Delays to hospital discharge create system-wide pressures by reducing acute hospital capacity. As increasing numbers of higher-acuity patients are discharged into home-based care requiring multiple daily visits, modelling capacity at the interface of acute and community care becomes more complex. We introduce a visits-based, time-driven simulation model that incorporates variable daily visit requirements. We compare the simulation with tractable analytic queueing models using a consistent set of parameters, showing how each approach can provide complementary perspectives on system behaviour. This comparison highlights how simplified analytic representations may not reflect congestion dynamics when resource use varies over time, particularly at high traffic intensities. Our model has been implemented in practice by NHS partners for routine planning. We demonstrate how incorporating time-varying resource requirements into visits-based models supports understanding of system dynamics and informs workforce and capacity planning in stretched systems.
Easier access to real-time data gives the opportunity for simulation to support operational decision-making based on the current state of the actual system. We describe a problem encountered in manufacturing where a maintenance team responds to downtime events by prioritising repair jobs with the aim of optimising the throughput of a production line. Our solution uses a complex, stochastic discrete event simulation model with a neural network regression metamodel as part of a multi-fidelity simulation optimisation packaged as a digital twin. We use the structure of the problem to suggest an aggregation of the state-space in order to fit the metamodel more efficiently. Aggregation allows us to reduce the dimensionality of the system state variable and the decision space while still maintaining a satisfactory level of predictability. We describe the framework as the Aggregate Decision and State-space Multi-fidelity Metamodelling Approach (ADSMMA). Results suggest that ADSMMA, which could be generalised to other examples, makes simulation optimisation feasible for operational decision-making. In addition, where the metamodel is accurate, it can provide good results without running the high-fidelity DES model in the simulation optimisation step.
As natural disasters intensify, timely evacuation planning becomes increasingly critical. This study develops a spatial Decision Support System (DSS) for tsunami evacuation planning, grounded in a terrain-aware Discrete Event Simulation (DES) model. The DSS integrates Geographic Information Systems (GIS) with DES to capture elevation- and slope-informed pedestrian travel times, enabling context-specific evacuation analysis. Evacuation performance is assessed through three metrics: average evacuation time (efficiency), maximum individual evacuation time (equity), and evacuation success rate within a critical 20-min threshold (effectiveness). The methodology is applied to a case study of Bahia de Caraquez, Ecuador, where scenario analysis shows that improving shelter accessibility, applying terrain-aware routing, and accounting for mobility conditions enhance evacuation performance, particularly in remote and terrain-constrained areas. To extend these findings, Response Surface Methodology (RSM) and Desirability-based Optimization Methodology (DOM) are used to approximate nonlinear mobility-performance relationships and to identify operating conditions under which efficiency, equity, and effectiveness remain jointly favourable. Overall, the framework links simulation outputs to actionable planning guidance, supporting evidence-based evacuation decisions in tsunami-prone regions.
Animated visualisations of discrete event simulation (DES) models, which show entities flowing through a system and using resources, are valuable for both stakeholders and modellers. For stakeholders, who are often simulation novices, they enhance understanding and help to build trust in a model. For modellers, animations can support model verification and validation. While code-based free and open source (FOSS) modelling solutions are gaining traction, solutions for generating these animated visualisations are lacking, with users being tied to specific DES frameworks or required to generate a solution from scratch. Until now, this has meant commercial off-the-shelf (COTS) software, which has invested in animation capabilities, maintains a significant advantage. This paper introduces vidigi, a domain and framework-agnostic Python library that can generate these animations. Given simple event logs, vidigi undertakes all required data transformations to produce an animation in a format that can be embedded in interactive web interfaces, included in web-based reports, or distributed as a standalone file. By lowering the technical barrier to producing animations, vidigi addresses a crucial factor limiting the uptake of code-based FOSS modelling solutions. This helps to advance the adoption of transparent, reusable and reproducible open science techniques in the field of simulation modelling.
This study investigates how behavioral factors, grounded in the Theory of Planned Behavior (TPB), influence CEOs' resource allocation decisions between exploration and exploitation, and how these decisions impact organizational performance. An agent-based simulation model replicates the Iranian pharmaceutical market, incorporating three agent types: company agents (modeled using TPB), competitor agents (based on the Miles and Snow typology), and a market agent (structured around Porter's Five Forces). Strategy effectiveness is evaluated through the Fuzzy Delphi Method, and the model is calibrated using eight years of real-world data through a genetic algorithm. The results show that CEOs who rely more heavily on personal attitude and perceived behavioral control (PBC)-rather than subjective norms-achieve greater market share, even without increasing resources. Furthermore, adapting CEO behavior based on competitor performance (i.e. dynamic learning) significantly enhances strategic outcomes. These findings highlight the role of behavioral adaptability in fostering exploration within dynamic markets. Further empirical validation across different organizational settings is needed to enhance the model's generalizability. The study provides actionable insights for managers aiming to improve strategic agility by integrating behavioral traits and environmental dynamics into resource allocation decisions.
