Simulation Modelling Practice and Theory最新文献_第3页

A GIS-integrated agent-based simulation framework for modeling and evaluation of police patrol operations 基于gis集成agent的警察巡逻行动建模与评估仿真框架

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-25 DOI: 10.1016/j.simpat.2025.103233

Yasaman Ghasemi , Yuan Zhou , Sina Zare , Victoria C.P. Chen

This study presents a Geographic Information Systems (GIS)-integrated agent-based simulation (ABS) framework designed to evaluate police patrol deployment and shift scheduling under realistic operational constraints. The model integrates empirical Intergraph Computer-Aided Dispatch (I/CAD) data, GIS-based travel-time routing, and shift-level scheduling logic within a unified ABS environment. It captures dynamic interactions among patrol units, incident locations, and time-varying service demand.

A series of scenario-based experiments investigate the effects of key operational parameters, shift length (8-hour vs. 10-hour), patrol force size, and routing logic (shortest vs. fastest path) on system performance indicators such as response time, officer utilization, and workload balance. Results show that 10-hour shifts consistently improve response efficiency compared to 8-hour shifts, while larger patrol sizes enhance workload equity without significantly reducing delays. The model also quantifies the trade-offs between workforce expansion and scheduling strategy.

The simulation is calibrated using real-world patrol data from the Arlington Police Department, Texas, and validated through both historical benchmarks and synthetic call-arrival profiles. The model offers a configurable and adaptable simulation-based planning framework for urban public-service operations. The proposed framework demonstrates how agent-based simulation, enriched with spatial routing and empirical scheduling data, can support tactical decision-making in complex, service-driven systems.

本研究提出了一个地理信息系统（GIS）集成的基于agent的仿真（ABS）框架，旨在评估现实操作约束下的警察巡逻部署和轮班调度。该模型在统一的ABS环境中集成了经验的Intergraph计算机辅助调度（I/CAD）数据、基于gis的行程时间路由和班次级调度逻辑。它捕获巡逻单位、事件地点和时变服务需求之间的动态交互。一系列基于场景的实验研究了关键操作参数、轮班长度（8小时vs 10小时）、巡逻部队规模和路由逻辑（最短vs最快路径）对系统性能指标（如响应时间、警官利用率和工作负载平衡）的影响。结果表明，与8小时轮班相比，10小时轮班可以持续提高响应效率，而更大的巡逻规模可以提高工作量公平，但不会显著减少延误。该模型还量化了劳动力扩张和调度策略之间的权衡。模拟使用来自德克萨斯州阿灵顿警察局的真实巡逻数据进行校准，并通过历史基准和合成呼叫到达配置文件进行验证。该模型为城市公共服务运营提供了一个可配置和适应性强的基于仿真的规划框架。提出的框架展示了基于智能体的仿真，丰富了空间路由和经验调度数据，如何在复杂的服务驱动系统中支持战术决策。

{"title":"A GIS-integrated agent-based simulation framework for modeling and evaluation of police patrol operations","authors":"Yasaman Ghasemi , Yuan Zhou , Sina Zare , Victoria C.P. Chen","doi":"10.1016/j.simpat.2025.103233","DOIUrl":"10.1016/j.simpat.2025.103233","url":null,"abstract":"<div><div>This study presents a Geographic Information Systems (GIS)-integrated agent-based simulation (ABS) framework designed to evaluate police patrol deployment and shift scheduling under realistic operational constraints. The model integrates empirical Intergraph Computer-Aided Dispatch (I/CAD) data, GIS-based travel-time routing, and shift-level scheduling logic within a unified ABS environment. It captures dynamic interactions among patrol units, incident locations, and time-varying service demand.</div><div>A series of scenario-based experiments investigate the effects of key operational parameters, shift length (8-hour vs. 10-hour), patrol force size, and routing logic (shortest vs. fastest path) on system performance indicators such as response time, officer utilization, and workload balance. Results show that 10-hour shifts consistently improve response efficiency compared to 8-hour shifts, while larger patrol sizes enhance workload equity without significantly reducing delays. The model also quantifies the trade-offs between workforce expansion and scheduling strategy.</div><div>The simulation is calibrated using real-world patrol data from the Arlington Police Department, Texas, and validated through both historical benchmarks and synthetic call-arrival profiles. The model offers a configurable and adaptable simulation-based planning framework for urban public-service operations. The proposed framework demonstrates how agent-based simulation, enriched with spatial routing and empirical scheduling data, can support tactical decision-making in complex, service-driven systems.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"147 ","pages":"Article 103233"},"PeriodicalIF":3.5,"publicationDate":"2025-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145738583","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Optimization of urban mobility processes through the integration of process mining 通过一体化流程挖掘优化城市交通流程

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-25 DOI: 10.1016/j.simpat.2025.103232

Selsabil Ines Bouhidel, Nabil Belala

We introduce a dual-log process mining approach for jointly modeling and optimizing behaviors in Vehicular Ad Hoc Networks (VANETs) and urban road traffic. Simulation event logs from SUMO (traffic dynamics) and NS2 (network communications) are synchronized, preprocessed, and mined using Fuzzy Miner and Petri-net discovery in the ProM tool to produce interpretable process models. These models uncover critical anomalies, congestion hotspots, CO

_{2}

emissions peaks, and packet-delivery bottlenecks and drive a continuous feedback loop that adaptively tunes routing protocols and eco-driving strategies in real-time. Experimental evaluation demonstrated the framework’s effectiveness in identifying recurring high-emission behaviors, communication bottlenecks, and incomplete packet flows across a large-scale VANET and traffic simulation dataset. The process models significantly improved behavioral interpretability and reduced the time required for manual analysis and anomaly tracing. Future work will extend this approach with predictive modules and online mining capabilities for enhanced adaptability in dynamic VANET environments.

