model, their suitability to reflect the real operational performance of the supply chain and suggest ways to improve it.

This paper addresses a multi-echelon inventory management problem with a complex network topology where deriving optimal ordering decisions is difficult. Deep reinforcement learning (DRL) has recently shown potential in solving such problems, while designing the neural networks in DRL remains a challenge. In order to address this, a DRL model is developed whose Q-network is based on radial basis functions. The approach can be more easily constructed compared to classic DRL models based on neural networks, thus alleviating the computational burden of hyperparameter tuning. Through a series of simulation experiments, the superior performance of this approach is demonstrated compared to the simple base-stock policy, producing a better policy in the multi-echelon system and competitive performance in the serial system where the base-stock policy is optimal. In addition, the approach outperforms current DRL approaches.

Container transports within ports are characterized by mostly manual trucks and many handling operations in relatively small areas. Accordingly, they incur a disproportionately large cost in maritime transport chains. One way to reduce these costs is to use automated trucks in a port-internal transport system. Such systems have only been used on terminals, but not within whole ports. Thus, it is important to determine the design parameters of such transport systems. Discrete-event simulation is particularly suitable for investigating planned systems and controls in logistics. However, the costs of such systems are usually neglected. Therefore, a simulation-based cost model is used in this study to determine the cost-effectiveness of automated trucking systems. It is shown which factors possess the greatest influence on the cost-effectiveness of port-internal container transports. Furthermore, it can be estimated for the first time which cost savings can be achieved by using automated trucks for port-internal container transports.

This paper proposes a large-scale logistics network simulation method to reduce package delivery delay and package loss caused by the sudden increase of package transportation demand during large-scale promotion activities such as 11.11 and 6.18. We develop a large-scale logistics network simulation software for a large logistics enterprise. According to its actual logistics network, we establish its equivalent virtual logistics network in the simulation software. Then we simulate and adjust the virtual logistics network in advance. At last we regulate the actual logistics network according to the virtual network. As a result, we reduce the transportation time, the transportation distance, and the transportation costs for the logistics enterprise. The simulation software can complete the simulation of 500 million package distribution of a month in less than 30 minutes on a personal computer.

COVID-19 resulted in some of the largest supply chain disruptions in recent history. To mitigate the impact of future disruptions, we propose an integrated hybrid simulation framework to couple nonstationary demand signals from an event like COVID-19 with a model of an end-to-end supply chain. We first create a system dynamics susceptible-infected-recovered (SIR) model, augmenting a classic epidemiological model to create a realistic portrayal of demand patterns for oxygen concentrators (OC). Informed by this granular demand signal, we then create a supply chain discrete event simulation model of OC sourcing, manufacturing, and distribution to test production augmentation policies to satisfy this increased demand. This model utilizes publicly available data, engineering teardowns of OCs, and a supply chain illumination to identify suppliers. Our findings indicate that this coupled approach can use realistic demand during a disruptive event to enable rapid recommendations of policies for increased supply chain resilience with controlled cost.

This paper presents a framework for integrating a discrete mode choice model with agent-based simulation. The integrated framework provides a more realistic representation of long-haul freight transport and is applied to the real-world scenarios of moving freight from ports to inland destinations via road, rail, and inland waterways. It incorporates a mode choice component that captures demand shifts between modes in response to different different policy and vehicle technology interventions. The objective is to investigate the financial and environmental impacts of introducing new vehicle technologies and associated energy sources under different future scenarios in a UK multimodal freight system.

This study aims at identifying the best strategy to temporarily store products within a buffer area in an Italian ceramic tile company. The storage policy is analyzed to maximize the storage capacity, facilitate operators' activities, and, consequently, improve the warehouse logistics performance. A discrete event simulation was conducted using Salabim, a Python based open-source software, in order to determine the best policy. We compare the performance of the current storage policy, based on technical production properties of products, and a newly proposed one, based on products' downstream destination. The results suggested that the proposed strategy significantly improves the performance of the buffer area management. The approach can be applied to different applications, contributing to the literature on simulation-based decision-making in material management. Furthermore, the study provides a functional case study showing the potential and achievable results of Salabim for modeling complex systems.

Production Planning and its parameterization is critical to fulfil customer demands and to successfully react on changes in high volatile markets. Therefore, demand updates should be considered to improve production planning. In this paper the performance of two production planning methods MRP (Material Requirements Planning) and RPS (Reorder Point System) are compared in a multi-item single stage system where customer orders are updated in a rolling horizon manner. Applying a simulation study, we investigate the performance of MRP and RPS for biased and unbiased forecast information and discuss the difference in the optimal planning parameters. The study shows that for a production system with underbooking and low demand uncertainty, RPS method is superior, in all other scenarios MRP outperforms RPS. For overbooking scenarios, the results show that MRP leads to overall cost improvements ranging from 8% to 30%.

This study aims to optimize the order fulfillment process in a Robotic Mobile Fulfilment System warehouse by improving the order batching and the pick order assignment in order-picking activities using a simulation approach. The order-to-station assignment considers the association between the new order and the in-progress order at the station instead of random assignment. The proposed model aims to maximize the total throughput, maximize the pile-on value, and minimize the required number of pods. The proposed model is compared with a baseline scenario. The result shows that the proposed model significantly decreases the number of required pods by 40%, increases the pile-on by 60%, and increases the throughput by 4%. This result proves that the proposed strategy can improve the efficiency of the order-picking process by ensuring every order and/or batch of orders always goes to the picking station with the most similar order.

At a time when supply chain disruptions are on the rise, supply chain managers are often overwhelmed by a simple question: How resilient is my supply chain and how can the status quo be improved? We present a case study of a manufacturing company in Central Europe that uses a two-step approach to help managers answer these questions. In the first stage, Key Resilience Areas (KRAs) are applied to transactional data to identify critical elements of the supply chain and their potential impacts. In the second stage, process mining is used to analyze the root causes of the identified impacts. In the case study, we reveal vulnerable locations and relevant product characteristics of the material flows of the company's inbound network, and process mining is used to analyze why, for example, a single sourcing strategy was chosen for a critical supplier.

Supply chains (SCs) collaborate in production and consumption across the world. SC management techniques attempt to optimize and balance supply chain operations. SC simulation can help support this by exploring “what-if” scenarios across key performance indicators, particularly when SCs are subject to potentially disruptive events. Resilience is the capacity for an enterprise to survive, adapt, and grow in the face of turbulent change. Change engenders SC vulnerabilities and management control attempts to create SC capabilities to address them. We are investigating the feasibility of creating a generic SC Simulation framework that represents sources of vulnerability and resilience and allows decision makers to explore potential capabilities to address them. This article reports progress on the first step of this study towards the creation of a conceptual model of SC resilience.

Developing a comprehensive model is an effective approach for gaining insight into and analyzing complex systems such as transportation yards. Following this approach, a data-driven agent-based simulation model has been developed for transportation yards at Amazon which captures the features and processes of the system. By simulating different scenarios and using simulation output performance indicators like yard/parking slip/dock door utilization, entry/exit gate queue, and late departure count, this model helps to identify potential bottlenecks, inefficiencies, and risks in the system. This information is used for strategic decision making and/or improving the system. Furthermore, the user can find ways to increase the yards’ daily maximum volume process capacities through multiple ‘what-if’ scenarios. This model is performed with mean absolute error (MAE) and root mean square error (RMSE) of 6% and 7% respectively. This paper presents the overview, current use cases and future works for improvement of the simulation model.

The recent advancement in hardware computation speed has allowed stochastic microscopic traffic simulators to be embedded in signal optimization systems. In this study, stochastic perturbation simulation approximations (SPSA), an efficient difference-typed gradient-based searching, has been applied in the signal solver of a signal optimization system due to (i) its lower required total number of replications and (ii) the capability to conduct variance reduction technique (VRT). The case study has shown that the objective value, in terms of road users’ delay, indeed improves over iterations. Since the gradient-based method may be trapped in the local optimal, this study has further applied the shotgun mechanism that allows better solutions in the subject stage to proceed to the next stage. By further offering the shotgun process, the quality of the solution can be further improved.

The proliferation of data collection from smart cities has resulted in an exponential growth in the volume of measurements available for analysis. However, collecting all parameters concurrently at the same location is not feasible due to the complex nature of the real world. We present an innovative methodology that enriches asynchronous time series data from a variety of sources to facilitate data enrichment and city-wide behavior simulation. A case study on OpenDataBCN attests to the efficacy of this approach via an XGBoost model, predicated on geographical coordinates and timestamp disparities. The consolidation of data from different sources improves the richness and granularity of information at disposal for analysis, thereby revealing previously hidden patterns and relationships, exhibiting new insights and underscoring the potential of this methodology for sustainable and efficient data enrichment processes as well as new possibilities for simulation based on smart city datasets.

The disruptions experienced by the last mile delivery processes during the SARS-CoV-2 pandemic have inevitably raised the dilemma of alternative last mile approaches in Urban Logistics (UL). Self-Collection Delivery Systems (SCDS) suppose an improvement for both courier companies and customers, providing flexibility of time-windows and reducing overall mileage, delivery time and, gas emissions. Drawing a distinction from previous works involving hybrid modeling for automated parcel lockers (APL) network design, this study integrates a System Dynamics Simulation Model (SDSM) to forecast e-commerce demand in Pamplona (Spain), and considers the scalability of the model for other cities. A bi-criteria Facility Location Problem (FLP) is proposed and solved with an ε-constraint method, where ε is defined as the level of coverage of the total demand, and four different cases of demand coverage are run. The simulation and demand forecast was carried out using Anylogic software, being CPLEX the optimization solver.

The growth of urban population and the proliferation of mobility options in big cities are adding to the complexity of comprehending how people move about and how efficiently they do it. Understanding how traffic patterns change throughout the day is essential for legislators, public administrations, and other stakeholders, as it has a direct impact on citizens' quality of life by, for instance, increasing greenhouse gas emissions and noise pollution. In this context, simulation becomes an essential tool for grasping the emerging dynamics of urban transportation, citizens' mobility patterns, and traffic flow bottlenecks. This work presents a complete data modelling pipeline for generating the population, network and transportation demand that is fed to a multi-modal traffic simulation of the city of Barcelona using MATSim and open-access statistical data sources. The model is calibrated, the results are obtained, and future applications of the developed tool are outlined.

Fulfillment centers in the E-commerce industry are highly complex systems that houses inventory and fulfill customer orders. One of the key processes at these centers involves translating customer demands into trucks and yard operations. Truck yards with operational issues can create delays in customer orders. In this paper, we show how a scalable cloud-based hybrid simulation model is used to improve yard operations, optimize flow and design, and forecast yard congestion. Cloud experimentation along with automated database connectivity allows any user to run simulation analyses to derive data driven operational decisions. We tested the model on two real world case studies, which results in cost savings for the organization. This paper also proposes a robust automated framework for setting simulation validation benchmarks and measuring model accuracy.

In road freight transport, booking unfavorable time windows (TW) through time window management systems (TWMS) for loading or unloading trucks at the loading dock often leads to avoidable long tours. Therefore, this paper investigates, based on an agent-based simulation framework, the efficiency gains and improvements in vehicle routing with TW constraints that can be achieved by a reallocation of unfavorable TWs using a one-sided Vickrey-Clarke-Groves mechanism. A branch-and-cut algorithm is used to evaluate the value of a TW in the context of a pickup and delivery problem with time windows and to generate a bid for the auction. A winner determination problem is solved for conducting the auction. We show that a reallocation of unfavorable TWs leads to distance savings for the considered tours of the auction winners of 13% on average. Further, we can show that the TWMS provider can benefit by operating the mechanism on an electronic marketplace.

Cross-docking is a warehousing strategy that (ideally) moves goods from inbound docks directly to outbound docks. In reality, goods often need to be temporarily stored. Cross-docking is typically set up as a block-stacking warehouse (BSW), where goods are stored directly on the ground. Autonomous mobile robots (AMRs) could significantly reduce BSW costs. To deploy AMR systems to BSWs, five interlaced decision problems, including the storage location assignment problem (SLAP), need to be solved. Because of the combinatorial complexity of BSWs, and the absence of pertinent use case data and fitting simulation software, this is a challenging task. This work seeks to alleviate these gaps by (1) extending SLAPStack, a fine-grained open-source BSW simulation framework to accommodate cross-docking, (2) providing CROSSStacks, a real-world cross-docking dataset, and (3) evaluating two dual command cycle SLAP strategies as of yet untested for BSWs. One of the approaches outperforms a naive cross-docking SLAP strategy.

In this paper, we explore the potential of multi-agent reinforcement learning (MARL) for managing the driving behavior of autonomous guided vehicles (AGVs) in production logistics environments with single-lane tracks, where deadlocks pose a significant challenge. We build upon previous work and adopt a MARL approach using the Proximal Policy Optimization (PPO) algorithm. We conduct a thorough hyperparameter search and investigate the impact of varying numbers of agents on the performance of the AGVs. Our results demonstrate the effectiveness of the MARL approach in addressing deadlocks and coordinating AGV behavior, as well as the scalability of the learned policy to different numbers of agents. The Bayesian optimization process and increased iteration count contribute to improved performance and more stable learning curves.

As the adoption of autonomous mobile robots in warehouses and other industrial environments continues to increase, there is a need for methods that can effectively dispatch robots to meet system demand. Real-time dispatching of autonomous mobile robots can be very complex, but simple rule-based methods are typically used for this task. In this paper, a reinforcement-learning-based dispatching method for intralogistics (RLDI) is proposed. RLDI is warehouse layout independent and takes into consideration task due dates and the travel distance. The algorithm is trained and tested in a simulation environment that represents a small warehouse. Monte Carlo simulation analysis is used to explore the capabilities and limitations of the established RLDI. The performance of the method is compared to the shortest distance dispatching rule in single and multi-agent environments under various levels of due date tightness. Experimental results demonstrate the potential for using reinforcement learning methods for warehouse dispatching.

Traffic simulators are used to generate data for learning in intelligent transportation systems (ITSs). A key question is to what extent their modelling assumptions affect the capabilities of ITSs to adapt to various scenarios when deployed in the real world. This work focuses on two simulators commonly used to train reinforcement learning (RL) agents for traffic applications, CityFlow and SUMO. A controlled virtual experiment varying driver behavior and simulation scale finds evidence against distributional equivalence in RL-relevant measures from these simulators, with the root mean squared error and KL divergence being significantly greater than 0 for all assessed measures. While granular real-world validation generally remains infeasible, these findings suggest that traffic simulators are not a deus ex machina for RL training: understanding the impacts of inter-simulator differences is necessary to train and deploy RL-based ITSs.

The p-hub median problems (pHMPs) are a well-researched topic within the fields of Operations Research and Industrial Engineering. These problems have been found to have a wide range of practical applications in various areas such as logistics, retailing, and Internet computing. These applications have made pHMPs an important area of study, leading to numerous research efforts aimed at solving different variations of the problem. This paper presents a simheuristic algorithm for solving the uncapacitated version of the pHMP with stochastic travel times. The proposed approach combines simulation with biased-randomized heuristics to generate high-quality solutions quickly. The proposed method is validated by testing it on huge benchmark instances, which include stochastic travel times. The results demonstrate the efficiency of the proposed approach for this particular problem variation. The simulation-optimization approach provides a promising solution to a practical problem that arises in many real-world applications.

Wind energy constitutes a main contributor to clean and renewable energy. While the offshore sector received much attention from research and industry, onshore wind farms still make up the largest share of installation projects. Thereby, onshore installations retain similar wind speed restrictions as their offshore counterparts but additionally introduce limits and wait time restrictions between installation operations. This article proposes extending a planning method initially designed for offshore wind farms to cover these additional requirements and proposes a simulation model capable of evaluating the resulting plans. The results show that the extended approach prevents violations of these requirements, mitigates the influence of weather forecast uncertainties, and provides efficient plans for installation operations.