The study proposes a trivariate Poisson regression model based on the trivariate Poisson distribution (TPD), assuming a constant covariance. Towards this investigation, a simulated dataset involving ten thousand dependent observations from a TPD was randomly generated. Ten thousand independent observations were generated from independent standard normal distributions. Parameter estimation was performed using maximum likelihood estimation. The Newton $ - $- Raphson (NR) algorithm is used to solve the resulting equations from the maximum likelihood estimation procedure. An assessment of the efficiency and bias associated with parameter estimators generated through the NR algorithm was conducted to evaluate the quality of the resultant estimators. Additionally, computational implementation was carried out using R and Python programming environments to determine whether these programming tools generated better outcomes with comparable computational times. Results from both software environments showed similarities, with Python proving faster computational time and competitive, efficient estimators compared to R. The model proposed has future applicability across different fields that need analysis of three dependent observations together with their related common independent observations and dependencies.
Quick-commerce puts significant stress on last-mile delivery systems in dense urban environments, where disruptions such as partial hub failures, localised congestion, or workforce shortages can reduce service reliability. This study develops a simulation-based framework to examine the resilience of clustering-based customer-hub assignment strategies under urban disruption conditions using a realistic road network representation of Kolkata and behaviourally grounded synthetic demand. Fifteen widely used clustering techniques are evaluated across multiple disruption scenarios under flexible and rigid reassignment policies. System performance is assessed using On-Time Delivery (OTD) and Fill Rate, enabling resilience to be characterised in terms of performance stability and degradation behaviour. The analysis reveals pronounced heterogeneity in resilience across clustering approaches. Soft and medoid-based methods which include Fuzzy C-Means and K-Medoids achieved their best resilience performance by maintaining OTD rates between 84-86% and Fill Rates over 92% during conditions of moderate disruption. Under severe hub-capacity disruptions, these methods preserve 8-14% higher OTD and 6-11% higher Fill Rates compared with centroid-based approaches. The results demonstrate that clustering logic is a structural determinant of last-mile resilience and provide a scalable decision-support framework for designing disruption-tolerant quick-commerce delivery networks.
Shared autonomous vehicles (SAVs) are increasingly integrated into urban transport, yet effective evaluation tools for fleet sizing and charging infrastructure are lacking. This study develops an agent-based discrete event simulation model to evaluate SAV service levels and costs. Tested on the Manhattan road network, the model simulates interactions between vehicles and infrastructure agents. Key results show: (1) a heuristic charging deployment algorithm improves passenger waiting times compared to even distribution; (2) for small fleets, more charging stations significantly boost performance, though this effect diminishes once fleet size exceeds 6,600; and (3) while increasing charging stations reduces unsatisfied demand for smaller fleets, larger fleets eventually saturate demand regardless of station density. These findings offer insights for optimizing SAV fleet management and infrastructure deployment.
Validation of a simulation model is a key stage in building its credibility. It demonstrates that the model offers a reliable representation of the real system and fulfils its intended objectives. Validation is a highly discipline-dependent process. Social simulations differ from business, engineering, or physics-based simulations because usually they do not refer to an objective, well-known reality described by established, universal background theories. This causes social simulations to require a different approach to the validation process. This paper focuses on the operational validation of social simulation models and, in particular, one group of these models, namely pension models. We present the application of the model-to-model-comparison method, which involves comparing the results of two simulation models developed using different approaches and different platforms. The article discusses the rationale for using this method in the validation of social simulation models. It also presents the results of pension model validation using statistical tests and analysis of model behaviour consistency for different input parameter values.
This paper explores the evolution of the electricity market in Eastern Africa, focusing on the integration of varying renewable energy sources. The problem addressed is the current market structure's inability to support short-term trade, which is essential for accommodating fluctuations in renewable energy production and ensuring supply. The Agent-based Market model for the Investigation of Renewable and Integrated energy Systems (AMIRIS) open-source software is employed to predict market participants' behaviours in a deregulated competitive market, considering pricing, bidding strategies, and capacity expansion decisions. The results indicate that a short-term market could enhance the integration of renewable energy, reduce costs, and improve service quality, while highlighting the need for policy interventions to ensure economic viability and optimal resource utilization.