我们介绍了一种双日志过程挖掘方法，用于联合建模和优化车辆自组织网络（VANETs）和城市道路交通中的行为。来自SUMO（流量动态）和NS2（网络通信）的仿真事件日志被同步、预处理，并使用ProM工具中的模糊Miner和Petri-net发现进行挖掘，以产生可解释的过程模型。这些模型揭示了关键异常、拥堵热点、二氧化碳排放峰值和数据包传输瓶颈，并驱动了一个持续的反馈循环，自适应地实时调整路由协议和生态驾驶策略。实验评估证明了该框架在识别大规模VANET和流量模拟数据集上反复出现的高排放行为、通信瓶颈和不完整数据包流方面的有效性。过程模型显著提高了行为的可解释性，减少了手工分析和异常跟踪所需的时间。未来的工作将通过预测模块和在线挖掘功能扩展这种方法，以增强动态VANET环境的适应性。

{"title":"Optimization of urban mobility processes through the integration of process mining","authors":"Selsabil Ines Bouhidel, Nabil Belala","doi":"10.1016/j.simpat.2025.103232","DOIUrl":"10.1016/j.simpat.2025.103232","url":null,"abstract":"<div><div>We introduce a dual-log process mining approach for jointly modeling and optimizing behaviors in Vehicular Ad Hoc Networks (VANETs) and urban road traffic. Simulation event logs from SUMO (traffic dynamics) and NS2 (network communications) are synchronized, preprocessed, and mined using Fuzzy Miner and Petri-net discovery in the ProM tool to produce interpretable process models. These models uncover critical anomalies, congestion hotspots, CO<span><math><msub><mrow></mrow><mrow><mn>2</mn></mrow></msub></math></span> emissions peaks, and packet-delivery bottlenecks and drive a continuous feedback loop that adaptively tunes routing protocols and eco-driving strategies in real-time. Experimental evaluation demonstrated the framework’s effectiveness in identifying recurring high-emission behaviors, communication bottlenecks, and incomplete packet flows across a large-scale VANET and traffic simulation dataset. The process models significantly improved behavioral interpretability and reduced the time required for manual analysis and anomaly tracing. Future work will extend this approach with predictive modules and online mining capabilities for enhanced adaptability in dynamic VANET environments.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103232"},"PeriodicalIF":3.5,"publicationDate":"2025-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145623831","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Scheduling mixed workloads with security requirements in a cloud-fog-mist computing environment 在云雾计算环境中调度具有安全需求的混合工作负载

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-22 DOI: 10.1016/j.simpat.2025.103231

Helen D. Karatza

Cooperating cloud-fog-mist computing frameworks have been methodically designed to balance computational efficiency and data privacy during the execution of complex applications with diverse security demands. To guarantee the proper execution of these applications, the implementation of security-aware scheduling strategies is crucial. This paper explores security-aware scheduling policies, with a focus on developing algorithms tailored for heterogeneous workloads, including both simple single-task jobs and Bags of Linear Workflows (BoLWs) with varying priority levels. Multi-criteria scheduling algorithms are utilized to handle tasks by priority in the three layers. These algorithms are evaluated under different conditions, including varying system utilization, security requirements, and task service demands. Building on the epoch policy discussed in prior research, which considers job security levels, we propose an enhanced epoch-based approach that also accounts for the number of virtual machines allocated to each BoLW job alongside its security requirements. Simulation results demonstrate the superior performance of this novel epoch strategy compared to the previously established approach.

协作的云-雾-雾计算框架被有条不紊地设计为在执行具有不同安全需求的复杂应用程序期间平衡计算效率和数据隐私。为了保证这些应用程序的正确执行，安全感知调度策略的实现至关重要。本文探讨了安全感知调度策略，重点是开发针对异构工作负载的算法，包括简单的单任务作业和具有不同优先级级别的线性工作流包（bolw）。在三层中采用多准则调度算法按优先级处理任务。这些算法在不同的条件下进行评估，包括不同的系统利用率、安全需求和任务服务需求。在先前研究中讨论的epoch策略的基础上（该策略考虑了作业安全级别），我们提出了一种增强的基于epoch的方法，该方法还考虑了分配给每个BoLW作业的虚拟机数量及其安全需求。仿真结果表明，与已有的方法相比，该策略具有更好的性能。

{"title":"Scheduling mixed workloads with security requirements in a cloud-fog-mist computing environment","authors":"Helen D. Karatza","doi":"10.1016/j.simpat.2025.103231","DOIUrl":"10.1016/j.simpat.2025.103231","url":null,"abstract":"<div><div>Cooperating cloud-fog-mist computing frameworks have been methodically designed to balance computational efficiency and data privacy during the execution of complex applications with diverse security demands. To guarantee the proper execution of these applications, the implementation of security-aware scheduling strategies is crucial. This paper explores security-aware scheduling policies, with a focus on developing algorithms tailored for heterogeneous workloads, including both simple single-task jobs and Bags of Linear Workflows (BoLWs) with varying priority levels. Multi-criteria scheduling algorithms are utilized to handle tasks by priority in the three layers. These algorithms are evaluated under different conditions, including varying system utilization, security requirements, and task service demands. Building on the epoch policy discussed in prior research, which considers job security levels, we propose an enhanced epoch-based approach that also accounts for the number of virtual machines allocated to each BoLW job alongside its security requirements. Simulation results demonstrate the superior performance of this novel epoch strategy compared to the previously established approach.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103231"},"PeriodicalIF":3.5,"publicationDate":"2025-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145623771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Simulation and evaluation of a hybrid trust–cryptographic protocol for UAV swarm communications 无人机群通信中混合信任-密码协议的仿真与评估

