Rebalancing Autonomous Electric Vehicles for Mobility-on-Demand by Data-Driven Model Predictive Control

The main goal of this paper is to present an end-to-end, data-driven framework for the control of Autonomous Electric Vehicles (AEV) for Mobility-on-Demand (MoD). We present a data-driven Model Predictive Control (MPC) algorithm that rebalances (i.e. preemptively repositions) the AEV fleet in order to meet the mobility demand in the near future. The algorithm consists of Mixed Integer Linear Programming (MILP) that leverages the short-term forecast of the mobility demand as well as the charging station availability in order to optimally rebalance the AEV fleet. The proposed algorithm is evaluated by means of simulations with the New York City (NYC) taxi data. The proposed algorithm outperforms other state-of-the-art rebalancing strategies by reducing the mean customer wait time by 82.3% and the number of rejected requests by 94.6% for a given fleet size and a number of charging stations.