Ride-Hailing Service aware Electric Taxi Fleet Management using Reinforcement Learning

Recently, the adoption of electric taxis (e-taxis) has become an essential option in many countries for the reduction of carbon emissions from large cities. However, it is generally not easy to design a sophisticated e-taxi management system due to the complex mixture of charging overheads, ride-hailing service quality for passengers, and uncertain traffic conditions. This paper proposes an Intelligent E-taxi ride-Hailing Service (I-EHS) controller that maximizes the satisfaction of served passengers while guaranteeing reliable charging for each e-taxi. Our controller integrates the reinforcement learning (RL) based e-taxi dispatcher and the heuristic-based e-taxi allocator, so as to derive the delicate e-taxi control with acceptable training overhead. Through the experiments based on the OpenAI-Gym framework, we show that our I-EHS controller efficiently finds the solution without prior knowledge of the traffic environment.