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-20 DOI: 10.1016/j.simpat.2025.103230

Raju Singh

In mission-critical environments that require secure, scalable, and resource-efficient communication, Flying Ad Hoc Networks (FANETs) are increasing in utility. This paper proposed a Python-based simulation framework to analyse a Hybrid Trust–Cryptographic (HTC) protocol designed for unmanned aerial vehicle (UAV) swarm networks. The framework couples’ lightweight cryptographic primitives: Elliptic Curve Cryptography (ECC), AES-GCM, and ECDSA, with an adaptive trust management mechanism that qualifies UAV behaviour in a dynamic way. The trust–key coupling strategy is feedback-driven; declining trust will evoke key refresh or revocation on a pre-emptive basis to address the threats of collusion and insider attacks. Parameter values are validated against existing available cryptographic profiling benchmarks on embedded hardware platforms to ensure realism in modelling computational cost. The simulation environment is built under Gauss–Markov mobility and probabilistic attack model and has scalability with UAV nodes up to 200. The results show an increase in resilience and efficiency with almost 14 % higher packet delivery ratio, 17 % lower end-to-end latency, and 92 % of malicious node detection accuracy, also keeping energy overhead below 15 %. These results establish that adaptive trust evaluation coupled with lightweight cryptographic operations creates an optimal trade-off between security assurance and system performance. With an emphasis on reproducibility, this proposed simulation framework should thus serve as a benchmark for future research into secure communication systems for large-scale UAV swarms.

在需要安全、可扩展和资源高效通信的关键任务环境中，飞行自组织网络（fanet）的效用越来越大。本文提出了一种基于python的仿真框架来分析为无人机（UAV）群网络设计的混合信任-密码（HTC）协议。该框架将轻量级密码原语：椭圆曲线密码（ECC）、AES-GCM和ECDSA与自适应信任管理机制耦合在一起，该机制以动态方式限定无人机的行为。信任-密钥耦合策略是反馈驱动的；信任的下降将在先发制人的基础上唤起密钥更新或撤销，以解决共谋和内部攻击的威胁。参数值根据嵌入式硬件平台上现有可用的加密分析基准进行验证，以确保建模计算成本的真实性。仿真环境采用高斯-马尔可夫机动和概率攻击模型，具有200个节点的可扩展性。结果表明，弹性和效率都有所提高，数据包传递率提高了近14%，端到端延迟降低了17%，恶意节点检测准确率提高了92%，同时能源开销也保持在15%以下。这些结果表明，自适应信任评估与轻量级加密操作相结合，可以在安全保证和系统性能之间实现最佳权衡。由于强调可重复性，因此，该提出的仿真框架应作为未来大规模无人机群安全通信系统研究的基准。

{"title":"Simulation and evaluation of a hybrid trust–cryptographic protocol for UAV swarm communications","authors":"Raju Singh","doi":"10.1016/j.simpat.2025.103230","DOIUrl":"10.1016/j.simpat.2025.103230","url":null,"abstract":"<div><div>In mission-critical environments that require secure, scalable, and resource-efficient communication, Flying Ad Hoc Networks (FANETs) are increasing in utility. This paper proposed a Python-based simulation framework to analyse a Hybrid Trust–Cryptographic (HTC) protocol designed for unmanned aerial vehicle (UAV) swarm networks. The framework couples’ lightweight cryptographic primitives: Elliptic Curve Cryptography (ECC), AES-GCM, and ECDSA, with an adaptive trust management mechanism that qualifies UAV behaviour in a dynamic way. The trust–key coupling strategy is feedback-driven; declining trust will evoke key refresh or revocation on a pre-emptive basis to address the threats of collusion and insider attacks. Parameter values are validated against existing available cryptographic profiling benchmarks on embedded hardware platforms to ensure realism in modelling computational cost. The simulation environment is built under Gauss–Markov mobility and probabilistic attack model and has scalability with UAV nodes up to 200. The results show an increase in resilience and efficiency with almost 14 % higher packet delivery ratio, 17 % lower end-to-end latency, and 92 % of malicious node detection accuracy, also keeping energy overhead below 15 %. These results establish that adaptive trust evaluation coupled with lightweight cryptographic operations creates an optimal trade-off between security assurance and system performance. With an emphasis on reproducibility, this proposed simulation framework should thus serve as a benchmark for future research into secure communication systems for large-scale UAV swarms.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103230"},"PeriodicalIF":3.5,"publicationDate":"2025-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145623773","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Exploring the complexity of pedestrian dynamics in skiing: A modelling and simulation framework 探索滑雪中行人动力学的复杂性：一个建模和仿真框架

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-14 DOI: 10.1016/j.simpat.2025.103225

Buchuan Zhang , Chuan-Zhi Thomas Xie

As a distinct form of pedestrian motion, skiing possesses a long-standing history, yet the recurrent occurrence of ski-related accidents underscores the necessity of deeper inquiry into this dynamic system. In light of such a need, the present study adopts a modelling and simulation perspective to construct a framework for analysing skier trajectories and performance, with explicit consideration of the complex interactions between human behaviour, varying environmental and physical conditions. To this end, in specific, a cellular automaton (CA)-based model was developed, incorporating six critical factors: slope angle, surface friction, boundary constraints, terrain curvature, aerodynamic drag, and directional inertia. Probabilistic decision rules combined with physics-based speed updates enabled realistic skier movement simulations across a discretized slope grid. The simulation shows that slope angle predominantly drives skier speed, while surface friction and aerodynamic drag reduce efficiency by increasing resistance and prolonging descent. Boundary effects, though minor under wide-slope conditions, help confine lateral motion and influence path shaping. Terrain curvature impacts turning dynamics, especially on rough or irregular surfaces, while inertia enhances straight-line speed but reduces adaptability. The study underscores the importance of capturing both environmental and behavioural interactions to accurately model downhill skiing dynamics and provides detailed insights into the mechanisms shaping skiing efficiency, offering a powerful tool for advanced skier simulation and slope performance analysis. This study presents a cellular automaton (CA)-based modelling framework for simulating skier dynamics. Model integrates six environmental factors – slope, friction, boundary, curvature, aerodynamic drag, and inertia – to reproduce realistic motion patterns on alpine slopes. This study primarily focuses on the dynamics of a single skier, while multi-agent interactions will be explored in future work.