Drone delivery is a popular logistics method for e-commerce businesses due to its efficiency and convenience, especially for last-mile delivery and emergency situations in areas with poor infrastructure. However, the uncertainty of customer demands can affect transportation costs in the long run, making it vital to design an effective delivery system. To tackle this issue, we propose a two-stage stochastic model that minimizes the sum of fixed and expected operating costs. The first stage minimizes the total cost of the delivery system, including the facilities fixed costs and expected operating costs, while the second stage arranges drones' routes according to simulated demands to estimate the minimal expected transportation cost and penalty cost. Since this stochastic programming has infinite scenarios, we deploy a sample average approximation method to estimate its bounds. Additionally, we use a heuristic simulation framework to find a satisfactory solution in an acceptable time.

Digital twins have become an important element in smart manufacturing. As any other product, digital twins also have a lifecycle, starting from specifying the requirements of the digital twins until their decommissioning. As part of the Manufacturing and Industry 4.0 track of the Winter Simulation Conference (WSC), the purpose of this panel is to discuss the state of the art in digital twins with a special emphasis on the operations and maintenance of manufacturing digital twins during their lifecycles. The panelists come from academia, industry, and government with experience in the digital-twin landscape of the manufacturing industry in the United States, Europe, and Asia. This paper provides a collection of the statements from each panelist with the objective of initiating a deeper discussion during the panel session and inspiring researchers in the simulation community with their perspectives on the use of digital twins for smart manufacturing.

To facilitate a rapid response to pandemic threats, this paper focuses on developing a mechanistic simulation model for in vitro transcription (IVT) process, a crucial step in mRNA vaccine manufacturing. To enhance production and support industry 4.0, this model is proposed to improve the prediction and analysis of IVT enzymatic reaction network. It incorporates a novel stochastic molecular reaction queueing network with a regulatory kinetic model characterizing the effect of bioprocess state variables on reaction rates. The empirical study demonstrates that the proposed model has a promising performance under different production conditions and it could offer potential improvements in mRNA product quality and yield.

We study a highly complex scheduling problem that requires the generation and optimization of production schedules for a multi-product biomanufacturing system with continuous and batch processes. There are two main objectives here; makespan and lateness, which are combined into a cost function that is a weighted sum. An additional complexity comes from long horizons considered (up to a full year), yielding problem instances with more than 200 jobs, each consisting of multiple tasks that must be executed in the factory. We investigate whether a rolling-horizon principle is more efficient than a global strategy. We evaluate how cost function weights for makespan and lateness should be set in a rolling-horizon approach where deadlines are used for subproblem definition. We show that the rolling-horizon strategy outperforms a global search, evaluated on problem instances of a real biomanufacturing system, and we show that this result generalizes to problem instances of a synthetic factory.

Data-Driven simulation models are valuable tools to improve the accuracy of the models and enable them to transition to real-time predictive analytics tools. Adding AI (Artificial Intelligence) and ML (Machine Learning) enables those model to provide feedback and real-time optimization in un-attended environment. This paper details the steps and benefits that were used to implement such system in a large filling and packaging manufacturing setting, from initial randomized models to full real-time digital twin systems. Final models were used to optimize (real-time and offline) changeover, CIP (Clean in Place), production, filling lines, and material handling.

Semiconductor manufacturing is a complex, costly process involving a long sequence of operations on limited, expensive equipment. Recent chip shortages and their impacts have highlighted the importance of semiconductors in the global supply chains and how reliant on those our daily lives are. Due to the investment cost, environmental impact, and time scale needed to build new factories, it is difficult to ramp up production when demand spikes. This work introduces a method to successfully learn to schedule a semiconductor manufacturing facility more efficiently using deep reinforcement and self-supervised learning. We propose the first adaptive scheduling approach to handle complex, continuous, stochastic, dynamic, modern semiconductor manufacturing models. Our method outperforms the traditional hierarchical dispatching strategies typically used in semiconductor manufacturing plants, substantially reducing each order’s tardiness and time until completion. Consequently, our method yields a better allocation of resources in the semiconductor manufacturing process.

In shipyards, newly supplied steel plates from steel-making companies are stored in steel stockyards until they are retrieved according to the pre-determined cutting schedule. Steel plates are grouped into lots, and all steel plates of the identical lot are retrieved and transported into the cutting workshop at the same time. In this study, we developed the two-stage stacking algorithm to minimize the workload of overhead cranes for the rehandling work in the retrieval process. In the proposed algorithm, a reinforcement learning-based agent which learns the stacking policy in the simulation environment determines the initial stacking location of the steel plates only considering the cutting schedule. After the initial arrangement of steel plates is created, steel plates are reshuffled using the simulated annealing considering both the cutting schedule and lot information.

This paper presents an approach that combines data-driven digital twins (DTs) and deep reinforcement learning (DRL) to address the challenges of online optimization of scheduling problems, focusing specifically on the classic job shop scheduling problem. Traditional approaches to solving such problems often encounter limitations in handling uncertainties and dynamic environments. In this study, we explore the integration of DTs and DRL to enhance decision-making in scheduling problems. We investigate the adaptability of a Graph Neural Network model within the DRL framework, enabling the agent to learn optimal scheduling policies through interactions with the DT. The potential of this convergence to tackle modern scheduling complexities offers insights into the future of operations management.

Accelerated with the developments in the context of Industry 4.0, a new trend has established itself in the manufacturing industry within the last two decades. Companies started to offer integrated solutions such as Product-Service Systems (PSS). While the provision of PSS enables benefits like business model innovation or strengthening competitiveness, the exploitation of these benefits depends heavily on the decisions in the operative service delivery planning. This, however, is a complex task due to the huge solution space. Analytical methods reach their limitations when trying to find the optimal solution. Though different optimization algorithms were elaborated for this problem, the evaluation of their solutions is overly simplified, and thus, their expressiveness for the uncertain and dynamic reality remains questionable. This paper addresses these issues by demonstrating the modeling of an adaptive simulation model that can be used to gain a realistic evaluation of operative service delivery plans in PSS.

The reduction of carbon dioxide emissions is a major goal of the European Union and energy storage is a core aspect to reach this goal. However, the production of lead-acid batteries is very energy consuming. Based on a case company production system and data, we develop a simulation model for the most energy-intensive lead-acid battery production steps, i.e., ripening and drying of lead plates. As both processes have some non-controllable stochastic aspects, the planned process times for both steps are a crucial factor for overall energy consumption. Too low or too high planned process times either lead to energy wasting for re-warm-up or to unnecessary energy consumption during processing. Simulation results reveal a significant energy reduction potential when optimizing planned process times, which increases when process uncertainty decreases. In addition, also the post-maturation and post-drying times are found to have a high influence on overall energy consumption.

Compared to other manufacturing industries, the shipbuilding industry has high uncertainties and volatility in resources such as manpower, space, and equipment. The labor-intensive, expansive yard spaces, and enclosed working areas of the shipbuilding industry make it difficult to aggregate and analyze data. The research effort presented in this extended abstract focuses on gathering production data using IoT technology and schedule simulation for the intent of reduction in uncertainty of project management. The gathered data from automated equipment can be employed to monitor production performance and conduct data-driven production management. It is possible to prevent from decreasing production performance and excluding input of batch production performance unrelated to actual work information. In addition, we use simulation to find the optimal solution for the purpose of load leveling in the process of establishing an LNG CCS manufacturing plan.

This paper considers the problem of real-time dispatching of a fleet of heterogeneous automated guided vehicles (AGVs) with battery constraints. The AGV fleet is heterogeneous in terms of material handling capabilities; some can tow loads, some can lift loads while others manipulate loads with the assistance of a robotic arm. Transport requests arrive in real-time and include a soft time window, with late delivery incurring tardiness costs. Transport requests need to be assigned to a capable AGV based on required material handling capabilities with the objective to minimize a weighted sum of tardiness costs of transport requests and travel costs of AGVs. In this paper, an AGV-specific linear dispatching rule (LDR) learning approach is proposed to assign AGVs to randomly arriving transport requests in real time over a finite horizon. The proposed approach is compared with a heuristic policy from practice by using real-world data provided by our industry partner.

The continued acceleration of e-commerce growth present a challenge for fulfillment centers to manage growing SKU counts and increased demand volatility while continuing to satisfy customer delivery expectations and maintain control over costs. Many fulfillment centers are turning to automated solutions such as Autonomous Mobile Robots in an effort to increase throughput and efficiency from existing facilities. AMRs move throughout the warehouse environment guidance-free and can be deployed bringing goods to person, bulk material movement and can work collaboratively with employees for picking applications. For warehouse operations management teams and AMR solution providers, identifying the optimum fleet size and deployment logic for current and projected demand is a crucial step in a successful adoption of this technology. Data-Driven modelling and simulation can be a useful asset when evaluating different solutions and requirements before installation as well as identifying opportunities for increased efficiency or expansion in existing operations.

In a flexible production line capable of producing various product types within a single assembly line, an efficient parts supply is critical. The kitting feeding policy, implemented in the flexible production line, aims to kit and supply the necessary parts to the production line without delay. This study investigates the kitting feeding operation for Samsung Electronics’ surface-mount device production line. To facilitate the timely supply of parts required for surface-mount device production, Samsung Electronics introduced a robotic mobile fulfillment system-based automated kitting system. This research proposes a sequential decision-making framework to address the kitting operation optimization problem, as well as a kitting scheduling algorithm within the proposed framework. A simulation environment has been implemented to verify the performance of the proposed framework and algorithm through a series of experiments. The experimental results indicate that the proposed framework enhances operational performance and maintains stability, even as the problem size expands.

This objective of this simulation was to estimate the impact of optimizing parts slotting on picking throughput within the existing Raymond Corporation lift truck manufacturing facility warehouse in Iowa. The simulation demonstrated that slotting can results in a 67.89% increase in picking throughput. This increase exceeded production requirements and eliminated the need to outsource picking.

Increasing product customization and a continual need for higher productivity has led to more complex automotive vehicles being built in more compressed spaces. The material delivery networks supporting these processes have also had to adapt to deliver a wider variety of parts in smaller packaging at an increasing frequency. The author will discuss the development and analysis of an automotive delivery network simulation with a focus on delivery times, the resources required, the data model used to drive the simulation and the analytical techniques used during the project. The presentation will also include a discussion on the model construction, the time required to construct the model, and the challenges encountered in the project.

JFE Steel Corporation has developed an ore logistics optimizer to reduce transportation costs. Because the Japanese steel industry imports large quantities of raw materials, the huge cost of ship freight and demurrage fees has become a problem. This work presents the ore carrier scheduler which was developed using metaheuristics methods to minimize logistics costs. A strategy of consolidating various iron ore brands at a junction spot that super-large carriers can enter is suggested. A digital twin that represents the stockyard in the steelworks is developed using a discrete simulator to verify the feasibility of operations, confirming the possibility of reducing costs by more than 10 % by utilizing this system.

The performance and reliability of production systems is greatly affected by sudden breakdowns. In order to avoid these unforeseen interruptions, predictive maintenance (PdM) systems are being widely used to predict failures and prevent outages by maintenance. The performance of PdM systems however depend heavily on precision and recall of prediction results. In the worst case, missing or false alarms can actually worsen the performance of an production system instead of improving it. We present a new procedural model which specifically focus on the imperfection of such PdM systems and estimate the impact of this unwanted property on the performance and economic aspects of a production system. The model is presented in all steps needed for implementation and evaluation and demonstrated in a realistic use case examining an interlinked production system with a simulation-based approach.

The concept of Digital Twins has gained significant attention in recent years due to its potential for improving the performance of production systems. One promising area for Digital Twins is Smart Maintenance, enabling the simulation of different strategies without disrupting operations in the real system. This study proposes a high-level framework to integrate Digital Twins to support Smart Maintenance data-driven decision making in production lines. We implement, then, a case study of a lab scale drone factory to demonstrate how the production line performance evaluation is made under different what-if maintenance scenarios. The effects of this Smart Maintenance decision analysis approach were evaluated according to Key Performance Indicators from literature. The identified contributions are: (i) Digital Twin demonstrator focused on smart maintenance; (ii) implementation of smart maintenance data-driven decision analysis concepts; (iii) design and evaluation of what-if maintenance scenarios.

Digital twin has emerged as a key technology in the era of smart manufacturing and holds significant potential for maintenance. However, gaps remain in understanding stakeholders' requirements and how this technology support maintenance-related decisions. This paper aims to identify stakeholders' requirements for digital twin implementation and examine the role of digital twin in supporting maintenance actions and decision-making process. Semi-structured interviews and a workshop involving manufacturing practitioners and researchers were conducted to attain these goals. Furthermore, an in-depth qualitative analysis of the interview data was carried out. The results shed light on the current state of digital twin adoption, implementation challenges, requirements, supported decisions and actions, and future demand characteristics. By integrating the findings from the literature review and interview analysis, this study outlines the requirements for the digital twins as expressed by industry stakeholders that will be used and tested in the drone factory digital twin model.

The management of production line is a challenging task due to the high level of uncertainty in demand, which can lead to unbalanced utilization of resources. This may result in a potential deterioration of management satisfaction in terms of cost-effectiveness. Therefore, it requires efficient tools to optimize resource utilization. With such inherent needs, this paper presents a simulation-based decision support framework for garments industries. The Discrete Event Simulation (DES) is used to model different scenarios for the operational processes. The procedure focuses on the line balancing technique, which aims to eliminate bottlenecks and optimize the production process by balancing the workload. The results of this study demonstrate the effectiveness of the line balancing technique in improving line efficiency, reducing the idle time of the operators, and increasing productivity. The simulation was developed using AnyLogic simulation software. The outcome of the process is thoroughly evaluated and justified using a case study.

Scheduling of flat block assembly in a shipyard is crucial for productivity performance due to the high level of workload. This problem is commonly known as the permutation flowshop scheduling problem (PFSP) in operation research, which has been extensively studied in various papers since the 1950s. However, existing solutions often involve simplifying real-world problems with certain assumptions, limiting their practical applicability. In recent times, constraint programming (CP) has emerged as a strong alternative to exact algorithms and has been successfully applied to various PFSP, addressing the limitations of exact algorithms. In light of this, our study proposes a two-step optimization process to overcome the existing limitations composed of a CP and discrete-event simulation(DES).

This paper proposes a digital twin architecture for a flow shop assembly line to maximize productivity and reduce quality costs. The proposed digital twin architecture consists of five major modules; Synchronization module to synchronize a real factory and the digital twin, Monitoring module to provide intuitive information visualization, Event calendar initialization module to initialize the factory state at any given time to the starting point of the CPS (Cyber-Physical System) simulation, CPS simulation module to identify potential production losses, and Decision-making module to take proactive actions to avoid anticipated production losses. The proposed digital twin architecture has been implemented for a home appliance factory of LG Electronics Co., Ltd. In South Korea, and shows significant improvements in terms of productivity, quality cost, and energy efficiency.

Researchers are investigating innovative techniques and tools to improve operational production planning, as manufacturing processes are increasingly influenced by new product demands, innovation, and cost-effectiveness. Backward-oriented discrete event simulation (SimBack) is one such tool that has shown great promise in this area. However, conducting multiple simulation runs for backward simulation can be time and resource-intensive, hampering its efficiency. To address this issue, this paper proposes an automated approach for executing and evaluating simulation experiments within the framework of backward-oriented discrete event simulation for scheduling and capacity planning. The authors illustrate their approach by applying it to a simulation model of the Semiconductor Manufacturing Testbed 2020 (SMT2020).

