FleetRL: Realistic reinforcement learning environments for commercial vehicle fleets
Enzo Cording, Jagruti Thakur
Abstract
Reinforcement Learning for EV charging optimization has gained significant academic attention in recent years, due to its ability to handle uncertainty, non-linearity, and real-time problem-solving. While the number of articles published on the matter has surged, the number of open-source environments for EV charging optimization remains small, and a research gap still exists when it comes to customizable frameworks for commercial vehicle fleets. To bridge the gap between research and real-world deployment of RL-based charging optimization, this paper introduces FleetRL as the first customizable RL environment for fleet charging optimization. Researchers and fleet operators can easily adapt the framework to fit their use-cases, and assess the impact of RL-based charging on economic feasibility, battery degradation, and operations.