作为一种独特的行人运动形式，滑雪有着悠久的历史，然而与滑雪有关的事故的反复发生强调了对这一动态系统进行更深入研究的必要性。鉴于这种需求，本研究采用建模和仿真的角度来构建分析滑雪者轨迹和表现的框架，明确考虑人类行为、变化的环境和物理条件之间复杂的相互作用。为此，具体而言，开发了一个基于元胞自动机（CA）的模型，该模型包含六个关键因素：斜坡角、表面摩擦、边界约束、地形曲率、气动阻力和方向惯性。概率决策规则与基于物理的速度更新相结合，可以在离散的斜坡网格上模拟真实的滑雪者运动。模拟结果表明，坡角主要驱动滑雪者速度，而表面摩擦和气动阻力通过增加阻力和延长下降时间来降低效率。边界效应虽然在宽坡度条件下较小，但有助于限制横向运动并影响路径形成。地形曲率影响转向动力学，特别是在粗糙或不规则的表面，而惯性提高了直线速度，但降低了适应性。该研究强调了捕捉环境和行为相互作用的重要性，以准确地模拟下坡滑雪动力学，并提供了形成滑雪效率的机制的详细见解，为高级滑雪者模拟和斜坡性能分析提供了强大的工具。本研究提出了一种基于元胞自动机（CA）的模拟滑雪者动力学的建模框架。模型集成了六个环境因素-坡度，摩擦，边界，曲率，空气动力学阻力和惯性-再现现实的运动模式在高山斜坡上。本研究主要关注单个滑雪者的动态，而多智能体交互将在未来的工作中进行探索。

{"title":"Exploring the complexity of pedestrian dynamics in skiing: A modelling and simulation framework","authors":"Buchuan Zhang , Chuan-Zhi Thomas Xie","doi":"10.1016/j.simpat.2025.103225","DOIUrl":"10.1016/j.simpat.2025.103225","url":null,"abstract":"<div><div>As a distinct form of pedestrian motion, skiing possesses a long-standing history, yet the recurrent occurrence of ski-related accidents underscores the necessity of deeper inquiry into this dynamic system. In light of such a need, the present study adopts a modelling and simulation perspective to construct a framework for analysing skier trajectories and performance, with explicit consideration of the complex interactions between human behaviour, varying environmental and physical conditions. To this end, in specific, a cellular automaton (CA)-based model was developed, incorporating six critical factors: slope angle, surface friction, boundary constraints, terrain curvature, aerodynamic drag, and directional inertia. Probabilistic decision rules combined with physics-based speed updates enabled realistic skier movement simulations across a discretized slope grid. The simulation shows that slope angle predominantly drives skier speed, while surface friction and aerodynamic drag reduce efficiency by increasing resistance and prolonging descent. Boundary effects, though minor under wide-slope conditions, help confine lateral motion and influence path shaping. Terrain curvature impacts turning dynamics, especially on rough or irregular surfaces, while inertia enhances straight-line speed but reduces adaptability. The study underscores the importance of capturing both environmental and behavioural interactions to accurately model downhill skiing dynamics and provides detailed insights into the mechanisms shaping skiing efficiency, offering a powerful tool for advanced skier simulation and slope performance analysis. This study presents a cellular automaton (CA)-based modelling framework for simulating skier dynamics. Model integrates six environmental factors – slope, friction, boundary, curvature, aerodynamic drag, and inertia – to reproduce realistic motion patterns on alpine slopes. This study primarily focuses on the dynamics of a single skier, while multi-agent interactions will be explored in future work.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103225"},"PeriodicalIF":3.5,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

LMP-Opt: A simulation-based hybrid model for dynamic job scheduling and energy optimization in serverless computing LMP-Opt：一种基于仿真的混合模型，用于无服务器计算中的动态作业调度和能源优化

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-13 DOI: 10.1016/j.simpat.2025.103227

Jasmine Kaur , Inderveer Chana, Anju Bala

Serverless computing has revolutionized cloud platforms by enabling developers to deploy applications without the burden of managing infrastructure. However, challenges such as workload unpredictability, inefficient job scheduling, and high energy consumption remain critical concerns. To address these issues, this paper introduces LMP-Opt, a simulation-driven hybrid model that integrates Long Short-Term Memory (LSTM) for workload prediction, Multi-Agent Deep Q-Learning (MADQL) for job scheduling, and Proximal Policy Optimization (PPO) for fine-tuning scheduling decisions. Firstly, LSTM predicts workload patterns by capturing temporal dependencies, enabling efficient resource provisioning, and reducing performance degradation caused by unpredictable workloads. Secondly, MADQL utilizes multiple agents to optimize job scheduling by dynamically adjusting allocation strategies in response to workload fluctuations. Third, PPO refines scheduling policies by balancing exploration and exploitation, improving stability and convergence in decision-making. The proposed approach has been validated using ServerlessSimPro, a specialized simulation environment, and is further tested in AWS Lambda to ensure applicability to real-world serverless platforms. Extensive experiments using an e-commerce transaction processing workload demonstrate that LMP-Opt significantly improves system performance. The simulation results show a reduction in the average response time by 4.79% over MADQL and 6.09% over PPO, in addition to savings in energy consumption of 4.35% and 6.14%, respectively. The model also improves cost efficiency, CPU utilization, and resource scalability by reducing node requirements. These results confirm the value of hybrid learning-based simulation models for optimizing scheduling and energy efficiency in serverless computing environments.