Simulation can reach computational limits, especially when running large-scale experiments. One possibility to counter this issue is distributed simulation. Recent developments in containerization and container orchestration technologies, such as Kubernetes, provide a stable and scalable infrastructure, that can serve distributed simulation. Although these solutions exist, applications within the simulation community remain scarce. Thus, in this paper, we present the general setup of such an infrastructure and discuss the application of an example case. Adding to the existing literature, we present our path forward and insights with different versions, as well as the efforts needed to construct similar implementations. As a result, we showcase the speed-up of simulation experimentation. We aim to provide a helpful foundation for others in our community to weigh the effort and benefit of such a system for their own projects.

This paper presents a case study demonstrating the application of LineLab, a mathematical production system modeling tool, to optimize production system configurations and the ramp-up trajectory for novel mass timber building modules. The modeling tool can efficiently co-optimize a large number of variables, such as machine count, work-in-progress (WIP) count, average wait times, and throughput, thus helping to narrow down a broad design space. Sidewalk Labs, a Google company, faced unique challenges related to new product development, high-mix production, and phased ramp-up. This case study highlights the use of this mathematical optimization tool, and its integration with other simulation methodologies, resulting in an optimized digital pipeline for modeling the production scale-up for mass timber buildings. The insights provided contribute to the advancement of production optimization techniques and their applications across various industries.

Supply chains face a myriad of adverse risks that impact their daily operations and make them vulnerable. In addition, supply chains continue to grow in size and complexity which further sophisticates the problem. Lack of a structured approach and limitations in existing risk management methods contribute towards effective mitigation strategies not being properly developed. In this paper, we develop a discrete event simulation modelling approach to quantify the performance and risk assessment of a manufacturing supply chain in Sweden which is under the impact of risks. This approach could support decision makers by prioritizing risks according to their performance impact and facilitating the development of mitigation strategies to enhance the resilience of the supply chain. The conceptual digital model can also be used to generate synthetic data to build an artificial intelligence-enhanced predictive demonstrator model to showcase capabilities for building data-driven resilience of the supply chain.

Contract manufacturers face challenges with short-term orders, cost pressures, and diverse customer requirements. Customer trends in the automotive industry intensify these challenges with reduced batch sizes and individual customization. Traditional analytic planning methods are insufficient for handling the complexity of modern manufacturing processes. Computational power alone cannot overcome this obstacle, careful modeling of production processes and resources is essential. Simulative approaches have been developed to address similar problems. In this use case, we aim to adapt and implement these approaches for a leading automotive contract manufacturer. A comprehensive assessment will then verify the adapted approach’s viability and potential.

In recent times, manufacturing industries and their related supply chains have faced growing internal and external pressures. Due to the complex nature of global supply chain networks and the increased frequency of disruptive events, there is a pressing need to implement digital tools to support these industries. Digital twins have gained significant interest from industry and research communities due to their ability to provide valuable services in the short term. While there have been many contributions on digital twin-based methodologies for system design and production planning and control, the use of digital twins in supply chain management still needs to be improved. This paper presents an overview of the existing contributions on digital twins for supply chains. Starting from a preliminary literature review on the topic, relevant works are selected and used to identify insights on the current development level and future research opportunities.

Production Planning and Control (PPC) plays a key role in stabilizing and improving manufacturing processes under external and internal uncertainties by providing transparency in the whole system. This study focuses on PPC with internal uncertainties such as losses of work-in-process products during a contact lens manufacturing process. Although such losses are expected, the yield rates are uncertain and vary at different production stages. A hybrid agent-based simulation (ABS) and discrete-event simulation (DES) approach was utilized to resemble the underlying dynamics of the manufacturing system with uncertain yield rates. The results of the simulation experiments demonstrated that a simple average yield approach for production planning would cause potential backlogs and extra holding costs for the excess inventory. The proposed hybrid simulation could be used to support the decision-making process on a weekly basis to help a production planning team make a schedule that would improve efficiency and customer satisfaction.

Within the area of manufacturing planning and control there is a long ongoing debate on when and if decisions should be integrated to a centralized model or split to separate planning levels. While a centralized monolithic model is capable of solving separate decisions simultaneously, a hierarchical approach offers more degrees of freedom since a local planner always has more accurate information. The focus of this paper is on the design and mathematical assumptions of optimization models for overtime and order release decisions in order to cope with different degree of demand and process uncertainty. We execute the optimal decisions within a simulation model of a multi-stage, multi-product stylized flow shop. Our results show that a fully centralized is outperformed by a hierarchical design and that planning order release quantities centrally in combination with flexible overtime planning yields the lowest costs for high process uncertainty on the shop floor.

The nesting problem in the shipbuilding industry calls for an increase in the utilization rates of plates and a decrease in the scrap ratio. To improve the efficiency of part nesting in ship design, this paper proposes an approach that uses a reinforcement learning algorithm to determine an efficient arrangement of parts. We frame the ship nesting problem as a Markov Decision Process (MDP) to apply the Proximal Policy Optimization (PPO) model, a reinforcement learning algorithm. A case study on a real-life nesting design is provided to validate and compare the proposed approach.

Airbus established a digitalization strategy to enhance logistics and production processes using model-based systems engineering, including material flow simulation. To store and reuse simulation model, holdup quality standards and support logistics planning, novel to the aerospace industry, a logistics simulation repository is being developed. This is supported by presenting ongoing simulation studies.

Service companies have the challenge to analyze a large number of documents in order to extract relevant information for decision making. Such analysis can be made automatically reducing drastically the time amount and human effort needed. However, the computer system must ensure that the analysis of each document will be completed within a specified period of time which depends on the type of the document. A real case study is presented in this paper where the objective is to propose a new scheduling model for a computer system with 6 servers with a total of 384 logical cores. The arrival of documents is aperiodic and the processing time stochastic though processing time estimation can be done based on the number of pages and the type of the document. A simulation model has been developed to analyze the quality of each algorithm. A delay maximum time (DMT) algorithm is also proposed.

Canned water is one of the thriving markets in the food and beverage industry. Given the tight competition in this market, realistic analysis in such production lines has become even more attractive for all participating parties. In this paper, we apply a KPI-driven simulation-based approach to a smart production plant of a key player in the European beverage market. The project covers realistic discrete-event modeling and analysis of the system together with the suggested scenario-based optimization for performance improvement. Here, the smart line management system is modeled and re-coded while considering machine characteristics, failures, and their overall influence on the production process. Our proposed optimized scenario demonstrates noticeably better results in all performance indicators when compared to the existing state of the system. The total increment of the production speed reaches up to 45 percent, resource utilization is evenly optimal, and the overall work-in-progress inventory is reduced significantly.

The upcoming decentralized production systems seem to be promising in Industry 4.0 assembly to handle the challenges of highly individual products. Matrix production characterized by freely linked workstations and an advanced automation level are highly flexible. That is why many efforts have already been made to explore the advantages compared to existing flow shop production systems, but also the additional challenges arising from this new paradigm. One of these challenges is the synchronization of main product and supply part flow at the individual workstations during order scheduling. This paper presents a new approach of integrating logistics support processes into the scheduling of the main product flow to consider the part supply in the decisions taken during scheduling avoiding waiting times. We compare our integrated approach with the existing decoupled scheduling approach, based on a “bicycle assembly” scenario. The results are promising particularly when part supply is a bottleneck.

A Reinforcement Learning (RL) model is applied for photolithography schedules with direct consideration of reentrant visits. The photolithography process is mainly regarded as a bottleneck process in semiconductor manufacturing, and improving its schedules would result in better performances. Most RL-based research do not consider revisits directly or guarantee convergence. A simplified discrete event simulation model of a fabrication facility is built, and a tabular Q-learning agent is embedded into the model to learn through scheduling. The learning environment considers states and actions consisting of information on reentrant flows. The agent dynamically chooses one rule from a pre-defined rule set to dispatch lots. The set includes the earliest stage first, the latest stage first, and 8 more composite rules. Finally, the proposed RL approach is compared with 7 single and 8 hybrid rules. The method presents a validated approach in terms of overall average cycle times.

Scheduling the diffusion area in a front-end wafer fab poses challenges. This industrial case focuses on scheduling diffusion at Renesas’ Palm Bay Fab, which is always seeking scheduling system improvements. Transitioning to an advanced system, considering fab-wide impacts on diffusion batching, enhances Key Performance Indicators (KPIs). Our A.I. scheduler utilizes optimization, heuristics, and live data updates every five minutes. Collaborating with FabTime integrates the scheduler with the fab’s MES, ensuring frequent updates. It optimizes batching, tool allocation, and launch times, aligning with Renesas’ objective to balance competing goals. Initial results show a 36% and 13% increase in diffusion batch sizes at clean and expensive furnace toolsets. The minor impact on cycle time reflects the scheduler’s focus on batching efficiency. This approach improves efficiency and meets Renesas’ goals, marking a positive step in optimizing their wafer fab operations.

Digital twin-based Production Scheduling (DTPS) is a process in which a digital model replicates a manufacturing system, known as a “Digital Twin (DT)”. DT is essentially a virtual representation of physical equipment and processes that are connected to the physical environment using an online data-sharing infrastructure within the Manufacturing Execution System (MES). In the case of reactive scheduling, DT is used to detect fluctuations in the scheduling plan and execute rescheduling plans. In proactive scheduling, it is used to simulate different production scenarios and optimize future states of production operations. Replicating detailed simulation models in most PS cases is highly computationally intensive, which negates against the main goal of DT (online decision making). Thus, this research aims to examine the possibility of using data-driven models within the DT of a Flexible Job Shop (FJS) production environment aiming to provide online estimations of PS metrics enabling DT-based reactive/proactive scheduling.

Impact of timelinks to semiconductor manufacturing has risen due to shrinking technology sizes. Their operational control defines on the one hand how good the time restrictions are met and on the other the impact to fab capacity. This paper discusses both aspects and the influencing factors like uptime stability, length of the timelink etc. A control approach is proposed, evaluated, and discussed. Furthermore, a monitoring system is introduced that enables for fast decision making and optimization of the control parameters. Finally, a simulation study is done for evaluating different parameters and impact of influencing factors.

This paper presents a methodology for predicting queue times in semiconductor fabrication, where numerous complex and costly pieces of equipment are utilized. Queue time, occurring between continuous single or multi-processes, is a crucial factor affecting the quality of wafers, which can significantly impact costs. While most semiconductor fabrications use queue time limits as a key dispatching factor, some wafers may still be scrapped or reworked. By predicting queue times, we can reduce unnecessary waste by blocking or re-dispatching wafers. Two approximations are proposed and compared based on accuracy and prediction time: a machine learning model trained using experimental results and a multi-resolution simulation model with varying fidelity levels. The simulation model is validated using the SMAT2022 data set.

In partnership with GlobalFoundries we have significantly advanced Processing Time (PT) and machine availability prediction in fabrication plants, utilizing an attention based neural network. This model is integrated into an MLOps pipeline consisting of data collection, preprocessing, training and deployment. The data is augmented with features such as chamber usage and process sequences. Compared to the current model, which calculates average processing times over a predefined context, our approach has reduced the Mean Absolute Error (MAE) of PT predictions by 43% to 80% across the crucial areas: Etch, Diffusion, and Deposition. The model also produces high quality predictions for the remaining tools. The model is in the process of being implemented in the FAB to improve scheduling, dispatching, and improve crucial Key Performance Indicators (KPIs) such as cycle time and throughput.

Access to more information and sophisticated analytics enables warehouse management to make better data-driven decisions. In our study, we develop a simulation-regression metamodel to help warehouse managers plan workforce, space, and equipment requirements for a leading semiconductor capital equipment company. More specifically, we use historical inbound and outbound demand records and performance metrics (such as workers’ hourly productivity and moving rates) to predict the space, workforce, and equipment required for different operation stages in the warehouse facility. We implement the simulation model in Python. Simulation experiments provide insights on resource planning under different demand scenarios and supply constraints.

Ensuring reliable and timely customer deliveries is crucial to supply chain management. The ability to meet delivery commitments is essential for maintaining customer satisfaction. Despite the importance of delivery commitments, there is a lack of standard measurement techniques for evaluating their quality. Therefore, this paper introduces the term Commitment Quality (CQ) and develops a CQ matrix that can be used to measure the quality of delivery commitments. The CQ matrix provides a comprehensive set of quantitative measures to evaluate different aspects of delivery commitments. Finally, a numerical example based on an order data sample from a semiconductor manufacturer is presented and discussed. The proposed framework aims to standardize the CQ, enhancing transparency in delivery commitments.

The bullwhip effect (BWE), a well-known phenomenon in supply chain management since it was first identified in 1958, is causing significant economic damage after disruptions. While the role of human factors in BWE has been widely recognized, the impact of different replenishment policies on BWE mitigation has not been thoroughly investigated. This paper presents a study on the impact of reach-based Kanban systems on the BWE in supply chains containing suppliers with intrinsically non-reducible long cycle times, such as those in the semiconductor industry. Our findings suggest that a reach-based replenishment system acts as a BWE accelerator after significant disruptions, which can end up in line-downs downstream. We propose a change to absolute stock targets for replenishment policies during disruption to mitigate this aspect of the BWE root cause for supply chain with long cycle time suppliers to reduce the risk of line downs.

Competitiveness in the semiconductor industry requires continuous management of product rollovers, the process of introducing new products and retiring older ones to maintain market share. This paper presents a decentralized decision-making framework to coordinate product rollover decisions using Lagrangian decomposition of a centralized model using quadratic coordination errors in the subproblem objectives, and a decentralized heuristic that recovers the feasible solutions from the relaxed ones obtained from the Lagrangian procedure. Experimental results show that this decentralized framework delivers promising results, obtaining near-optimal solutions in modest CPU times.

Data-driven (DD) production planning formulations for semiconductor wafer fabrication facilities (wafer fabs) are studied in this paper. These formulations are based on a set of system states representing the congestion behavior of the wafer fab with work in process and resulting output levels. We establish two DD formulations with inventory considerations. The first variant is a shortfall-based chance-constrained formulation that considers safety stocks at the finished goods inventory level. The second variant is a simple scenario-based stochastic program where the objective function reflects the expected inventory holding and backlog cost under uncertainty. The two variants are compared with the conventional DD formulation in a rolling horizon environment using a simulation model of a large-scaled wafer fab. The simulation experiments demonstrate that the stochastic program achieves the largest profit under all experimental conditions.

The design and the implementation of a multi-agent system (MAS) for a borderless fab scenario is described. In such a scenario, lots are transferred from one wafer fab to a nearby one to perform process steps of the transferred lots. Production planning is carried out individually for each of the wafer fabs. The modeling of the available and requested capacity in the production planning models of the participating wafer fabs is affected by the lot transfer. The transfer of the route information from one wafer fab to another to automatically generate the linear programming models is described. Production planning is carried out in a rolling horizon setting using a cloud-based infrastructure. We show by simulation experiments with the MAS with a correct modeling of the capacity in production planning results in improved profit compared to a setting where the lot transfer is not taken into account in the planning formulations.

An OEM who maintains a fleet of complex systems strives for high system availability for its customers. Frequently failing components lead to system unavailability and high maintenance costs. Consequently, the OEM might decide to upgrade components. We develop a model that quantifies the impact of the introduction of an upgraded component on the OEM's costs and number of failures to define the best implementation strategy. Using a Markov process, we evaluate four policies differing in the roll-out strategy of new parts, either immediate or corrective, and the phase-out strategy of old parts, either rework or salvage. The model is used in a case study at ASML. We conclude that, in the case study, reworking is preferred over salvaging as the phase-out strategy and corrective replacements are generally preferred over immediate replacements for the roll-out strategy.

