Joint Rebalancing and Charging for Shared Electric Micromobility Vehicles with Human-system Interaction

The use of shared electric micromobility vehicles, such as bikes and scooters, has become increasingly popular. It leads to the problem of management (i.e., rebalancing and charging). Existing approaches typically assume that all vehicles have an equal chance of being selected for a ride, which is not practical. To overcome this limitation, we propose a reinforcement-learning-based framework incorporating human-system interaction. We first predict the likelihood of each vehicle being selected, then integrate this prediction into the reinforcement learning framework. The aim is to create a more realistic simulation process to guide policy learning more effectively. Our experimental results demonstrate the effectiveness of incorporating human-system interaction.