无服务器计算使开发人员能够部署应用程序而无需管理基础设施，从而彻底改变了云平台。然而，诸如工作负载不可预测性、低效的作业调度和高能耗等挑战仍然是关键问题。为了解决这些问题，本文引入了LMP-Opt，这是一种仿真驱动的混合模型，它集成了用于工作负载预测的长短期记忆（LSTM），用于作业调度的多代理深度q -学习（MADQL）和用于微调调度决策的近端策略优化（PPO）。首先，LSTM通过捕获时间依赖性、支持有效的资源供应和减少不可预测的工作负载导致的性能下降来预测工作负载模式。其次，MADQL利用多agent根据工作负载的波动动态调整分配策略来优化作业调度。第三，PPO通过平衡勘探和开采，提高决策的稳定性和收敛性来细化调度策略。所提出的方法已经使用专门的模拟环境ServerlessSimPro进行了验证，并在AWS Lambda中进行了进一步测试，以确保适用于现实世界的无服务器平台。使用电子商务事务处理工作负载的大量实验表明，LMP-Opt显著提高了系统性能。仿真结果表明，与MADQL相比，平均响应时间减少了4.79%，与PPO相比减少了6.09%，此外还分别节省了4.35%和6.14%的能耗。该模型还通过减少节点需求来提高成本效率、CPU利用率和资源可伸缩性。这些结果证实了基于混合学习的仿真模型在无服务器计算环境中优化调度和能源效率的价值。

{"title":"LMP-Opt: A simulation-based hybrid model for dynamic job scheduling and energy optimization in serverless computing","authors":"Jasmine Kaur , Inderveer Chana, Anju Bala","doi":"10.1016/j.simpat.2025.103227","DOIUrl":"10.1016/j.simpat.2025.103227","url":null,"abstract":"<div><div>Serverless computing has revolutionized cloud platforms by enabling developers to deploy applications without the burden of managing infrastructure. However, challenges such as workload unpredictability, inefficient job scheduling, and high energy consumption remain critical concerns. To address these issues, this paper introduces LMP-Opt, a simulation-driven hybrid model that integrates Long Short-Term Memory (LSTM) for workload prediction, Multi-Agent Deep Q-Learning (MADQL) for job scheduling, and Proximal Policy Optimization (PPO) for fine-tuning scheduling decisions. Firstly, LSTM predicts workload patterns by capturing temporal dependencies, enabling efficient resource provisioning, and reducing performance degradation caused by unpredictable workloads. Secondly, MADQL utilizes multiple agents to optimize job scheduling by dynamically adjusting allocation strategies in response to workload fluctuations. Third, PPO refines scheduling policies by balancing exploration and exploitation, improving stability and convergence in decision-making. The proposed approach has been validated using ServerlessSimPro, a specialized simulation environment, and is further tested in AWS Lambda to ensure applicability to real-world serverless platforms. Extensive experiments using an e-commerce transaction processing workload demonstrate that LMP-Opt significantly improves system performance. The simulation results show a reduction in the average response time by 4.79% over MADQL and 6.09% over PPO, in addition to savings in energy consumption of 4.35% and 6.14%, respectively. The model also improves cost efficiency, CPU utilization, and resource scalability by reducing node requirements. These results confirm the value of hybrid learning-based simulation models for optimizing scheduling and energy efficiency in serverless computing environments.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103227"},"PeriodicalIF":3.5,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529298","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

HCGN: A Hierarchical Causal-Graph Network for sustainable communication and coordination in edge–fog systems 边缘雾系统中可持续通信与协调的层次因果图网络

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-13 DOI: 10.1016/j.simpat.2025.103229

Shahed Almobydeen , Gaith Rjoub , Jamal Bentahar , Ahmad Irjoob , Muhammad Younas

In cloud computing systems, the proliferation of intelligent edge devices necessitates novel communication and coordination protocols that can operate under significant bandwidth and latency constraints. This necessity is driven not only by performance requirements but also by the growing imperative for sustainable computing, as inefficient communication is a primary driver of resources consumption in large-scale systems. This paper introduces the Hierarchical and Causal-Graph Network (HCGN), a framework designed for efficient, sustainable, and decentralized decision-making in large-scale edge computing environments. HCGN integrates a hierarchical control paradigm, mapping naturally to edge-fog architectures, with a Graph Neural Network (GNN) that learns a bandwidth-efficient communication policy between edge nodes. Furthermore, a novel Causal Credit Assignment Module (CCAM) enables intelligent and sustainable resource allocation by quantifying each node’s true causal contribution to system-wide objectives, ensuring that computational and communication resources are directed to the most effective parts of the network. We demonstrate through extensive simulations, including a novel edge-based collaborative video analytics task, that HCGN significantly outperforms traditional communication protocols in terms of task success rate, communication overhead, and robustness to network degradation. Our results validate HCGN as a scalable and resource-aware solution building the next generation of sustainable decentralized edge-fog-based systems.

