Training deep neural networks on edge devices reduces latency and preserves privacy, but limited compute and bandwidth require a trade-off between accuracy and processing time. This paper uses the deep learning multi-processing agent-based simulation framework to model distributed deep learning coordination under heterogeneous edge conditions and quantify the trade-off between communication cost and accuracy. Deep learning multi-processing is a lightweight simulator in which each agent simulates an edge node with distinct computational power for exchanging model parameters. To simulate the network dynamics of edge learning systems, deep learning multi-processing models different communication topologies among the machine learning models hosted on edge nodes participating in federated learning, focusing on how communication strategies for exchanging model parameters affect distributed training under resource constraints. Simulating the network aspect of distributed deep learning offers a cost-effective way to evaluate training performance under realistic edge conditions during the design phase of deep learning edge-enabled systems. The findings of such simulation are valuable for industry sectors developing intelligent systems for smart cities, including video-based traffic control, surveillance analytics and sensor-driven urban monitoring, where efficient and adaptive deep learning parallelization is required at the network edge.
Connected and automated vehicles are expected to play a key role in alleviating urban traffic congestion in transportation systems. However, coordinating traffic lights and connected and automated vehicles remains a challenging problem in mixed-autonomy environments. This paper proposes a modified proximal policy optimization–based decentralized multi-agent reinforcement learning framework to jointly optimize traffic signal control and eco-driving behavior. The model includes two types of agents, namely traffic light agents and connected and automated vehicle agents. Traffic light agents aim to improve traffic efficiency by optimizing signal-phase decisions. Connected and automated vehicle agents focus on adjusting vehicle acceleration and speed to enhance driving comfort and smoothness. Separate policy networks are designed for each agent type to capture their different control objectives. The proposed framework is evaluated on a realistic urban traffic network. Simulation results show that the decentralized model (multi-agent-proximal policy optimization) achieves the best overall performance. It significantly improves average speed compared to baseline methods. It also reduces average travel time and vehicle waiting time. Further analysis shows that increasing connected and automated vehicle penetration improves overall traffic efficiency. The best performance is achieved when all vehicles are connected and automated vehicles, indicating full system observability and coordination.
This article proposes HybridDWAVO, a deterministic local-planning framework that combines the dynamics-aware trajectory generation of the dynamic window approach with the predictive collision-avoidance capability of the velocity obstacle method. The proposed framework uses finite-horizon velocity obstacle filtering as a pre-evaluation feasibility step applied to the dynamic window approach–admissible velocity set. As a result, predictively unsafe candidates are removed before dynamic window approach trajectory scoring, and the dynamic window approach objective function is evaluated only over the remaining velocity obstacle–filtered safe candidates. The framework is evaluated in controlled simulations using deterministic dynamic window approach–based planner variants under matched parameter settings, obstacle trajectories, random seeds, and computational environment. To support a simulation-based assessment, the evaluation includes implementation-relevant factors such as robot-state estimation noise, obstacle-perception uncertainty, sensing and control latency, bounded actuation, and nonholonomic velocity constraints. The analysis combines repeated-trial ablation results, obstacle-density-dependent navigation outcomes, runtime and candidate-processing metrics, and representative trajectory-level case studies. The repeated-trial results indicate that the proposed pre-evaluation velocity obstacle–filtering variant improves goal-completion consistency and reduces collision counts relative to conventional dynamic window approach under the tested dynamic-obstacle conditions. In addition, the ablation and candidate-processing results suggest that placing velocity obstacle filtering before trajectory scoring reduces the number of candidates entering dynamic window approach objective-function evaluation while maintaining favorable navigation outcomes among the tested variants; the associated runtime results are interpreted within the specified simulation and workstation environment. Overall, the results provide simulation-based algorithmic evidence for the proposed dynamic window approach–velocity obstacle integration structure, while physical robot validation and broader benchmarking under real sensing, actuation, communication, and environmental conditions remain necessary future work.
The COVID-19 pandemic placed tremendous strain on medical resources. To assess intervention strategies and resource planning for metropolitan pandemic response, a hybrid System Dynamics–Discrete-Event Simulation (SD-DES) framework is developed, using Wuhan as a case study. With SD capturing macroscopic infection dynamics and policy intervention and DES simulating medical resource scheduling, bidirectional feedback between macro and micro levels enables a more holistic and comprehensive evaluation of policy effects. By simulating 132 representative scenarios, key factors including containment measures, medical resource capacity, mask adoption, vaccine rollout speed, and the timing of external medical support were examined. Our analysis provides a comprehensive perspective on pandemic prevention and control decision-making, an aspect often underrepresented in prior studies. The model offers an extensible computational framework for complex socioeconomic systems, such as pandemic emergencies, where dynamic and process complexities coexist. It can assist policymakers in enhancing the healthcare system’s preparedness and mitigating the spread of the pandemic. Future research could improve model robustness by incorporating advanced parameter estimation techniques, such as particle filtering. To address computational challenges, we also recommend implementing asynchronous module execution and optimized programming to improve simulation efficiency and scalability.
