Dynamic Deployment of Mobile Charging Trucks During Natural Disaster Evacuation: An Offline-to-Online Framework

2026-05-13 · Source: cs.MA updates on arXiv.org · Field: Transportation & Mobility — Autonomous Vehicles & Smart Transportation, Logistics & Freight Transportation, Transportation Infrastructure · Depth: Expert, extended

Summary

This study introduces the Adaptive Risk-aware Mobile Charging Truck Deployment (ARMD) framework, an offline-to-online system designed to dynamically deploy mobile charging trucks (MCTs) during large-scale electric vehicle (EV) evacuations. The framework aims to mitigate queue-induced risk exposure at fixed charging stations (FCSs) by addressing challenges like adaptive deployment, decentralized decision-making, and anticipatory network response. ARMD divides the problem into risk-aware MCT allocation among FCSs and dynamic routing of MCTs. It formulates allocation as a decentralized partially observable Markov decision process (Dec-POMDP) and uses a multi-agent proximal policy optimization (MAPPO)-based policy, pre-trained offline and refined online. For routing, a spatio-temporal travel time predictor (STPM) supports rolling-horizon route updates. Evaluated in a simulated hurricane evacuation in Hillsborough County, Florida, ARMD reduced average risk exposure by up to 71.1% in demand perturbation scenarios and 39.3% to 60.5% in infrastructure failure scenarios compared to baselines without MCTs, demonstrating its effectiveness and robustness.

Key takeaway

For emergency management agencies planning for large-scale EV evacuations, implementing the ARMD framework can significantly enhance charging infrastructure resilience. You should consider tailoring the simulator with local data to pre-train allocation policies, as this approach consistently outperforms static and rolling-horizon optimization methods, especially under severe demand perturbations and infrastructure failures. This adaptive system will enable more flexible and timely mobile charging support, reducing evacuee risk exposure.

Key insights

Dynamic deployment of mobile charging trucks significantly reduces EV evacuee risk during disasters.

Principles

• Combine offline learning with online adaptation for robust real-time decision-making.
• Decentralized control with periodic global observation enhances scalability and coordination.
• Anticipatory routing using spatio-temporal predictions improves response efficiency.

Method

The ARMD framework uses a MAPPO-based policy for decentralized MCT allocation, pre-trained offline and fine-tuned online. It integrates a spatio-temporal travel time predictor for rolling-horizon route updates, minimizing queue-induced risk exposure.

In practice

• Pre-train policies in simulators using local network data.
• Implement retrieval-augmented online fine-tuning for policy adaptation.
• Utilize dynamic graph convolution and global attention for travel time prediction.

Topics

• Mobile Charging Trucks
• Electric Vehicle Evacuation
• Decentralized POMDP
• Multi-Agent Reinforcement Learning
• Spatio-Temporal Travel Time Prediction

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Impacts of Electric Vehicle Charging Regimes and Infrastructure Deployments on System Performance: An Agent-Based Study — Effective EV charging infrastructure planning requires considering user behavior and multiple charging regimes.
• Adaptive mine planning under geological uncertainty: A POMDP framework for sequential decision-making — Sequential belief updating transforms geological uncertainty into an active component of value creation in mine planning.
• EvoDrive: Pareto Evolution for Safety-Critical Autonomous Driving via Self-Improving LLM Agents — EvoDrive uses LLM-based agentic evolution to generate diverse, safety-critical autonomous driving scenarios by balancing adversariality and realism.
• Multi-Agent Reinforcement Learning from Delayed Marketplace Feedback for Objective-Weight Adaptation in Three-Sided Dispatch — Reinforcement learning can safely adapt dispatch objective weights in complex, real-world marketplaces using delayed feedback.
• Transit Network Design with Two-Level Demand Uncertainties: A Machine Learning and Contextual Stochastic Optimization Framework — Integrating machine learning with contextual stochastic optimization improves transit network design by modeling two levels of demand uncertainty.
• Smarter Charging Could Add Years to Your Battery’s Life — An adaptive EV charging algorithm extends battery life by adjusting current based on real-time state of health.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.MA updates on arXiv.org.