在云计算系统中，智能边缘设备的激增需要能够在显著带宽和延迟限制下运行的新型通信和协调协议。这种必要性不仅受到性能需求的驱动，还受到对可持续计算日益增长的需求的驱动，因为低效的通信是大规模系统中资源消耗的主要驱动因素。本文介绍了层次和因果图网络（HCGN），这是一个为大规模边缘计算环境中高效、可持续和分散决策而设计的框架。HCGN集成了分层控制范式，自然映射到边缘雾架构，并使用图神经网络（GNN）学习边缘节点之间的带宽高效通信策略。此外，一个新颖的因果信用分配模块（CCAM）通过量化每个节点对系统范围目标的真正因果贡献来实现智能和可持续的资源分配，确保计算和通信资源被定向到网络中最有效的部分。我们通过广泛的模拟，包括一个新的基于边缘的协作视频分析任务，证明了HCGN在任务成功率、通信开销和对网络退化的鲁棒性方面明显优于传统通信协议。我们的研究结果验证了HCGN是一种可扩展和资源感知的解决方案，可以构建下一代可持续的分散式边缘雾系统。

{"title":"HCGN: A Hierarchical Causal-Graph Network for sustainable communication and coordination in edge–fog systems","authors":"Shahed Almobydeen , Gaith Rjoub , Jamal Bentahar , Ahmad Irjoob , Muhammad Younas","doi":"10.1016/j.simpat.2025.103229","DOIUrl":"10.1016/j.simpat.2025.103229","url":null,"abstract":"<div><div>In cloud computing systems, the proliferation of intelligent edge devices necessitates novel communication and coordination protocols that can operate under significant bandwidth and latency constraints. This necessity is driven not only by performance requirements but also by the growing imperative for sustainable computing, as inefficient communication is a primary driver of resources consumption in large-scale systems. This paper introduces the Hierarchical and Causal-Graph Network (HCGN), a framework designed for efficient, sustainable, and decentralized decision-making in large-scale edge computing environments. HCGN integrates a hierarchical control paradigm, mapping naturally to edge-fog architectures, with a Graph Neural Network (GNN) that learns a bandwidth-efficient communication policy between edge nodes. Furthermore, a novel Causal Credit Assignment Module (CCAM) enables intelligent and sustainable resource allocation by quantifying each node’s true causal contribution to system-wide objectives, ensuring that computational and communication resources are directed to the most effective parts of the network. We demonstrate through extensive simulations, including a novel edge-based collaborative video analytics task, that HCGN significantly outperforms traditional communication protocols in terms of task success rate, communication overhead, and robustness to network degradation. Our results validate HCGN as a scalable and resource-aware solution building the next generation of sustainable decentralized edge-fog-based systems.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103229"},"PeriodicalIF":3.5,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529385","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The impact of import container flow characteristics on port operational efficiency 进口集装箱流动特性对港口作业效率的影响

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-12 DOI: 10.1016/j.simpat.2025.103228

Agostino Bruzzone , Alessia Giulianetti , Marco Gotelli , Anna Sciomachen

In this paper, we analyze different scenarios for container flows arriving at marine terminals to different destinations in the hinterland. The aim of the study is to verify how the type of import containers — standard, hazardous, and refrigerated — and their size affect the operational efficiency of the terminal. Relevant performance indicators, such as container dwell time, average and maximum number of waiting containers, and equipment utilization rate, are evaluated. To this end, we present a discrete-event simulation study that, although generalizable to any port, refers to a terminal in the port network of Genoa (Italy). The number of considered scenarios, illustrated in this paper, are taken from a synthetic data generator for logistics flows and used in Witness Horizon v.24 simulation software environment to execute independent runs at a steady state condition. To the authors’ knowledge, this is the first time that a sensitivity analysis based on the variation in the types of containers is presented. The performed simulation experiments can be of great interest to various port stakeholders. Indeed, the results show that the percentage composition of the type of import container over the annual time horizon considered has an impact on the indicators under analysis, favoring a more balanced distribution. However, again in relation to the same indicators, the variation in container size appears to be negligible. The study highlights how advance knowledge of the type of import containers can support port terminal management in terms of efficient management and optimization of resources, providing specific advice on the operational decisions concerning equipment and block yard allocation.

在本文中，我们分析了集装箱流到达内陆不同目的地的不同情景。研究的目的是验证进口货柜的种类（标准货柜、危险货柜及冷藏货柜）及其大小如何影响码头的运作效率。评估集装箱停留时间、平均和最大等待集装箱数、设备利用率等相关性能指标。为此，我们提出了一个离散事件模拟研究，虽然可以推广到任何港口，但指的是热那亚（意大利）港口网络中的一个码头。本文中所示的考虑场景的数量取自物流流的合成数据生成器，并在Witness Horizon v.24仿真软件环境中用于在稳态条件下执行独立运行。据作者所知，这是第一次提出基于容器类型变化的敏感性分析。所进行的模拟实验可以引起各种港口利益相关者的极大兴趣。事实上，结果表明，在所考虑的年度时间范围内，进口集装箱类型的百分比构成对所分析的指标有影响，有利于更平衡的分配。然而，就同样的指标而言，容器大小的变化似乎可以忽略不计。该研究强调了进口集装箱类型的预先知识如何能够在有效管理和优化资源方面支持港口码头管理，并为有关设备和堆场分配的操作决策提供具体建议。

