Learning to Route Electric Trucks Under Operational Uncertainty

A team of researchers from MIT and TU Delft, including Stavros Orfanoudakis, Ziyan Li, Ruixiao Yang, and others, have proposed a learning-based framework for stochastic electric truck routing under charging constraints and operational uncertainty. Published on arXiv (2604.26566), the work tackles the complex coupled logistics and energy problem where electric trucks must make routing decisions that remain feasible under limited battery range, long charging times, variable travel and energy consumption, and competition for shared charging infrastructure. The authors note that traditional heuristics-based methods are impractical at scale for this problem.

The problem is formulated as an event-driven semi-Markov decision process solved via Reinforcement Learning. Key innovations include a graph-based representation of system state and feasible decisions, along with a rule-based action mask that restricts policies to operationally admissible actions, improving training efficiency. The team also developed an event-driven simulation environment supporting both RL training and benchmarking against heuristic and mathematical programming baselines. Computational experiments across a range of fleet sizes showed that the proposed algorithm consistently outperforms baselines and attains performance close to optimization benchmarks in many settings, while preserving high success rates under charging congestion and uncertainty.