An Energy-Aware Optimal Electric Taxi Dispatching Solution Using Deep Reinforcement Learning

Electric vehicles (EVs) are increasingly adopted worldwide due to their energy efficiency and near-zero emissions, offering a greener alternative to traditional transportation. While recent advancements have significantly extended EV driving ranges and improved charging infrastructure, operational challenges such as charging logistics and energy management for electric taxis (e-Taxis) still require careful consideration to ensure reliable service. Addressing this issue is critical for ensuring uninterrupted service and customer satisfaction in e-taxi operations. In this work, an energy-aware optimal e-taxi dispatching solution is proposed, utilizing Deep Q-Network (DQN) reinforcement learning to minimize energy consumption while optimizing fleet operations. The proposed dispatching algorithm aims to reduce the frequency of e-taxi recharging, decrease customer waiting times, and lower empty taxi mileage, ultimately enhancing the efficiency and sustainability of e-taxi services. The proposed algorithm demonstrated a significant reduction in the average Bellman error with an increase in the average cumulative reward across the training episodes. This confirms its effectiveness in minimizing the energy consumption of the e-taxi fleet while maximizing the covered demand.