{"title":"The impact of import container flow characteristics on port operational efficiency","authors":"Agostino Bruzzone , Alessia Giulianetti , Marco Gotelli , Anna Sciomachen","doi":"10.1016/j.simpat.2025.103228","DOIUrl":"10.1016/j.simpat.2025.103228","url":null,"abstract":"<div><div>In this paper, we analyze different scenarios for container flows arriving at marine terminals to different destinations in the hinterland. The aim of the study is to verify how the type of import containers — standard, hazardous, and refrigerated — and their size affect the operational efficiency of the terminal. Relevant performance indicators, such as container dwell time, average and maximum number of waiting containers, and equipment utilization rate, are evaluated. To this end, we present a discrete-event simulation study that, although generalizable to any port, refers to a terminal in the port network of Genoa (Italy). The number of considered scenarios, illustrated in this paper, are taken from a synthetic data generator for logistics flows and used in Witness Horizon v.24 simulation software environment to execute independent runs at a steady state condition. To the authors’ knowledge, this is the first time that a sensitivity analysis based on the variation in the types of containers is presented. The performed simulation experiments can be of great interest to various port stakeholders. Indeed, the results show that the percentage composition of the type of import container over the annual time horizon considered has an impact on the indicators under analysis, favoring a more balanced distribution. However, again in relation to the same indicators, the variation in container size appears to be negligible. The study highlights how advance knowledge of the type of import containers can support port terminal management in terms of efficient management and optimization of resources, providing specific advice on the operational decisions concerning equipment and block yard allocation.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103228"},"PeriodicalIF":3.5,"publicationDate":"2025-11-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529386","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Dynamic firefighting route planning for efficient evacuation in complex subway stations: A deep learning-enhanced robust optimization approach 复杂地铁站高效疏散的动态消防路线规划：一种深度学习增强的鲁棒优化方法

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-07 DOI: 10.1016/j.simpat.2025.103223

Jinli Wei , Chunyue Cui , Xiaoxia Yang

The enclosed spaces and high-density population in subway stations significantly complicate evacuation during fires, thus increasing the difficulty of emergency response. To enhance fire rescue capabilities, this study conducts robust optimization modeling for firefighting routes from costs of station facility layout, passenger flow distribution, smoke propagation patterns, and human resource expenditure. Firstly, the BKA-GRU deep learning method is designed to calculate passenger passage time at critical nodes such as gates, improving the rationality of firefighting route design. Secondly, a firefighting value function based on the importance of fire nodes is constructed, making the firefighting routes more conducive to efficient and safe passenger evacuation. Thirdly, a box-based intersection polyhedron uncertainty set is employed to model the uncertainties in firefighting travel time and firefighting time, enhancing the adaptability and robustness of the routes. Fourthly, the advanced Ivy algorithm combined with Gurobi is adopted to solve the developed robust optimization model, enabling rapid identification of efficient and stable firefighting routes in complex environments. Finally, both quantitative and qualitative analyses are used to comprehensively evaluate firefighting effectiveness. The results indicate that: (i) The BKA-GRU prediction model exhibits high accuracy and reliability in predicting node passage time. (ii) The robust optimization model for firefighting routes significantly reduces fire by-products, shortens passenger evacuation time, and mitigates congestion. (iii) The firefighting route design achieves significant improvements in temperature control and visibility enhancement, effectively improving the fire environment and enhancing rescue efficiency and safety. This study provides an innovative solution for fire rescue in complex environments.

地铁车站空间封闭、人口密集，使火灾时的疏散变得更加复杂，增加了应急响应的难度。为了提高消防救援能力，本研究从车站设施布局成本、客流分布成本、烟雾传播方式成本和人力资源支出成本等方面对消防路线进行鲁棒优化建模。首先，设计了BKA-GRU深度学习方法，计算登机口等关键节点的旅客通过时间，提高消防路线设计的合理性；其次，构建基于火灾节点重要性的消防价值函数，使消防路线更有利于高效、安全的乘客疏散。第三，采用基于框的交叉口多面体不确定性集对消防行程时间和消防时间的不确定性进行建模，增强了路径的自适应性和鲁棒性；第四，采用先进的Ivy算法结合Gurobi算法对所建立的鲁棒优化模型进行求解，实现了在复杂环境下快速识别高效稳定的消防路线。最后，采用定量分析和定性分析相结合的方法对消防效能进行综合评价。结果表明：(1)BKA-GRU预测模型在预测节点通过时间方面具有较高的准确性和可靠性。（ii）稳健的消防路线优化模型显著减少了火灾副产物，缩短了乘客疏散时间，缓解了拥堵。（三）消防路线设计在温度控制和能见度增强方面有明显改善，有效改善了火灾环境，提高了救援效率和安全性。本研究为复杂环境下的火灾救援提供了一种创新的解决方案。

{"title":"Dynamic firefighting route planning for efficient evacuation in complex subway stations: A deep learning-enhanced robust optimization approach","authors":"Jinli Wei , Chunyue Cui , Xiaoxia Yang","doi":"10.1016/j.simpat.2025.103223","DOIUrl":"10.1016/j.simpat.2025.103223","url":null,"abstract":"<div><div>The enclosed spaces and high-density population in subway stations significantly complicate evacuation during fires, thus increasing the difficulty of emergency response. To enhance fire rescue capabilities, this study conducts robust optimization modeling for firefighting routes from costs of station facility layout, passenger flow distribution, smoke propagation patterns, and human resource expenditure. Firstly, the BKA-GRU deep learning method is designed to calculate passenger passage time at critical nodes such as gates, improving the rationality of firefighting route design. Secondly, a firefighting value function based on the importance of fire nodes is constructed, making the firefighting routes more conducive to efficient and safe passenger evacuation. Thirdly, a box-based intersection polyhedron uncertainty set is employed to model the uncertainties in firefighting travel time and firefighting time, enhancing the adaptability and robustness of the routes. Fourthly, the advanced Ivy algorithm combined with Gurobi is adopted to solve the developed robust optimization model, enabling rapid identification of efficient and stable firefighting routes in complex environments. Finally, both quantitative and qualitative analyses are used to comprehensively evaluate firefighting effectiveness. The results indicate that: (i) The BKA-GRU prediction model exhibits high accuracy and reliability in predicting node passage time. (ii) The robust optimization model for firefighting routes significantly reduces fire by-products, shortens passenger evacuation time, and mitigates congestion. (iii) The firefighting route design achieves significant improvements in temperature control and visibility enhancement, effectively improving the fire environment and enhancing rescue efficiency and safety. This study provides an innovative solution for fire rescue in complex environments.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103223"},"PeriodicalIF":3.5,"publicationDate":"2025-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529299","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SimEdgeAI: A deep reinforcement learning framework for simulating task offloading in resource-constrained IoT networks SimEdgeAI：用于模拟资源受限物联网网络中任务卸载的深度强化学习框架

