An adaptive large neighborhood search method for rebalancing free-floating electric vehicle sharing systems

Abstract

Free-floating electric vehicle sharing systems allow users to pick up an available electric vehicle (EV) and return it to any permissible parking location within a service area. Such service flexibility can drive a severe spatial imbalance between vehicle availability and trip demands. Hence, it is an important part of their operations to relocate the EV fleet to meet the next day’s demand with sufficient battery levels. This relocation operation involves a complicated routing problem for a fleet of shuttles to transport the staff drivers who recharge, if necessary, and relocate the EVs to proper demand locations. Characterized by unique hierarchical and interdependent decisions, the EV relocation and shuttle routing problem poses significant computational challenges for large-scale problems. We devise an efficient algorithm that adapts the Adaptive Large Neighborhood Search (ALNS) metaheuristic framework to overcome such unique challenges. The algorithm is tested on two sets of data: randomly generated data and real-world EV-sharing usage data in Amsterdam. The results validate the efficiency and effectiveness of our ALNS algorithm. In addition, the analysis of total operational cost and waiting time percentage provides practical recommendations to decision-makers on choosing the mode of staff transportation (e.g., shuttles vs. personal mobility such as scooters). Lastly, our numerical results also highlight the usefulness of our ALNS method, which is quite flexible to be applied to a dynamic environment where some EV demands are removed or added in the course of EV relocation operations.