Long waiting times in the pre-examination stage of eye clinics can reduce patient satisfaction, while inefficient utilization of staff and equipment reduces operational efficiency. Effective management of this complex system is challenging due to sequential and patient-specific care pathways, non-stationary patient arrivals, and heterogeneous service times. This study employs Discrete Event Simulation (DES) to analyse multiple configurations of the pre-examination rooms, calibrated with real data from a high-volume ophthalmology clinic. The study compares flexible and dedicated resource strategies in terms of two metrics: average patient waiting time and nurse utilization. Three configurations are analysed within this context. Closing one room in the afternoon, which was simple to implement, saves one nurse-day, but has little effect on waiting. Equipping both rooms with flexible resources reduces nurse requirement by half but increases waiting times due to reduced concurrency. Finally, equipping both rooms with dedicated instruments, which reduces total waiting by roughly 20 h per month but requires more nurses. Our results show that simulation-based analysis provides a useful basis for evaluating trade-offs between flexibility, concurrency, and resource utilization, particularly in resource-limited healthcare systems.
Discrete Event Simulation has emerged as a crucial technology enabling manufacturing industries to model and optimize complex production processes in a virtual environment. This study investigates the role of simulation tools in industrial manufacturing using quantitative and qualitative survey data from professionals across several industries. The goal is to understand how simulation tools influence decision-making and what practical benefits they provide. Responses have been coded and categorized using thematic analysis. Initial findings show that production simulation improves equipment interaction assessments, supports change management through visualization, and helps identify unforeseen issues such as collision detection. The study also highlights challenges companies face in using simulation efficiently. Respondents reported that simulation improves decision-making by enabling clearer communication and better prediction of production outcomes. Views on the future of discrete event simulation are positive, with most respondents seeing it as a key tool for manufacturing digitalization over the next decade. This study presents a framework summarizing the benefits of production simulation and insights into improving and integrating it into industrial operations. The contributions highlight the need for organizational transformation to reach Industry 4.0 maturity. These findings offer useful insight for manufacturing companies considering implementing simulation technology.
Urban transportation systems increasingly depend on heterogeneous real-time data, yet most routing and scheduling methods rely on single-modal inputs and static fusion, limiting adaptability to dynamic traffic, demand, and weather conditions. This paper proposes an adaptive multimodal data fusion framework integrating Graph Neural Networks (GNNs) and Reinforcement Learning (RL) for large-scale vehicle scheduling and path planning. Multimodal inputs-including GPS, onboard sensors, traffic flow, weather, and user demand-are encoded into modality-specific features and fused through a lightweight attention-based module that dynamically estimates quality and relevance scores to generate time-varying weights with regularization. The fused representation is processed by a temporal GNN to capture multi-hop spatial dependencies and short-term dynamics, producing compact network embeddings. These embeddings form the state of a Proximal Policy Optimization (PPO) agent that learns fleet-level routing and dispatching policies under legal, capacity, energy, and time-window constraints enforced by feasibility masks and a safety layer. Experiments on two city-scale networks (up to 23,400 nodes and 10,000 vehicles) demonstrate reductions of up to 22.8% in travel time, 25% in energy use, and 43.3% in user delay, with inference latency below 200 ms, confirming scalability and robustness.
Discrete Event Simulation (DES) methods are widely used as effective tools for analysing the dynamics of systems in various industrial sectors. In the shipbuilding industry, DES is applied across diverse production management scenarios. Unlike conventional manufacturing, shipyards face the challenge of prioritizing the analysis and partial improvement of existing layouts. This need is heightened by constantly varying product and planning information, which demands a specialized methodology. This study addresses two key limitations in DES research for shipyard layout analysis: economic challenges from expensive commercial software and data-processing inefficiencies due to handling extensive input data. To overcome these, we propose an open-source based layout simulation framework. The development includes algorithms for efficient input and modelling of product information, planning data, factory/road information. User experience is further enhanced through an intuitive C# GUI seamlessly integrated with the Python-based layout simulation executable. The final application will validate its practical utility by analysing layouts of large Korean shipyards. This research offers a pragmatic and cost-effective solution tailored to shipbuilding's unique requirements, advancing DES applications in the field.