We study an order-lot pegging problem in semiconductor supply chains. The problem deals with assigning already released lots to orders and with planning wafer releases to fulfill orders if there are not enough lots in the wafer fabs. The objectives are minimizing the total tardiness of the orders and minimizing the total cost. We are interested in computing the set of Pareto-optimal plans. Based on a mixed-integer linear formulation, a ϵ-constraint method is proposed for small-sized problem instances. Moreover, a non-dominated sorting genetic algorithm (NSGA)-II algorithm is designed for tackling larger problem instances within a reasonable amount of computing time. We perform computational experiments with the ε-constraint method for small-sized problem instances and with the NSGA-II scheme for small- and medium-sized problem instances.

This case study presents a novel approach for modeling a fab, which allows for more rapid results than traditional simulation, while optimizing various variables like tool count or throughput, and capturing equipment sharing between co-produced devices. This modeling method was applied at InchFab, a foundry that uses ultra-small substrate sizes to allow for more flexibility and lower costs when fabricating small production quantities. The new approach was used to find the cost-optimal rate achievable for a primary product on certain tool counts - and then the cost-optimal rate of a secondary product, without any changes to equipment count. Using novel types of analyses and sensitivity figures, we demonstrate that it can be economically sensible to add a product to a fab that is already producing the cost-optimal quantity of a base product. This is an important finding, as some fabs consider offering additional foundry services on existing equipment.

In the semiconductor industry, many retrofits are being made to improve the production efficiency of manufacturing facilities. However, due to the nature of the data provided by the cluster tool, which is a semiconductor manufacturing facility, engineers have some limitations in utilizing it. To address this issue, it is necessary to introduce a digital twin model that can verify the performance of the semiconductor process cluster tool in a virtual environment, and to apply optimal mass production conditions based on this predictive data in the operational stage. In this study, we propose a digital twin model that visualize congestion factors during wafer transfer and evaluate the productivity of cluster tools.

This paper presents a study on the priority mix planning problem in semiconductor fabrication using simulation. The objective of the study is to analyze the impact of the mix of different lot types associated with their priority on the cycle time of the Implantation workshop. We have specifically analyzed the waiting time lots and the associated speed up or speed down on a work-center. The tests were conducted using Anylogic 8 on industrial instances from STMicroelectronics Crolles. Results shows that a speedup of more than 300% for high priority lots and speed down of less than 10% is possible if the proportion high priority lots is kept under 10%. This study initiates a first step toward a better priority mix management which has a strategic central place of in the semiconductor industry.

In modern manufacturing environments, the digital transformation to smart factories cannot be achieved without data-driven methods like discrete, event-driven simulation. This paper provides an overview of existing current simulation applications at Infineon Dresden in this area, especially on short-term simulation for production control and long-term simulations to forecast process flows in the wafer fabrication facilities. Furthermore, it illustrates the current status of research activities in the area of backward simulation for operational decision support for order scheduling by some latest research results.

Duplicate reticles provide a fab with an opportunity to mitigate the impact of catastrophic reticle damage or the need for offsite repair/cleaning and provide the necessary capacity for products in a high volume manufacturing environment. Implementation of a management system for duplicate reticles helps to maintain a minimum number of run paths while ensuring availability of multiple reticles to process lots simultaneously. Dedicating the duplicate reticles each to a group of exposure tools prevents duplicate reticles from ending up on the same exposure tool, and managing this dedication by tool/reticle inhibits has proven to be an effective method of distributing the WIP between the exposure tools while minimizing the management of the layer supported by those duplicate reticles.

Digital factories have been recognized as a paradigm with considerable promise for improving manufacturing performance. Digital Twins have emerged as a powerful tool to improve control performance for large-scale smart manufacturing systems. We argue that DT-based smart factories are vulnerable to attacks that use the DT to damage the system while remaining undetectable, specifically in high-cost processes, where DT technologies are more likely to be deployed. As an instructive example, we consider smart semiconductor processes with focus on photolithography. To this end, we formulate a static optimization problem to maximize the damage of a cyber-attack against a photolithography digital twin that minimizes detectability to the process controller. Results demonstrate that this problem formulation provides attack policies that successfully reduce the throughput of the system at trade off of increased detectability to a common process control technique. Results encourage more research in the domain, especially to face scalability and policy-like solutions.

As simulating semiconductor manufacturing grows complex, model reuse becomes appealing since it can reduce the time incurred in developing future models. Also, considering a large network of the semiconductor supply chain, knowledge sharing can enable the efficient development of simulation models in a collaborative organization. Such necessity of reusability and interoperability of simulation models motivates this paper. We will address these challenges through ontological modeling and linking of the simulation components. The first application is generating reusable ontologies from simulation models. Another discussed application is ontology matching for knowledge sharing between simulation components and a meta-model of the semiconductor supply chain. The proposed approach succeeds in automatically transforming simulation into reusable knowledge and identifying interconnection in a semiconductor manufacturing system.

This paper surveys the industrialization of an advanced optimization engine that was developed by Planimize and put into production in the cleaning and diffusion work center of the most advanced factory of a semiconductor manufacturing company. Hundreds of lots requiring several thousands operations in the work center must be scheduled on about 150 machines, while taking complex constraints into account, in particular hundreds of time constraints, and optimizing a collection of criteria. The optimization engine provides significantly better results, runs significantly faster, and can handle much larger problem instances than the previous Constraint Programming optimization engine used in the factory.

The scheduling problem we study in this paper is known as a multiple orders per job (MOJ) (Mason et al. 2004) problem which is encountered in a few different industries including front-end semiconductor manufacturing. We look at the MOJ scheduling problem in a two-stage permutation flowshop with some real-world constraints with the goal of minimizing the makespan. We use a MIP solver and various heuristics to solve this NP-hard scheduling problem for various stage configurations and bottleneck types. For moj(ipm-ipm) the makespan was minimized by the MIP solver regardless of the bottleneck type for over 90% of the small-sized problem instances. When the heuristics minimized the makespan, the Slope heuristic was the fastest NEH heuristic was the slowest for over 90% of the large-sized problem instances.

Dispatch rules are commonly used to schedule lots in the semiconductor industry. Previous studies have indicated that adapting dispatch rules can improve overall factory performance. Machine learning has proven useful in learning the relationship between manufacturing situations and dispatch rules. However, using only snapshot data at a given point in time to generate features for these models does not account for trends in the manufacturing situation, which can be represented as time series data. To address this issue, the proposed method generates features from time series data and combines them with features from snapshot data to train machine learning models for dispatch rule prediction. The results demonstrate the effectiveness of this methodology, as the combination of features from both types of data achieves the highest prediction accuracy. Simulation results show that this approach can adapt the dispatch rule according to the manufacturing situation and achieve a comparable factory performance.

Semiconductor manufacturing and supply chain processes are one of the most complex but can be considered at the same time also as one of the most rewarding processes in the world. In thousands of detailed unit chemical and physical processes in cleanrooms and under statistical process control chips on wafers emerge and are assembled and tested to components. The Modeling and Analysis of Semiconductor Manufacturing (MASM) conference embedded in the annual Winter Simulation Conference (WSC) was, is, and will be key to understand the optimization and simulation challenges in this domain.

The operating curve management targeting a low variability value and thus enabling a low flow factor - thus speed - and high utilization - thus a good cost position - at the same time has been an early achievement. With discrete-event, agent based, and system dynamic simulations on the four levels (machine, fab, internal and external supply chain) solution options for complex interactions could be proposed based on sophisticated mathematical models running on simulation testbeds like the MIMAC models and their successors. Accurate planning and advanced scheduling, available to promise (ATP) generation and usage with traditional or artificial intelligence (AI) / deep learning (DL) methods requires a huge amount of real data or qualified synthetic data (QSD).

The semantic web for semiconductor and supply chain containing semiconductors bears the potential to enable the provision of those urgently needed QSD in volume, (integrated) complexity and accuracy needed. Quantum bit (qubit) based algorithm could provide the speed for the next and over-next generation for optimization and simulation in our domain.

This panel assembles a number of prominent representatives from industry and academia to discuss how semiconductor supply chains in times of increasing geopolitical risks can be made more resilient through Modeling and Analysis of Semiconductor Manufacturing (MASM) techniques and enabling software solutions.

Monitoring the state of semiconductor equipment is crucial for ensuring optimal performance and preventing downtime. In previous studies, researchers have attempted to derive a health index that represents the overall condition of the equipment as a single index. However, these studies have often relied solely on time-series data from each sensor, neglecting other important viewpoints engineers consider when monitoring the equipment. To address this limitation, we propose a multi-view data set specifically designed for semiconductor equipment, which incorporates process, trend, and spatial data. In addition, we present a framework for deriving a hierarchical health index based on a multi-view data set. The hierarchical structure is derived using a hierarchical spectral clustering method, and an autoencoder-based health index is used. We have verified the effectiveness of our approach with real data sets, demonstrating its potential as a valuable tool for monitoring the condition of semiconductor equipment.

The complexity of manufacturing process data has made it more challenging to extract useful insights. Data-analytic solutions have therefore become essential for analyzing and optimizing manufacturing processes. Path modeling, also known as structural equation modeling, is a statistical approach that can provide new insights into complex multivariate relationships between process variables from different stages of the manufacturing process. The incorporation of expert process knowledge and subsequent interpretation of model results can facilitate communication between stakeholders, promoting lean manufacturing and achieving the sustainability goals of Industry 5.0. This paper describes the use of a path modeling algorithm called Process Partial Least Squares (Process PLS) to gain new insights into the relationships between equipment data from several machines within the semiconductor manufacturing process. The methods used in this study can assist manufacturers in understanding the relations between different machines and identify the most influential variables that may be used to develop soft-sensors.

This paper presents a novel modeling framework for semiconductor fabrication facilities (FABs) that integrates production and material handling systems. Because the productivity of semiconductor FABs is significantly influenced by their material-handling systems, existing research has focused on optimizing operational logic considering both aspects. However, the scale and complexity of modern FABs make implementation of fully integrated models challenging, resulting in slow simulation speeds for long periods. To address this issue, we propose a multi-resolution modeling framework that creates material-handling system models at two distinct resolution levels, enabling fast, fully integrated FAB models while accounting for material-handling effects. Experimental results demonstrated accelerated simulation completion compared to single-resolution models while maintaining consistent results. The proposed method provides a practical approach for semiconductor FABs to investigate long-term phenomena and urgent decision-making problems while considering both production and material-handling systems.

This study presents a self-supervised learning based framework for TFT-LCD defect classification in semiconductor smart manufacturing. Utilizing the Swapping Assignments between Views (SwAV) model trained on 1,000,000 unlabeled TFT-LCD images, the framework achieves an overall top-1 accuracy of 0.709 and precision of 0.7812 in downstream task of classifying 13 types of TFT-LCD defects. Compared to using SwAV pre-trained weighs on ImageNet, proposed domain-specific self-supervised learning model significantly outperforms, emphasizing the importance of domain-specific training. The framework offers manufacturers a cost-efficient decision support system, enhancing TFT-LCD defect classification quality.

In the highly dynamic world of semiconductor manufacturing, planning analysts are asked to analyze variations between weekly production plans with the goal of identifying a resolution in a landscape involving elaborate optimization models with significant interdependence between data elements. We propose a solution to effectively analyze the weekly planning engine output and identify the data elements with significant contribution to the outcome. An explainable Machine Learning model is trained and deployed to simulate the behavior of the planning engine. Each model execution can be explained to identify the features with the most significant contribution to prediction. The resulting application contributes to a timely resolution to the production plan deviation, while generating significant productivity gains.

Queue-Time Constraints (QTCs) set a maximum waiting time for lots between consecutive process steps. In semiconductor manufacturing, exceeding these limits results in yield loss, rework, or scrapping. Managing QTCs is challenging due to the need for lots to wait until there is available capacity for the final step. Specifically, accurately calculating the capacity is computationally expensive, making it difficult to handle large instances. Our research addresses the scalability of QTC management in real fabs with numerous constraints. We propose a deep Reinforcement Learning (RL) solution to handle lot release into the QTC. We describe the infrastructure developed for RL training using actual fab data, assess the performance of our RL approach, and compare it to three baseline solutions. Our empirical evaluation demonstrates that the RL method surpasses the baselines in key performance metrics including queue-time violations, while requiring negligible online compute time.

Live Virtual Constructive (LVC) exercises are becoming ubiquitous for training and mission rehearsal in the military domain because the use of LVC provides the most realistic environment available short of actual military operations. The mixture of live exercises with simulated components (constructive simulations and virtual simulators) allows for the creation of a context for the training or rehearsal that is richer and more representative of the real world than would be possible with only live events. This ability to embed live activity into synthetic environment to provide realism has attracted the interest of the Test and Evaluation community (T&E) and recently there are increasing efforts to start including LVC in the T&E tool suite. This talk will discuss some of the work we have been doing at the Naval Postgraduate School with LVC and how these environments can be used to assess and improve system and human resilience in operational environments.

This paper emphasizes the integration of sound tactical behavior in the generation of realistic military simulations, which includes the definition of combat tactics, doctrine, rules of engagement, and concepts of operations. Recent advances in reinforcement learning (RL) enable RL agents to generate a wide range of tactical actions. A multi-agent ground combat scenario is used in this paper to demonstrate how a machine learning (ML) application generates strategies and issues commands while following a given objective. Natural language is used to issue doctrines and objectives to improve communication between the human advisor and the ML agent. This allows us to embed objectives and existing doctrines into the reasoning of an artificial intelligence (AI). The research demonstrates the successful integration of natural language to enable an agent to achieve different objectives. This groundwork will enhance RL agents' ability in the future to uphold the doctrines and rules of military operations.

There is significant variation in shooting ability among U.S. Army soldiers, which is often overlooked in combat simulations. This study introduces a Monte-Carlo model to estimate the dispersion of a soldier's shot group based on their marksmanship score. This model is used to assess the impact of marksmanship on a squad's performance through two analyses. The first analysis employs a dueling model to examine various marksmanship skills between dueling teams, offering insights into overmatch requirements. The second analysis uses an agent-based combat simulation to investigate the influence of marksmanship on squad performance in a dueling scenario in addition to tactical rural and urban missions. The results reveal that marksmanship becomes increasingly crucial in enhancing lethality and survivability as the distance between combatants grows. Notably, superior marksmanship skills are particularly vital in offensive, rural operations. These findings emphasize the significance of marksmanship and its implications for military requirements and tactical decision-making.

The escalation of the Russian invasion in Ukraine, characterized by the deployment of conventional weapon systems, inflicts significant morbidity and mortality on the victims. It is imperative to ascertain optimal medical practices and disaster response strategies throughout the battlefield to minimize casualties and safeguard the well-being of medical and disaster responders. The challenges posed by large-scale battlefield threats can rapidly overwhelm healthcare providers due to the sheer number of victims, which can result in the depletion of medical supplies and insufficient training and resources. To address these issues, we utilized the SIMEDIS simulator to establish and implement a battlefield scenario involving an open-air artillery strike in a field. Mortality rates were calculated based on the application of bleeding control measures and the distribution policy for allocating victims to medical treatment facilities. Controlling hemorrhage remains the most crucial factor influencing mortality outcomes.

During military conflicts, the number of casualties is likely to exceed medical capabilities. For best treatment results, the patients must be distributed according to their needs to the available resources such as medical facilities and means of transportation. Computer simulations are used to verify and optimize current medical planning. However, recent models lack the capability of testing a wide range of decision rules. In this paper, we address this issue and propose a modular simulation concept whose components can be adapted and exchanged independently. Using modular submodels to control the simulated objects, we enable the implementation of a wide range of object behavior. A prototype implementation of the proposed concept is presented, showing the effects of applying different dispatching rules in an evacuation scenario.

