The transition toward electric mobility presents unprecedented challenges in vehicle routing and energy management, particularly under real-time operational constraints. This paper presents a comprehensive comparative analysis of optimization methodologies for autonomous electric-vehicle routing and charging, addressing the critical need for integrated decision-making frameworks that balance travel-time minimization with battery preservation. We formulate the problem as an optimization model incorporating traffic conditions, charging station availability, battery state-of-charge evolution, and minimum charge constraints to mitigate range anxiety. To solve this computationally challenging problem, we develop and evaluate multiple solution approaches: heuristic algorithms including Time Efficient Routing and Charging (TERC), TERC2, and Dynamic TERC2; classical reinforcement learning through Q-learning and transfer Q-learning; deep reinforcement learning via deep Q-networks (DQL); and state-of-the-art metaheuristics including Simulated Annealing (SA) and Memetic Algorithm (MA). Each approach is implemented and tested on networks of varying scale with systematically distributed charging infrastructure under dynamic traffic conditions. Experimental results demonstrate that deep reinforcement learning consistently outperforms alternative methods, achieving 15-25% reductions in travel time compared to tabular Q-learning and 7.0-8.1% reductions compared to MA and 13.1-16.1% compared to SA on larger networks, while maintaining superior battery charge trajectories and faster convergence. Heuristic methods exhibit competitive performance in smaller networks but show limitations as problem scale increases. Transfer Q-learning, initialized with heuristic solutions, achieves accelerated convergence compared to standard Q-learning, though it does not surpass deep Q-networks in final performance. The systematic comparison reveals that neural function approximation provides significant advantages in capturing the complex state space dynamics inherent in EV routing problems. These findings contribute to the advancement of intelligent transportation systems and provide practical insights for implementing real-time EV routing solutions in urban environments.
Ebrahimi et al. (Mon,) studied this question.