IF 3.5 2区计算机科学 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Simulation Modelling Practice and Theory

Pub Date : 2025-11-07 DOI: 10.1016/j.simpat.2025.103226

Waseem Abbass , Nasim Abbas , Uzma Majeed

The rapid growth of latency-sensitive Internet of Things (IoT) applications necessitates intelligent and scalable task offloading strategies in edge computing environments operating under dynamic workloads and limited energy resources. This paper introduces SimEdgeAI, a novel Deep Reinforcement Learning (DRL) framework that formulates task offloading as a stochastic decision-making problem over a multi-discrete action space, effectively capturing the trade-offs among local execution, edge offloading, and task dropping. The framework adopts an actor–critic architecture enhanced with a Gumbel–Softmax-based policy representation, enabling differentiable and stable learning over discrete actions. The actor network produces temperature-controlled stochastic policies, while the critic estimates long-term utilities based on system-wide features such as queue lengths, transmission delays, and energy states. A multi-objective reward function penalizing latency violations, excessive energy use, and fairness deviations guides the agent towards globally efficient and equitable offloading decisions. Extensive evaluations demonstrate that SimEdgeAI reduces average task latency by up to 35% and energy consumption by 25% compared to baseline methods including Deep Deterministic Policy Gradient (DDPG), Centralized DQN (C-DQN), and Greedy policies. It achieves over 91% deadline satisfaction and superior fairness measured by Jain’s index across edge clients. Ablation and sensitivity analyses confirm the contribution of each architectural component, while visualization studies underline the framework’s multi-objective consistency. These results highlight SimEdgeAI as an effective and fair solution for real-time, large-scale IoT–edge task offloading problems.

对延迟敏感的物联网（IoT）应用的快速增长需要在动态工作负载和有限能源下运行的边缘计算环境中采用智能和可扩展的任务卸载策略。本文介绍了SimEdgeAI，这是一个新颖的深度强化学习（DRL）框架，它将任务卸载作为一个多离散动作空间上的随机决策问题，有效地捕获了局部执行、边缘卸载和任务丢弃之间的权衡。该框架采用基于gumbel - softmax的策略表示增强的参与者-评论家体系结构，使离散行为的可微分和稳定学习成为可能。参与者网络产生温度控制的随机策略，而评论家则根据系统范围的特征（如队列长度、传输延迟和能量状态）估计长期效用。一个惩罚延迟违规、过度能源使用和公平性偏差的多目标奖励函数引导智能体做出全局高效和公平的卸载决策。广泛的评估表明，与包括深度确定性策略梯度（DDPG）、集中式DQN （C-DQN）和贪婪策略在内的基线方法相比，SimEdgeAI将平均任务延迟减少了35%，能耗减少了25%。它达到了超过91%的最后期限满意度和卓越的公平性，由Jain的跨边缘客户指数衡量。消融和敏感性分析证实了每个架构组件的贡献，而可视化研究强调了框架的多目标一致性。这些结果表明，SimEdgeAI是实时、大规模物联网边缘任务卸载问题的有效和公平的解决方案。

{"title":"SimEdgeAI: A deep reinforcement learning framework for simulating task offloading in resource-constrained IoT networks","authors":"Waseem Abbass , Nasim Abbas , Uzma Majeed","doi":"10.1016/j.simpat.2025.103226","DOIUrl":"10.1016/j.simpat.2025.103226","url":null,"abstract":"<div><div>The rapid growth of latency-sensitive Internet of Things (IoT) applications necessitates intelligent and scalable task offloading strategies in edge computing environments operating under dynamic workloads and limited energy resources. This paper introduces SimEdgeAI, a novel Deep Reinforcement Learning (DRL) framework that formulates task offloading as a stochastic decision-making problem over a multi-discrete action space, effectively capturing the trade-offs among local execution, edge offloading, and task dropping. The framework adopts an actor–critic architecture enhanced with a Gumbel–Softmax-based policy representation, enabling differentiable and stable learning over discrete actions. The actor network produces temperature-controlled stochastic policies, while the critic estimates long-term utilities based on system-wide features such as queue lengths, transmission delays, and energy states. A multi-objective reward function penalizing latency violations, excessive energy use, and fairness deviations guides the agent towards globally efficient and equitable offloading decisions. Extensive evaluations demonstrate that SimEdgeAI reduces average task latency by up to 35% and energy consumption by 25% compared to baseline methods including Deep Deterministic Policy Gradient (DDPG), Centralized DQN (C-DQN), and Greedy policies. It achieves over 91% deadline satisfaction and superior fairness measured by Jain’s index across edge clients. Ablation and sensitivity analyses confirm the contribution of each architectural component, while visualization studies underline the framework’s multi-objective consistency. These results highlight SimEdgeAI as an effective and fair solution for real-time, large-scale IoT–edge task offloading problems.</div></div>","PeriodicalId":49518,"journal":{"name":"Simulation Modelling Practice and Theory","volume":"146 ","pages":"Article 103226"},"PeriodicalIF":3.5,"publicationDate":"2025-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145529300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0