Simulating pre-planned routes and dynamic threat avoidance routing represents a significant problem for operations analysts. Without methods to create operationally valid routes through automation, the analyst is generally faced with hard coding individual routes for multiple aircraft over the entirety of the mission set. This research developed, implemented, and analyzed threat avoidance routing based on Dijkstra's algorithm for aircraft attempting to operate in an anti-access area denial (A2AD) environment capable of dynamically updating the mission route as new threat information is learned. A designed experiment was conducted to determine the impact of grid parameters on operational effectiveness metrics and computational costs. Statistical analysis results show that the proposed algorithm produced the best operational performance with grid spacing set to 50% of the smallest surface to air missile (SAM) threat radius without incurring prohibitive computational costs.

Military planning operations deal with highly dynamic environments and a variety of complex optimization challenges. In order to support decision-makers in this process, innovative concepts are required that can automatically generate applicable solutions for certain aspects of mission planning. Such instruments can simplify the planning process, reduce risks, and lower operating costs. This paper presents a simulation-based optimization framework that addresses three problems in the context of aerial warfare planning: task assignment, scheduling, and route planning. These problems are tackled with interconnected heuristics based on either greedy approaches or genetic algorithms. Additionally, hierarchical task networks are employed to incorporate domain knowledge in form of tactical doctrines into the solution. Our simulation results confirm the viability of the proposed approach for small to medium-sized scenarios. However, further investigation with regard to the evaluation function and the simulation environment is required.

The Sortie Generation Rate (SGR) which refers to the number of sorties that can be generated per unit time, is a key indicator for evaluating the ability of an airbase. However, an aircraft carrier has many constraints compared to a land-based airbase, such as spatial and environmental constraints, making it difficult to apply existing land-based research to analyze aircraft carrier operations. On the other hand, the Sortie Generation Process (SGP) on an aircraft carrier is similar to a logistics/production system in that sorties are generated through aircraft. Therefore, this study proposes a framework for analyzing the SGP on an aircraft carrier using discrete event simulation and defines the classes that make up the simulation. In addition, SGP analysis simulations were implemented using the proposed framework and several experiments were performed to demonstrate the feasibility of applying the proposed framework in practice.

The U.S. Army's command and control modernization efforts rely upon an expeditionary, mobile, hardened, and resilient network. Dispersed network access and data availability are central to increasing the operational speed required for effective command and control. The Army must define its satellite communication (SATCOM) requirements to support network modernization. This paper proposes the Holistic Prioritized SATCOM Throughput Requirements (HPSTR) simulation that prioritizes and adjudicates SATCOM throughput requirements for operational military units. Additionally, the simulation evaluates the impact of a contested, degraded, and operationally limited (CDO) communication environment on force effectiveness. HPSTR addresses knowledge gaps concerning U.S. Army SATCOM activities in a large-scale combat operation (LSCO) to inform modernization decisions.

Conducting a measured response to cyber or information attack is predicated on attribution. When these operations are conducted covertly or through proxies, uncertainty in attribution limits response options. To increase attribution certainty in the information dimension, the authors have developed a suite of supervised machine learning models that attribute an emerging narrative to historical narratives from known actors. These models were first developed on simulated narratives produced with a Large Language Model. Once the supervised classification models were developed and tested on the simulated narratives, they are evaluated on known actor social media narratives from three known actors. The attribution models are language agnostic and offer one-vs-rest and multi-class options. All models performed at relatively high accuracy and can provide decision support for cyber response decisions.

Cyber threats are moving beyond human comprehension and reaction capability in a rapidly evolving world. Deep learning models for network intrusion detection are becoming evermore crucial in processing network traffic to filter benign content from malicious activity. However, novel attacks such as zero-days are becoming more frequent, demonstrating the need for robust deep learning models to flag attacks while providing predictive certainty guarantees. Therefore, detecting out-of-distribution (OOD) inputs at inference time is crucial to address the rapidly changing environment while keeping up with evolving cyber threats. We develop multi-class deep learning models for network intrusion detection, comparing deterministic with Bayesian neural networks estimated using Hamiltonian Monte Carlo. We also propose new uncertainty quantification scoring measures for performance evaluation to evaluate certainty in predictions. During our experimentation, our best performing proposed Bayesian deep learning model detected 89.1% and 86.9% of the OOD packets at the 5% and 0.1% significance levels, respectively.

Technological advancements have made autonomous aerial search using unmanned systems a promising approach to search and rescue, targeting, and other mission sets. A handful of standard flight paths are traditionally used for aerial search, but this research presents the Lissajous pattern as an alternative to these traditional paths that could potentially locate targets more quickly. This research considers a searching agent with imperfect detection capability and leverages Monte Carlo simulations to generate data for various flight paths. Each flight path is evaluated by cumulative density functions representing the time it takes an unmanned aircraft system (UAS) to reach some desired percent certainty of locating a randomly generated target in a search area. Results show that Lissajous curves are viable search paths for superior aerial target detection, particularly for evasive targets in a Reciprocal Gaussian sampling distribution.

An engagement between naval ships is defined as a multi-agent system with multiple ships interacting. Because of the limitations of conducting and analyzing engagement, it is common to use modeling and simulation or wargame simulations. Most of the existing wargame simulation studies focus on simulation frameworks rather than real-world applications and tend to focus on the evaluation of single entities that comprise a wargame. Thus, this study improves the reality of the simulation by modeling objects that constitute a complex engagement situation based on the simulation framework. In addition, developed analytical tools to automate and accelerate Monte Carlo simulations of engagement-level wargames that require large numbers of human and time resources. The developed simulations enable the application of various engagement scenarios to evaluate strategies and tactics. Furthermore, experiments are possible while altering the design parameters of the naval ship, which allows for the evaluation of the ship's performance in combat.

A predictive simulation is built on a conceptual model (e.g., to identify relevant constructs and relationships) and serves to estimate the potential effects of `what-if' scenarios. Developing the conceptual model and plausible scenarios has long been a time-consuming activity, often involving the manual processes of identifying and engaging with experts, then performing desk research, and finally crafting a compelling narrative about the potential futures captured as scenarios. Automation could speed-up these activities, particularly through text mining. We performed the first review on automation for simulation scenario building. Starting with 420 articles published between 1995 and 2022, we reduced them to 11 relevant works. We examined them through four research questions concerning data collection, extraction of individual elements, connecting elements of insight and (degree of automation of) scenario generation. Our review identifies opportunities to guide this growing research area by emphasizing consistency and transparency in the choice of datasets or methods.

There are significant differences between using systems as human-controlled tools to accomplish a specific task and using systems designed to “cooperate and partner” with humans to achieve capabilities beyond either side acting alone. The live, virtual, constructive (LVC) paradigm increasingly emphasized by the DoD has wide acceptance and is congruent with how the military thinks about training, evaluation, and mission rehearsal. Consequently, it may help address these challenges. This paper aims to overview the current LVC construct, challenges associated with human-AI teaming and intentional design of these dynamics to achieve new capabilities, and the resulting need to evolve the LVC construct to improve our pursuit of understanding and evaluation that leads to effective fielding.

Fuzzy Cognitive Maps (FCMs) are graph-based simulation models commonly used to model complex systems. They are often built by participants and aggregated to compare the viewpoints of homogenous groups (e.g., anglers and ecologists) and increase the reliability of the FCM. However, the default approach for aggregation may propagate the errors of an individual participant, producing an aggregate FCM whose structure and simulation outcomes do not align with the system of interest. Alternative aggregation methods exist; however, there are no criteria to assess the quality of aggregation methods. We define nine desirable criteria for FCM aggregation algorithms and demonstrate how three existing aggregation procedures from social choice theory can aggregate FCMs and fulfill desirable criteria, enabling the assessment and comparison of FCM aggregation procedures to support modelers in selecting an aggregation algorithm. Moreover, we classify existing aggregation algorithms to provide structure to the growing body of aggregation approaches.

The simulation-based analysis of process collaborations introduces significant challenges, such as the ability to focus on the interchange of information and data without disclosing any internal details of collaboration participants' processes. The use of distributed simulation (DS) provides good opportunities to face these challenges. However, properly using DS standards and technologies requires significant technical know-how and effort. This paper introduces a largely automated approach to carry out distributed simulations of process collaborations. The DS standard addressed by the paper is the High Level Architecture (HLA), which is used to analyze process collaborations specified by using the Business Process Model and Notation (BPMN). The degree of automation is obtained by using a low-code development paradigm based on automated model transformations that reduce the amount of manual effort required to code the HLA-based simulation. An example application is also discussed to underline the pros and cons of the proposed approach.

In this paper, we introduce a novel methodology for business process simulation, focusing on the incremental transformation of Business Process Modeling and Notation (BPMN) models enriched with Business Process Simulation Interchange Standard (BPSIM) elements into the Discrete Event System Specification (DEVS) formalism. The proposed method enhances the precision and consistency of simulations by systematically converting BPMN components and BPSIM characteristics into DEVS representations, using adaptable rules and templates. A major contribution of this work is the introduction of the Interaction Intermediate Model (I2M), a model that provides a visually lucid representation with significant semantics, effectively encapsulating BPMN and BPSIM simulation aspects. The resulting DEVS model ensures accurate, reliable, and interoperable simulations. We provide a thorough analysis of this methodology, emphasize its advantages, and validate its efficiency through a case study. This method, applicable across various sectors effectively bridging the gap between conceptual modeling and simulation methodologies.

Model simplification is the process of developing a simplified version of an existing discrete-event simulation (DES) to study the performance of specific system subcomponents relevant to the analysis. The simplified model is referred to as a 'metasimulation'. A widely used model simplification operation is abstraction, which involves replacing the subcomponents, not core to the analysis, from the parent DES model with random variables representing the lengths of stay in said subcomponents. However, the one-time computational cost of developing metasimulations via abstraction can itself be considerable, as the approach necessitates executing the parent model for generating the necessary data for developing the metasimulation. Thus, this study proposes a queuing-theoretic approach for estimating the computational runtime reduction (CRR) achieved through abstraction, wherein the prediction of CRR precedes the development of the metasimulation. Towards this, we present preliminary results from applying this approach for simplification of DES models made up of M/M/n workstations.

Forty years ago, in 1983, Lee Schruben proposed the Event Graph formalism and modeling language, subsequently defining the paradigm of Event-Based Simulation, in a precise way, which had been pioneered 20 years before by SIMSCRIPT. The purpose of this panel is for a group of Event Graph researchers both from Operations Research and Computer Science, including the inventor of Event Graphs and one of his former PhD students who has made essential contributions to their theory, to discuss their views on the history and potential of Event Graph modeling and simulation. In particular, the adoption of Event Graphs as a discrete process modeling language in Discrete Event Simulation and in Computer Science, and their potential as a foundation for the entire field of Discrete Event Simulation and for the fields of process modeling and AI in Computer Science is debated.

Existing grammars generate Finite Deterministic DEVS models, a restricted subset of DEVS. The proposed context-free grammar generates the unrestricted set of Classic DEVS models. The grammar is implemented in ANTLR, a powerful parser generator for reading, processing, executing, or translating structured text or binary files. ANTLR enables the efficient processing of the specifications needed for generating members of Classic DEVS with ports. Applications include an easier introduction to DEVS for students and easier translation between different DEVS implementations.

The Discrete Event System Specification (DEVS) formalism is a modular discrete-event modeling formalism. It has a formal specification in terms of systems theory and is supported by several efficient and usable simulator implementations. In these implementations, the DEVS formalism is often “grafted” onto an existing Object-Oriented programming language. Examples are C++ in the case of ADEVS and Python in the case of PythonPDEVS. To match this grafting, we present CLAVS, the CLAss diagram and deVS formalism and its instance counterpart ODVS, the Object Diagram and deVS formalism, and their visual notations. These languages use an automaton-like visual notation for Atomic DEVS models and a Class Diagram notation augmented with port information and event structure specification. An implementation of a visual CLAVS/ODVS modeling environment built on draw.io is presented. The use and usefulness of the formalism is demonstrated by means of a simple traffic model whose detailed specification is presented.

Harmful Algal and Cyanobacterial Blooms (HABs) constitute a relevant public health and ecological hazard due to their frequent production of toxic metabolites, which is increased by the current vulnerability of water resources to environmental changes such as global warming, population growth, and eutrophication. These blooms have been typically assessed by combining predictive models with manual collection. However, these processes are generally independent and do not provide data with sufficient resolution to apply proactive policies. In this work, we propose a novel and integrative framework to straightforwardly combine the conception, design, and deployment of advanced Early-Warning Systems (EWSs) that will allow us to automate all the processes involved in HABs detection and management and apply proactive policies. The framework is built upon solid Modeling and Simulation (M&S) principles, through Model Based Systems Engineering (MBSE) as the driving methodology and Discrete Event System Specification (DEVS) as the M&S formalism.

While there are many efforts to simulate technical systems in virtual environments and provide a visual interaction for applications such as training, authoring and analysis, the process of generating applications still requires a lot of manual work. This is particularly critical in the context of interactive Digital Twins for resilience, where uncertain events can occur and every malfunction or mistreatment of any part of the system needs to be modeled. This paper presents an approach to model such systems in a modular way by automating the generation of its components for a game engine and simulators based on a common specification. Component instances are then synchronized bidirectionally across applications to achieve interaction between the game engine and simulators. An example hydraulic system is implemented and tested to demonstrate our approach, which needs minimal manual work by using predefined components. The solution can be extended by integrating more components and simulations.

Cities and large urban areas face a myriad of challenges ranging from city planning, developing sustainable transportation, managing natural catastrophes, and mitigating communicable diseases. Addressing these challenges requires effective analysis and planning which in turn necessitates the use of sufficiently detailed models or "digital twins." Such detailed models that embody multifaceted demographic and city characteristics are challenging to generate. This paper presents our ongoing work to develop a novel model generation method and software suite called Cityscape, that fuses diverse real-world data sets to generate a digital twin for a given city. Specifically, our method combines data from authoritative sources including PUMS, PUMAs, and OpenStreet Map to generate the digital twin. We have used the city of Chicago (IL, USA) as a case study to verify and validate (with ~85% confidence) our proposed method.

The modern world requires accurate and efficient traffic modeling to facilitate commerce and ensure citizens' safety. Traffic simulations play an important role in this endeavor by allowing traffic engineers to test traffic systems and policies before implementing them. This requires traffic simulation models that have the ability to accurately represent real-world traffic systems, and which are also capable of re-calibrating model parameters when needed through online calibration. This work presents four contributions toward this endeavor. The data science system ScalaTion was extended with agent-based modeling and makes use of virtual threads for each vehicle, which improves the efficiency of simulations. The modeling, simulating, and data loading schema were all optimized to enhance the system performance as well. Additionally, a new arrival model strategy was implemented improving the accuracy of the model calibration phase.

Hybrid and cyber-physical systems create synergy by combining digital modules with analog implementations of signal processing operations typically implemented in the digital domain. We propose a domain-specific language (DSL), so-called FACT – Functional Algebra for Continuous Time, based on the algebraic properties of the General Purpose Analog Computer (GPAC), a theoretical model of computation recently updated as a continuous time equivalent of the Turing Machine. We lift the GPAC to a continuous time dynamics inside a black box semantics for understanding hybrid systems, which allows us to redefine continuous time semantics inspired by the functional reactive programming style. FACT leverages the type class mechanism from the Haskell functional programming language to implement operators that capture the proposed continuous time semantics. An speed-optimized working open-source implementation in the Haskell functional language is provided and was used to demonstrate how the language supports modeling and simulation.

Discrete event simulation, formalized as deductive modeling, has been shown to be effective for studying dynamical systems. Development of models, however, is challenging when numerous interacting components are involved and should operate under different conditions. Machine Learning (ML) holds the promise to help reduce the effort needed to develop models. Toward this goal, a collection of ML algorithms, including Automatic Relevance Determination are used. Parallel Discrete Event System Specification (PDEVS) models are developed for Single-stage and Two-stage cascade factories. Each model is simulated under different demand profiles. The simulated data sets are partitioned into subsets, each for one or more model components. The ML algorithms are applied to the data sets for generating models. The throughputs predicted by the ML models closely match those in the PDEVS simulated data. This study contributes to modeling by demonstrating the potential benefits and complications of utilizing ML for discrete-event systems.

A stylized fact is a simplified presentation of an empirical finding. When modeling and simulating complex systems and real data are sparse, stylized facts have become a key instrument for building trust in a model as they represent important requirements regarding the model’s behavior. However, automatically validating stylized facts has remained limited as they are usually expressed in natural language. Therefore, we develop a formal language with a custom syntax and tailored predicates allowing modelers to unambiguously and succinctly describe important (temporal) characteristics of simulation traces or relationships between multiple traces via statistical tests. The proposed formal language is able to express numerous facts from the literature in different application domains, as well as to automatically check stylized facts. If stylized facts are defined at the beginning of a simulation study, formally expressing and checking them can streamline and guide the development of simulation models and their successive revisions.

This panel discussion is designed to provide young scholars in the field of simulation with valuable insights into identifying suitable publication avenues for their research endeavors. Senior journal editors will serve as panelists and share their wealth of experience and perspectives. Journals represented include ACM TOMACS, IISE Transactions, INFORMS Journal on Computing, Journal of Simulation, Operations Research, and Stochastic Systems. Specifically, the panelists will introduce preferred topics, focuses, and future trends for each journal. Panelists will also share their own experiences and suggestions on the peer review process, such as how to navigate through revisions and rejections, and ethical policies. Young scholars will also learn the importance of serving the community as a reviewer, and what senior editors expect from reviewers.

To cut costs and drive innovation in product development, many projects have turned to remote worksites for construction component pre-fabrication. Fabricating pipe spools in shops eliminates delays due to weather and allows for better resource planning. This paper aims to optimize labor resource usage in a pipe spool manufacturing plant that fabricates three different types of spools. It utilizes historical data to implement a discrete-event simulation model. The proposed simulation model effectively reduced idle time and evenly distributed the workload. As a result, the overall fabrication time for all three spools was reduced, leading to a 22% decrease in active shop usage. This allowed subsequent jobs to commence earlier, giving the team more flexibility in meeting deadlines and addressing labor constraints. This research provides insights into how resource allocation plans can be created to maximize sustainability results, both socially (through improving working conditions and reducing workloads) and economically.

Construction is a notoriously hazardous industry and heavy alcohol use is common. This project creates an agent-based modeling (ABM) simulation exploring the impact of alcohol on safety outcomes. Simulation modeling is useful in occupational safety research because it generates immediate results and bypasses ethical concerns. Workers and foremen interact on a virtual jobsite with hazards present. Positive blood alcohol concentration (BAC) decreases hazard awareness and reaction time, and additionally decreases competency of foremen. Scenarios of baseline, increased, and decreased alcohol consumption are analyzed for changes in near misses, injuries, and fatalities. Additional scenarios of improved training and engineering controls are explored also for comparison. A decrease in alcohol consumption led to a significant reduction in injuries by up to 12%, and an increase had the opposite effect. Neither scenario significantly impacted fatalities due to fatalities' low base rate. Safety training had a comparable impact but improving engineering controls outweighed both.

This study introduces a deep learning-based method for indoor 3D object detection and localization in healthcare facilities. This method incorporates spatial and channel attention mechanisms into the YOLOv5 architecture, ensuring a balance between accuracy and computational efficiency. The network achieves an AP50 of 67.6%, an mAP of 46.7%, and a real-time detection rate with an FPS of 67. Moreover, the study proposes a novel mechanism for estimating the 3D coordinates of detected objects and projecting them onto 3D maps, with an average error of 0.24 m and 0.28 m in the x and y directions, respectively. After being tested and validated with real-world data from a university campus, the proposed method shows promise for improving disinfection efficiency in healthcare facilities by enabling real-time object detection and localization for robot navigation.

Municipal construction projects are often challenging and risk-prone due to unexpected underground conditions. Access to As-Built and As-Design data is essential to avoid budget overruns, schedule delays, and other construction disputes. However, coordinating field conditions with construction drawings can be difficult and lead to discrepancies. Traditional methods of denoting information onto the ground by surveyors and field workers have been limited in their ability to provide relevant information and support scaling up. These methods also create restrictions in data sharing and communication among workers and engineering teams. With the development and use of AR technology, our study proposes an augmented reality tool leveraging Google ARCore to assist construction engineers in a straightforward and efficient manner by displaying utility information, including pipe direction, type, slope, diameter, and material. The campus area of the University of Maryland College Park is used as a case study to demonstrate our approach.

Lean construction is a promising approach for performance improvement in the construction industry. Value stream mapping (VSM) is an essential lean tool for the process improvement of construction activities. However, VSM, regarded as a static pen-and-paper technique, requires repeating the VSM preparation for every improvement alternative. Therefore, dynamism can be introduced into VSM by developing computer simulation models, which is the study's objective. A VSM-based discrete event simulation (DES) template is presented in this paper for off-site construction activities. The model provides a virtual testing environment for the user to decide upon the potential time reduction in non-value-added (NVA) activities for the process improvement. The development and validation of the model is done based on the actual data from a precast production factory. The DES-VSM simulation model assists plant managers with the best possible NVA reduction strategy and accelerates lean implementation in the construction industry.

The new preemption capabilities added to the STROBOSCOPE simulation system are described and illustrated by two examples. The first example involves moving soil using two wheelbarrows and two laborers. It investigates the conditions for preemption to improve production by allowing the return of an empty wheelbarrow to interrupt loading and to start hauling a partially loaded wheelbarrow immediately. In the second example, two cranes unload barges bringing fill material for undersea land reclamation. When only one barge is available, it can unload using both cranes. When two or more barges become available, each barge unloads using one crane. Unloading a barge can switch between using one and two cranes multiple times, with the remaining unload time either cut in half or doubled each time. Modeling the multiple reallocations of cranes and the required time adjustments illustrates the new STROBOSCOPE preemption capabilities.

Detailed STROBOSCOPE simulations of earthmoving for the construction of a dam use the engineering calculations typically employed in heavy construction to estimate equipment performance based on the characteristics of the haul and return roads and the mechanical properties of actual models of heavy loaders and trucks. Sensitivity analysis investigates the total cost of truck combinations while considering the traffic effects of one or two bridges needed to cross a river along the haul route. This example can serve as a simulation model template to facilitate the wider acceptance of simulation in heavy construction practice.

Public-Private Partnership (PPP) programs have been adopted to leverage private funding for the development of public infrastructure and services, thereby relieving public fiscal pressure. However, the complexity and length of PPP contracts can lead to higher costs for the public sector. Using data from more than 700 PPPs that integrate the UK Private Finance Initiative and Private Finance 2 programs, this study analyzes the long-term financial implications of these programs using System Dynamics. Causal-loop diagrams were developed to illustrate the causal structures that generate the long-term financial effects of PPPs on the public sector. The paper offers potential strategies to enhance the performance of PPP programs. This study contributes to closing the research gap identified in previous research for more efficient PPP programs by uncovering their dynamics and offering suitable policies for governments to improve their outcomes.

Research and development (R&D) of biotechnology products is an iterative process typically characterized by a monolithic workflow in which a single organization takes a project from start to finish through many complex operations. This paper presents a discrete-event simulation methodology to explore an alternative disaggregated workflow in which R&D is managed by a single organization but individual operations are distributed among multiple organizations. This methodology is applied to a protein engineering R&D process to compare the monolithic and disaggregated workflows over a range of conditions and scenarios. Based upon a set of assumed parameters, results identify conditions favorable to either workflow and provide a first indication that the industry’s trend towards disaggregation may lead to improvements in development timelines. The methodology also provides a foundation for decision support tools that enable decision-makers to manage biotechnology R&D projects.

Six Sigma is a useful technique adopted in the construction industry to attain supreme quality levels by reducing the variability in the processes. However, rigorous field implementation of a Six Sigma methodology takes time, money, resources, and stakeholder commitment. This study develops a simulation-based framework that can mimic a Six Sigma implementation effort in a construction site using a discrete event simulation technique. Such a framework helps the decision makers to check the benefits of Six Sigma by assessing what-if scenarios for possible system improvement, even before expending the time and resources needed for field implementation of Six Sigma techniques. Therefore, through a combination of discrete event simulation and Six Sigma, the variations in a process at a construction project are eliminated. The results of this study can inspire construction managers to use simulation to understand Six Sigma implementation and improve the process or system to fulfill customer needs.

The Mersenne Twister (MT) is a pseudo-random number generator (PRNG) widely used in High Performance Computing for parallel stochastic simulations. We aim to assess the quality of common parallelization techniques used to generate large streams of MT pseudo-random numbers. We compare three techniques: sequence splitting, random spacing and MT indexed sequence. The TestU01 Big Crush battery is used to evaluate the quality of 4096 streams for each technique on three different hardware configurations. Surprisingly, all techniques exhibited almost 30% of defects with no technique showing better quality than the others. While all 106 Big Crush tests showed failures, the failure rate was limited to a small number of tests (maximum of 6 tests failed per stream, resulting in over 94% success rate). Thanks to 33 CPU years, high-quality streams identified are given. They can be used for sensitive parallel simulations such as nuclear medicine and precise high-energy physics applications.

Bail funds have a long history of helping those who cannot afford bail in order to wait for trial at home. They have also had a large impact on the verdict of the defendant. In this paper, we present the first stochastic model for capturing the dynamics of a community bail fund. Our bail fund model integrates traditional queueing models with classic insurance/risk models to represent the bail fund’s intricate dynamics. We employ simulation techniques to assess Gaussian-based approximations that estimate the probability of a defendant being denied access to the bail fund when it lacks the adequate funds to support them. Additionally, we propose a new simulation-based algorithm that leverages a deterministic infusion of capital as a control variable to stabilize the probability that defendants have access to the bail fund. Our simulation results reveal that our Gaussian-based approximations are suitable for moderately and highly active bail funds.

We explore the application of a Sensor Fusion Framework, called SAFE (Simple, Applicable, Extensible, and Flexible) to improve the reliability of measurements obtained from Inertial Measurement Unit (IMU) sensors. SAFE is built using a DEVS specification and the Cadmium tool. Measuring angular position is a difficult task due to the unreliability of gyroscopes and accelerometers, two sensors widely used to measure angles. Although angular position can be measured using imaging systems, these are costly, and not ideal for handheld and portable devices. An alternative solution is to use sensor fusion to fuse the readings of both accelerometer and gyroscope, obtaining reliable readings. We show the application of the SAFE methodology and the results of our case study showing the potential of this method.

This paper presents a virtual testbed for the development and verification of cyber-physical systems, integrating network simulation, physics, and hardware emulation within the multi-domain simulation framework VEROSIM. The testbed facilitates comprehensive software-in-the-loop testing, enabling accurate and reliable evaluation of control algorithms in complex, interconnected systems. The integrated approach simplifies simulation setup and model management, while allowing natural treatment of mobility and the use of sophisticated physical radio wave propagation models. The testbed also enables the simulation of various fault scenarios, supporting the assessment of system resilience and fault-tolerant strategies. A case study involving a capsule approaching the International Space Station demonstrates the effectiveness of the proposed testbed in capturing the interactions between software, hardware, and physical elements, and verifying the overall behavior of a cyber-physical system under adverse conditions.

The growing frequency of power outages has prompted increased interest in developing a more resilient power grid that can quickly recover from weather-related damage. At the distribution level, power restoration is a complex, multi-stage process involving multiple response entities. Providing utility stakeholders, government regulators, and the public with information about outage duration and estimated time to restoration is crucial. The research employs a multi-agent simulation approach, which allows for the simulation of decision-making behaviors among different entities and the incorporation of various uncertainties. Specifically, the study uses the open-source simulation package Mesa-Geo in conjunction with the Python language and constructs a road network using the open-source network extension pgRouting for routing queries. The research design includes several experiments focused on Florida as a case study, comparing repair crew sizes, power outage numbers, and road damage scenarios. The findings could offer valuable managerial guidance on resource allocation in the restoration process.

In recent years, there has been a significant increase in the deployment of Cyber-physical Production Systems (CPPS) across various industries. CPPS consist of interconnected devices and systems that combine physical and digital elements to enhance the efficiency, productivity, and reliability of manufacturing processes. Due to the continuous and fast-paced evolution of the behavior of CPPS, there is an increasing interest in generating data-driven Discrete-event Simulation (DES) models of such systems. The validation of these models, however, remains a challenge, and traditional approaches may be insufficient to ensure their accuracy. To address this challenge, we propose a framework for validating data-driven DES models of CPPS. We emphasize the importance of continuously monitoring the validity of data-driven DES models and updating them when necessary to ensure their accuracy over time. We, furthermore, demonstrate our proposed approach through a case study in reliability assessment and discuss challenges and limitations of our framework.

The Singular Value Decomposition (SVD) algorithm is ubiquitous in many fields of science and technology. It may be used embedded into other advanced algorithms, solvers or data processing chains. In those scenarios dealing with large data volumes expressed as a huge matrix, there is the need of a parallel SVD version to process it efficiently. We present some ideas and results obtained within the PETSc framework, which enable to design promising HPC scalable solvers. The focused SVD implementations have been taken from the SLEPc library, which is seamless plugged into PETSc to extend its capabilities. Besides, there is also a randomized SVD and wrappers to interface ScaLAPACK and others packages to extract singular triplets. This work assesses the strong scaling attained with these SVD implementations at extracting the leading singular values of a population of both sparse and dense matrices. A comparison of performance is provided.

Simulation experiments are crucial in conducting simulation studies. With simulation studies growing increasingly complex, simulation experiments are intertwined with steps of conceptual modeling, model building, analyzing data, and visualizing and interpreting results. Making the products of these various steps (assumptions, requirements, data, model components, and experiments) explicit has been shown to increase the reproducibility of simulation studies. Moreover, using an integrated environment that allows developing, organizing and documenting those products can facilitate their automatic reuse and exploitation. We explore Jupyter Notebook as an all-in-one solution for conducting and documenting a simulation study, and we present nbSimGen. This Jupyter Notebook extension lends support to modelers by automatically specifying and running suitable simulation experiments. It is based on an annotation vocabulary that, during the development of the conceptual model and the simulation model, allows users to mark portions of their notebook deemed relevant to the various simulation experiments to come.

Considering some challenges that prevent the expansion of discrete event simulation studies, such as financial constraints to invest in the data collection of large samples and to hire qualified people for data analysis and for developing complex models, this paper aims to propose a framework to support simulation studies where it is not widely used, adopting facilitated modeling. Since the facilitated DES frameworks in the literature focus on healthcare and face-to-face meetings, the present work offers a framework for simulation projects in production systems, which also supports online interventions. After its development, the FaMoSim (Facilitated Modeling Simulation) framework was applied in a real case to evaluate its applicability. In the application, it was possible to carry out a faster and more flexible online modeling process, create a simple computer model that does not require a complex data collection structure nor a specialist team, and assist the stakeholders in identifying improvements.

Good quality healthcare services require effective communication between the patient and the healthcare provider. This work will help improve the areas of healthcare systems automation and optimization by applying Speech Emotion Recognition (SER) in health consultations to prevent miscommunication between patients and healthcare providers. Crowd-Sourced Emotional Multimodal Actors Dataset (CREMA-D) was used to compare the performances of different machine learning models in classifying emotions. Before feeding the raw dataset to the models, exploratory data analysis was done to determine features that should be considered for future analysis. Our results showed that depending on the emotion, there are some syllables in the text that were emphasized or took time to be pronounced by the speaker. After data analysis, the dataset was fed into different models and determined that the Support Vector Machine (SVM) is a machine-learning model for SER.

The disruptive technology provided by large-scale pre-trained language models (LLMs) such as ChatGPT or GPT-4 has received significant attention in several application domains, often with an emphasis on high-level opportunities and concerns. This paper is the first examination regarding the use of LLMs for scientific simulations. We focus on four modeling and simulation tasks, each time assessing the expected benefits and limitations of LLMs while providing practical guidance for modelers regarding the steps involved. The first task is devoted to explaining the structure of a conceptual model to promote the engagement of participants in the modeling process. The second task focuses on summarizing simulation outputs, so that model users can identify a preferred scenario. The third task seeks to broaden accessibility to simulation platforms by conveying the insights of simulation visualizations via text. Finally, the last task evokes the possibility of explaining simulation errors and providing guidance to resolve them.

The overarching goal of beamline design is to achieve a high brightness electron beam from the beamline. Traditional beamline design studies involved separate optimizations of radio-frequency cavities, magnets, and beam dynamics using different codes and pursuing various intermediate objectives. In this work, we present a novel unified global optimization framework that integrates multiple physics modules for beamline design as simulation functions for a two-stage global optimization solver.

Radiation protection applications often require the creation of a large number of precise simulations of radiation-human body interactions. Our research is focused on creating RadPhantom, a new Geant4 application that constructs voxelized anthropomorphic phantom models. This allows for the standardized and reproducible generation of Geant4 simulations in cloud-based environments. We have incorporated existing and publicly accessible models into Meiga, a framework designed for the integration of Geant4-based applications. To standardize these simulations, guarantee their reproducibility, and adhere to the FAIR principles, we have developed an extended vocabulary schema using metadata and ontologies that align with current standards. By employing virtualization containers, we capitalize on the scalability and adaptability of public and federated clouds. In this paper, we detail our implementation, present some benchmarking results and comparisons with current methodologies, and discuss the potential applications for evaluating doses on commercial flights or assessing radiation shielding in neutron production facilities.

Chile is the world's leading copper producer, with more than 5.6 million tons produced in 2020. Most of the produced ore comes from open pit mines, whose extraction process consists of different subprocesses, with ore hauling incurring the highest operational cost. Tools to improve this subprocess are of paramount importance. Most tools use approaches that rely on optimization based on analytical methods. However, these fail to capture human behavior or to consider fine-grained details. To this end, we present a DEVS (Discrete-Event System Specification) simulation model. The formal definition of DEVS helps with the design and experimentation. DEVS modular interfaces allow users to extend the model easily to consider more entities, mine layouts, and dispatching policies. Simulations of the model delivered precise results compared to the literature, providing a valuable tool for decision-making in the mining industry.

Generally, bridge construction is one of the most complex structures in the construction industry due to the higher scalability and supply chain complexity. The modular bridge construction (MBC) technique is considered more advantageous in providing higher productivity, shorter schedules, and better quality. Current practices in managing MBC projects overlook dynamic behaviors among the relevant stakeholders and the interactions among various interacting systems, including manufacturing, logistics, and onsite assembly. To this end, this paper proposes a simulation-optimization framework to enhance MBC projects planning. The simulation module comprises discrete event simulation and agent-based modeling to model the interconnected behaviors of the MBC systems. The optimization module aims to improve the key performance indicators (KPIs) of MBC projects, including project cost, schedule, and sustainability. The proposed framework is validated by introducing an MBC case of a cable-stayed bridge. The generated solutions by the optimization model show possible significant enhancements in the identified KPIs.

Unbalanced production rates of activities and abundant resource allocation are the leading reason behind bottlenecks in processes and have been one of the causes that negatively affect projects leading to wasted resources. Many industries suffer from unbalanced resource workloads, where manufacturing takt times at some workstations are out of sync with preceding stations, consequently leading to an abruption in the workflow between activities. This research aims to assess the current state of the manufacturing process of a wall assembly line from material cutting to installation, identifying bottlenecks, and creating a framework that would contrast both cycles to finally propose a solution through simulation. A case was studied to propose innovative methods to improve the process flow and to eliminate any waste generated by bottlenecks. This will not only reduce the process duration but will also significantly increase cost expenditure since the amount of idle time and resources will be reduced.

Simulate Simulation Technology is a Brazilian consultancy company focused on developing discrete event simulation models and providing simulation training. Some of the simulation models developed over the last 20 years are classified by us as successful and impactful, with no relationship to their complexity, applicability level, or purpose. This article presents some of these models.

This paper addresses the problem of determining organizational adaptations to ensure business continuity. We propose a methodology to assess the impact of disruptions on a business model and evaluate interventions using System Dynamics archetypes. The methodology aims to contribute to making decision-making more effective and efficient in an uncertain scenario.

Managing effective visualisations for data analysis is critical to support informed decision making across multiple domains, which also requires the ability to interact with the data. This includes understanding data from real-world scenarios enriched with simulated virtual data, and the ability to assess the impact of user interventions on concurrently running simulation models. To address this, we propose a framework that combines a DEVS simulator with a game engine, allowing users to interact directly with the model during simulation runtime, while observing realistic visualisations of the generated data and system behaviour.

We present RustSim, a library for discrete-event process-oriented simulations designed and implemented in Rust programming language. It includes a broad set of classes to allow the user to implement simulation processes and process-oriented primitives. The flexible modular design of RustSim allows users to extend its functionality. In addition, RustSim includes mechanisms to avoid inconsistencies when applying state-changing primitives that other libraries in the language's ecosystem do not provide. We take advantage of Rust generators (coroutine equivalent) to implement process-oriented simulation primitives. Finally, the library's internal process handling structure is discussed in detail, including its implementation, how simulations are executed, and a case study with a highly detailed example of its use.

Stream processing platforms allow processing and analyzing real-time data. Several tools have been developed for these platforms to guarantee that the applications running on them are scalable, fast, and fault-tolerant and that they can be deployed on many processors. However, determining the proper number of processors suitable to hold a given stream processing-based software application is challenging, especially if the application is intended to serve a large user community. In this paper, we propose to model and simulate stream processing platforms for performance evaluation purposes. In our case study, we simulated a commonly used application for the analysis of Twitter streams with Storm. We evaluate its performance under different workloads. Our simulator supports profiling to measure various aspects of the application's performance. Results show that the simulator can replicate the metrics reported by the application running on a real platform with minimal error.

Routed DEVS (RDEVS) models improve traditional discrete-event models by enhancing the development of routing processes over predefined behaviors. In this paper, we demonstrate how a Software Engineering design pattern, specifically the Decorator pattern, was applied to the RDEVS formalism design to include event tracking into the models without altering their expected behavior. As a result, we provide a solution that allows getting structured data from RDEVS models at execution time.

Agent-Based Modeling and Simulation (ABMS) is a valuable tool for understanding infectious disease propagation. This study presents a hybrid ABMS approach to explore the transmission dynamics of vector-borne diseases (Dengue, Zika, and Chikungunya) in Bello, Colombia, incorporating geospatial characteristics. The model was developed with specific assumptions to validate its alignment with theoretical behavior. Our results demonstrate the temperature’s significant impact on disease spread. Particularly, Chikungunya exhibits distinct behavior compared to Dengue and Zika. While major infection peaks occur early in the simulation, subsequent spread diminishes due to the absence of reinfection considerations. This research represents an early stage of a larger project, laying the groundwork for future research to address computational challenges, enabling statistical analysis with multiple runs, and enhancing the model’s realism with seasonal temperature variations and geographical distributions. These findings will provide valuable insights for policymakers and disease control strategies in Colombia.

Using Agent-Based Models (ABM) for disease incidence may help decision-making processes. This work shows an ABM for cervical cancer detection. Our results show the relevance of social indicators.

Motivated by hospital parking problems that limit the access of patients and visitors, we study a hospital parking setting comprising an on-site parking lot with an occupancy-based dynamic tariff and a free shuttle service from an off-site free parking lot. We developed a discrete event simulation model to study the system’s dynamics and find the preferable coordinated tariff and shuttle schedule that maximize revenue for the contractor operating the hospital’s parking services under a predefined service level. We use a case study from Hadassah Medical Center in Ein Kerem, Jerusalem, to demonstrate the effectiveness of our method. Our results show that the coordinated solution provides significantly better performance: more than a 30% increase in service level, a 25% (about $5,000) increase in daily revenue, and a 53% decrease in average waiting time for a shuttle.

Numerical weather prediction (NWP) contributes significantly in the production of appropriate weather forecasts. These critical capabilities were still largely lacking in East Africa in the early 2010s and were recently established under the auspices of the WIMEA-ICT Project. The project introduced the use of the Weather Research and Forecasting (WRF) model in the region. This model was adopted by the National Hydro-meteorological Agencies and is largely being used as guidance in the operations. However, due to advances in technology, there is a need to build capacity in NWP data assimilation as well as Machine Learning to further improve the accuracy of weather and climatic predictions. Additional crop weather modelling studies will further inform agricultural productivity enhancement in the region.

Simulation models have been employed in developed countries for healthcare service operations management. However, leveraging simulation in developing countries is limited because healthcare operations management challenges are quite different due to scarcity of resources, high population numbers, high healthcare demand, and poor planning, implementation, monitoring and evaluation. This study, hence, aims to investigate the usage and adoption of simulation for healthcare operations management in developing countries and the challenges of using simulation in this context by studying the case of Ethiopia through a systematic literature review and survey.

Mobile crane location (MCL) in modular construction is a complex problem that affects both construction safety and efficiency. Sub-optimal MCL planning increases the number of crane relocations and the overall project cost. Interestingly, recently, research on crane operation planning and analysis focused on determining crane configurations, boom lengths, and radii to enable lifting given a crane location. However, with a large number of feasible locations, finding the best solution becomes a harder task. In this respect, finding a single crane location ensures an optimal lift plan, e.g., minimizing the number of pick-location. As a result, this paper aims to bridge this gap by providing a hybrid approach using heuristics, grid-based, and combinatorial optimization algorithms to find the least required lifting points. The proposed approach is tested on a case study of a modular building. The study contributes by minimizing the number of crane relocations to enhance budget and cost planning.

The Connected Farm is composed of several elements that communicate with each other through a 4G/5G Radio Base Station (RBS) placed in the middle of the farm. This RBS is connected to the Internet, allowing communication for all kinds of autonomous devices, performing uninterrupted tasks. This work simulates the Connected Farm environment for an autonomous drone. Our model intends to define when each drone needs to recharge its batteries, with no collusion regarding this recharging decision, reducing the drone's battery usage due to the absence of this communication.

Agent-based simulations have been used in modeling transportation systems for traffic management and passenger flows. In this work, we hope to shed light on the complex factors that influence transportation mode decisions within developing countries, using Colombia as a case study. We model an ecosystem of human agents that decide at each time step on the mode of transportation they would take to work. Their decision is based on a combination of their personal satisfaction with the journey they had just taken, which is evaluated across a personal vector of needs, the information they crowdsource from their prevailing social network, and their personal uncertainty about the discomfort of trying a new transport solution. We simulate different network structures to analyze the social influence for different decision-makers. We find that in low/medium connected groups inquisitive people actively change modes cyclically over the years while imitators cluster rapidly and change less frequently.

The Physical Internet (PI) is a recent paradigm in the supply chain management that proposes a framework in which standardization and optimization are key factors to raise supply chain efficiency, resilience, and sustainability. Strategic decisions are included in the PI, including the supply chain network design (SCND). In fact, structuring a (near) optimal design is essential to achieve the PI objectives. Additionally, disruptive events such as the COVID-19 pandemic, earthquakes, or terrorist attacks threaten the supply chains. These events are difficult to predict, but their effects can be simulated when addressing this problem. Hence, we propose a simulation-optimization approach that hybridizes a multi-objective multi-period mixed-integer program with discrete-event simulation to optimize both cost and resilience in the SCND. Furthermore, a network hyperconnection strategy is tested. Results show that both resilience and risk are improved after hyperconnecting the supply chain, especially when active edges are disturbed, but incur higher costs.

We present an application of the EB-DEVS modelling framework for agent-based complex adaptive systems to a systematic study of the international Product Space network in the field of Economic Complexity. The evolution of the production structure of agents (countries) becomes mutually determined by an emerging macroscopic network (resulting from the worldwide trade). This framework allows to make prospective analysis about the productive structure of countries.

Scheduling real systems is complicated because of the consideration of various working conditions. Although various combinatorial optimization methods, ranging from mathematical models, heuristics, metaheuristics, etc., have been developed, these methods could require a long computational time due to the complexity of the problems. This study proposes a framework to understand the system’s behavior using regression machine learning techniques. The considered system could be any type, e.g., the flow shop, job shop, and their variants, with a certain scheduling method. The framework consists of (1) the development of the simulation for generating the data and (2) how the data could be used for training the regression machine learning models. An example of the stochastic single-machine problem with the First-In-First-Out rule is considered. The framework could be used to simplify the process of understanding the system’s behavior without any necessity to solve the optimization problem, which could be time-consuming.

In the aftermath of a sudden catastrophe, First Responders (FR) strive to promptly reach and rescue victims. Simultaneously, individuals take roads to evacuate the affected region, access medical facilities or shelters, and reunite with their relatives. The escalated traffic congestion significantly hinders critical FR operations. In this study, we construct an Agent-Based Simulation (ABS) model that extends the existing models by incorporating FR agents, their allocated road map, and their interaction with evacuees in the network. Our model investigates individuals' evacuation times as well as FRs' rescue operation performance, provided that a subset of road segments are reserved for the explicit use of FRs. The decision-maker can allocate these segments manually within the simulation interface. Subsequently, the consequences are discovered through the earthquake scenario outputs of the ABS model, casting light on its real-world impact.

Earthquakes are one of the most common natural disasters and assessing the hazard levels of the affected regions and planning post-disaster operations, including search and rescue operations, are very critical. As the roads can be blocked due to an earthquake and debris removal may take time preventing critical rescue operations from starting, drone utilization has been increasing. Since the drones fly, it will be easier to assess the damage levels. However, drones have a major drawback, their batteries. In this study, we propose a scenario-based mathematical model to allocate a limited of batteries before the earthquake while computing the drones’ paths for each scenario maximizing the total expected priority scores. Our preliminary analysis shows that small instances can be solved very efficiently.

Owing to the increase in computing power, large-scale agent-based modeling (ABM) has been increasingly used in various fields. However, a complete and detailed individual population is challenging to obtain because of confidentiality concerns. Thus, modelers must adopt population synthesis to emulate the joint distribution of individual-level attributes of the actual population in the region of interest. Traditional population synthesis methods often exhibit issues regarding scalability and sampling zero. Therefore, this paper presents the use of a deep generative model called the denoising diffusion probabilistic model to generate new samples. Our proposed method uses the characteristics of deep generative model of generation from noise to generate a synthetic population, including sampling zero. In the experimental results, the standardized root mean squared error of our proposed model performed 2.130, which outperformed 2.381 of the deep learning-based population synthesis method, VAE, and 7.620 of the traditional population synthesis method, MCMC.

Quantum computing is a contemporary engineering discipline that innovatively overcomes computational burdens. This study applies quantum computing techniques to data analyses with input data issues. When a dataset has insufficient attributes and uncertainties, quantum embedding techniques contribute to the dimensional expansion of input vectors and the quantification of uncertainties. The converted qubits are linked to subsequent deep learning modules, and this architecture is used for accurate data analysis. This study proposes a quantum embedding technique and a corresponding quantum neural network (QNN) to better understand these processes. In this QNN architecture, input data are converted into corresponding qubits, which are transformed with quantum phase-operating modules. The quantum features pass through subsequent deep learning layers for more accurate data analyses. To demonstrate the effectiveness of the proposed model, a process model and relevant analyses are presented and compared with existing deep learning methods.

This paper details the simulation and performance results of a Sparse Hyper-Distributed Robust Efficient Neural Network (SpHyRE-Net) architecture that performs anomaly detection for real-world time-series data. SpHyRE-Net is an innovative, novel, low size, weight and power (SWaP) machine learning solution for devices operating at the tactical edge. It utilizes bit operations and sparse hyper-dimensional representations for bio-inspired learning via a Hebbian-like rule that results in a combined power-latency reduction of 2-orders of magnitude over ordinary deep networks. The paper details the application of SpHyRE-Net to real-world cell-traffic datasets as well as simulation requirements to minimize latency and memory use. Also discussed are the mechanisms necessary for implementing the architecture on an FPGA as a precursor to realization on a neuro-morphic ASIC with ultra-low power profile.

This paper introduces a conversational framework that enhances the usability of smart energy system simulations. This study is centered around OpenAI's Generative Pre-trained Transformer (GPT), a fine-tuned conversational model that allows users to communicate with the system in a natural way. Therefore, users can describe their simulation scenarios in plain language and GPT seamlessly translates these descriptions into Python scripts, used as inputs to the simulation environment, in our case, AnyLogic Simulation Software. Our framework is based on the i7-AnyEnergy core framework to compute distribution flows and relevant statistics. The proposed human-machine interface facilitates and accelerates simulation modeling, as demonstrated through the two scenarios we have provided in this paper. Overall, our conversational framework has the potential to significantly improve the user experience of smart energy system simulation environments. By simplifying the interaction between users and complex simulation models, we enable users to obtain valuable insights rapidly and more easily.

The explainability of large and complex simulation models is an open problem. We present a framework to analyze such models by processing multidimensional data through a pipeline of target variable computation, clustering, supervised classification, and feature importance analysis. As a use case, the well-known large-scale hydrology and crop systems simulator VIC-CropSyst is utilized to evaluate how climate change may affect water availability in Washington, United States. We study how snowmelt varies with climate variables (temperature, precipitation) to identify different response characteristics. Based on these characteristics, spatial units are clustered into six distinct classes. A random forest classifier is used with Shapley values to rank static soil and land parameters that help detect each class. The results also include an analysis of risk across different classes to identify areas vulnerable to climate change. This paper demonstrates the usefulness of the proposed framework in providing explainability for large and complex simulations.

Deep learning models have gained great success in many applications, but their adoption in financial and actuarial applications have been received by regulators with some treprdation. The lack of transparency and interpretability of these models leads to skepticism about their resilience and reliability, which are important factors to ensure financial stability and insurance benefit fulfillment. In this study, we use stochastic simulation as a data generator to examine deep learning models under controlled settings. Our study shows interesting findings in fundamental questions like "What do deep learning models learn from noisy data?'' and "How well do they learn from noisy data?''. Based on our findings, we propose an efficient nested simulation procedure that uses deep learning models as proxies to estimate tail risk measures of hedging errors for variable annuities. The proposed procedure uses deep learning models to concentrate simulation budget on tail scenarios while maintaining transparency in the estimation.

The problem of reinforcement learning is considered where the environment or the model undergoes a change. An algorithm is proposed that an agent can apply in such a problem to achieve the optimal long-time discounted reward. The algorithm is model-free and learns the optimal policy by interacting with the environment. It is shown that the proposed algorithm has strong optimality properties. The effectiveness of the algorithm is also demonstrated using simulation results. The proposed algorithm exploits a fundamental reward-detection trade-off present in these problems and uses an algorithm for the quickest detection of the model change. Recommendations are provided for faster detection of model changes and for smart initialization strategies.

Industry 4.0 induces an increasing demand of autonomous interaction between the units of production facilities, like work centers and transportation equipment. This has an impact on the requirements for production scheduling and control algorithms. These must be capable to adapt autonomously to changes on the shop floor. This paper presents a combination of Reinforcement Learning and discrete event simulation for controlling a flexible flow shop using a gantry robot system as transportation unit. In a gantry robot system parts are transported by carriages fitted with grippers that travel along rails from machine to machine. The presented agent learns autonomously the right control policy to move the carriages. It is shown that in cases the optimal policy can be determined the Reinforcement Learning based policy is optimal and in other cases the achieved throughput does slightly exceed the throughput gained by a heuristic priority rule for controlling the gantry robot.

A substantial part of fighter pilot training is simulation-based and involves computer-generated forces controlled by predefined behavior models. The behavior models are typically manually created by eliciting knowledge from experienced pilots, which is a time-consuming process. Despite the work put in, the behavior models are often unsatisfactory due to their predictable nature and lack of adaptivity, forcing instructors to spend time manually monitoring and controlling them. Reinforcement and imitation learning pose as alternatives to handcrafted models. This paper presents the Learning Environment for the Air Domain (LEAD), a system for creating and integrating intelligent air combat behavior in military simulations. By incorporating the popular programming library and interface Gymnasium, LEAD allows users to apply readily available machine learning algorithms. Additionally, LEAD can communicate with third-party simulation software through distributed simulation protocols, which allows behavior models to be learned and employed using simulation systems of different fidelities.

Scheduling is a fundamental task in each production facility with implications on the overall efficiency of the facility. While classic job-shop scheduling problems become intractable when the number of machines and jobs increase, the problem gets even more complex in the context of semiconductor manufacturing, where flexible production control and stochastic event handling are required. In this paper, we propose a Deep Reinforcement Learning approach for lot dispatching to minimize the Flow Factor (FF) of a digital twin of a real-world, stochastic, large-scale semiconductor manufacturing facility. We present the first application of Reinforcement Learning to an industrial grade semiconductor manufacturing scenario of that size. Our approach leverages self-attention mechanisms to learn an effective dispatching policy for the manufacturing facility and is able to reduce the global FF of the fab.

Effectively managing products at the end of their lifecycle is increasingly crucial as numerous systems adopt recovery strategies. However, many are limited to remanufacturing or recycling as the only recovery option. Effectively handling end-of-life products demands diverse approaches, including refurbishing and cannibalization. Sustainable recovery centers and manufacturers encounter challenges linked to uncertainties about the quantity and condition of returned products, which can disrupt operations and lead to bottlenecks. Our solution employs machine learning, specifically a CNN-LSTM model that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM), for predicting return product quantity and quality. Additionally, we utilize scenario-based simulations to proactively pre-identify and address bottlenecks within a short timeframe, especially within systems managing multiple recovery options or dealing with complex and hazardous materials.

An LNG cargo tank, referred to as the Cargo Containment System(CCS), encompasses several barriers intended for the storage of LNG at extremely low temperatures. In the case of the membrane-type CCS, each barrier is composed of insulation panels and membrane sheets. The CCS schematic arrangement endeavors to minimize the number of panels and sheets to enhance the manufacturing productivity. In this study, a combinatorial optimization approach is adopted to obtain the optimal CCS schematic arrangement. Then, a simulation environment is established to assess the arrangement results under diverse design conditions. By comparing the actual CCS design with the results of the proposed arrangement, the effectiveness of the proposed approach is valiated.

Simulation is widely used in several areas of knowledge, from engineering to biology, including physics and finance. It allows the evaluation of the model’s results under different conditions, enabling performance analysis and more assertive decision-making. However, simulation can be computationally intensive, especially when we consider complex models. To deal with this problem, metamodeling has been increasingly used as a simulation optimization technique. In this article, we propose a new adaptive metamodeling method for simulation optimization, which aims to achieve better results using fewer experiments. This method combines machine learning and metaheuristic techniques, allowing the identification of the most important regions of the search space, which can be explored more efficiently to obtain optimal solutions. The results achieved in a manufacturing problem show that the proposed method presents a significant improvement in the achieved objective function value, in comparison with the conventional benchmark method, without compromising the simulation execution time.

Reinforcement learning provides a mathematical framework for learning-based control, whose success largely depends on the amount of data it can utilize. The efficient utilization of historical samples obtained from previous iterations is essential for expediting policy optimization. Empirical evidence has shown that offline variants of policy gradient methods based on importance sampling work well. However, existing literature often neglect the interdependence between observations from different iterations, and the good empirical performance lacks a rigorous theoretical justification. In this paper, we study an offline variant of the natural policy gradient method with reusing historical observations. We show that the biases of the proposed estimators of Fisher information matrix and gradient are asymptotically negligible and reduce the conditional variance of the gradient estimator. The proposed algorithm and convergence analysis could be further applied to popular policy optimization algorithms such as trust region policy optimization. Our theoretical results are verified on classical benchmarks.

Having a sustained human presence on the lunar surface is a central objective of the Artemis Program, as it represents a key pre-requisite in resource mining operations on the Moon as well as an important steppingstone for future Martian exploration and colonization. Despite its importance, this endeavor has little precedent to rely on to inform the many challenges it needs to address. Digital Twin (DT), in recent years, has been employed in a wide range of applications. In this paper, we explore its usefulness in establishing the Artemis Base Camp. DT can be applied to various stages of the lifecycle of the lunar base development. We also identify several open questions that need be addressed before the digital twin can be utilized effectively in this project. In fact, addressing these questions could facilitate deploying DTs in use cases in a wider spectrum of industries and sectors.

Modelling and analysis of systems that are equipped with sensors and connected to the Internet are becoming more automated and less human-dependent. However, bringing expert knowledge into the loop along with data obtained from Internet of Thing (IoT) devices minimizes the risk of making poor and unexplainable decisions and helps to assess the impact of different strategies before applying them in reality. While Digital Twins are more of a data-driven simulation of the physical system, Cognitive Digital Twins bring the human dimension into the modelling and simulation. In this paper, we aim to emphasize the crucial role of explainability and the underlying rationale behind automated or interactive decision-making processes. Furthermore, we propose an initial framework that delineates the specific points within the feedback loop of a cognitive digital twin where human involvement can be incorporated.

The retail industry is undergoing a profound transformation, driven by technological advancements including AI and evolving consumer behaviors. However, what retail decision making lacks at present is knowledge of and integration of ways to factor in customer behavioral drivers in purchase decisions. We show how this can be done through a four-step approach that will create a behavior simulation model for retail use cases. We use a real world problem as a guiding example to explain our approach. Our approach enables retailers to use behavioral drivers to nudge customers and better explainability of the decisions.

In this research paper, we propose a digital twin prototype to improve mass vaccination centers in the Montreal region. This research is important because is it always challenging to define an optimal layout/capacity for healthcare operations, especially in an emergency mode (e.g., pandemic mode). Indeed, in such stressful situations, all managers are more concerned about the effectiveness of daily operations, regardless of their efficiency. Following a "design science" research approach, we developed (i) an IoT prototype for real-time patient tracking, (ii) a simulation model, and (iii) integrated them to build our digital twin prototype. Our institution's IoT lab was used as a testbed research environment for developing the IoT infrastructure and simulating the vaccination center. While the prototype was developed for vaccination centers, the approach can be used in any other multi-patient/multi flow operational environment where real-time visibility and simulation are required

The National Institutes of Health (NIH) main campus in Bethesda, Maryland currently contains 30 parking structures. On any given day, 12,000 vehicles enter the campus. NIH is planning for the south side of the campus to become the main parking areas for employees and visitors. Central to this vision is replacing a surface lot, which contains 241 parking spaces, with the construction of a greenfield six story parking structure that has a planned capacity of 1420 parking spaces. NIH wanted to prioritize the employee experience and emphasize the safety of pedestrians and vehicles. MOSIMTEC utilized simulation modeling to provide NIH with insight on the impact of various entrance and exit combinations into the parking structure. This presentation will further describe the project, the system being modeled, the inputs and outputs of the simulation tool and the outcome upon the design of the greenfield parking structure.

The pandemic period has witnessed a rapid growth of online delivery services in various sectors, especially in the domain of fresh produce e-commerce. Gousto, for instance, provides a meal subscription service where customers select their meals for a week, and subsequently receive a box containing all the required ingredients along with step-by-step cooking instructions for the chosen recipes. In light of recent economic difficulties worldwide, Gousto is prioritising its efficiency to reduce cost and to continue providing affordable meals to its customers. One key aspect for Gousto was to improve its station utilisation, through better routing of boxes throughout the factory. The use of simulation as a digital twin has been a key factor in the development of a new routing algorithm, that has now been put in production and has increased station utilisation by 20%, in line with the simulation's predictions.

Microgrids are localized power systems that can function independently or alongside the main grid. They consist of interconnected generators, energy storage, and loads that can be managed locally. Using SystemC-AMS, we demonstrate how microgrid components, including solar panels and converters, can be accurately modeled and simulated, along with their interactions. Real-time simulations are crucial for understanding microgrid behavior and optimizing components. This approach facilitates seamless integration with hardware prototypes and automation systems, supporting various development stages. Our study presents a best-case scenario for real-time simulation, assuming each loop takes less time than the simulation time step, with fallback to the previous value if data isn't received in time. This article introduces the first known real-time simulation strategy using SystemC-AMS, enabling the real-time simulation of microgrid components and integration with external devices. The implementation adopts a model-based design approach, creating increasingly complex systems with grid components and controllers.

Ionizing radiation refers to particles or photons that carry enough energy to remove electrons from atoms or molecules. Through ionizing interactions, radiation can have severe implications for human health and the environment, making it essential to develop effective strategies to manage the risks it poses. To display the potential benefits from the application of digital twin technologies to concerns regarding radioactive material in laboratory, university, and national defense settings, this paper presents the development of a digital twin framework, and potential use cases for the framework. The platform was demonstrated in two scenario studies. The first scenario involves a faux radiation-detecting glovebox used for lab safety education, while the second scenario addresses training for first responders in a nuclear defense and safety situation.

A critical facility's reliability relies heavily on its maintenance process's effectiveness. This process involves numerous sub-processes, which can be challenging to model due to uncertainties and complexities. System managers often seek a predictive tool, and this work extends a previous study that developed a digital twin of a nuclear facility's maintenance task process using data-driven and stochastic modeling, along with expert input. The authors extended the project's previous iteration by enhancing the bootstrapping technique and improving the model's fidelity.

Taking into account the crucial node of the logistics network, this paper concentrates on the layout design problem of logistics parks considering numerous uncertain factors during operations. To provide comprehensive support for park planners and managers, a simulation-optimization-powered toolbox is developed for decision-making, with core functions such as park layout design, construction quantity calculations, and performance evaluations. A case study demonstrates the toolbox's effectiveness in assisting users to achieve their desired layout designs, and the result shows that the optimized layout generated by the toolbox can lead to improvements of approximately 13%.

We consider the problem of scheduling automated guided vehicles (AGVs) in a stochastic warehouse management system. This problem was studied in the Case Study Competition of the 2022 Winter Simulation Conference. We propose a simulation optimization method that simultaneously optimizes dispatching and route planning for AGVs to enhance the system performance. Experimental results on two warehouse system simulation scenarios demonstrate that the proposed method outperforms the default method.