Abstract:

This paper introduces a deep reinforcement learning (DRL) approach for determining energy-optimal routes for electric logistic vehicles, aiming to minimize operating costs. The Energy-Minimization Electric Vehicle Routing Problem (EM-EVRP) is formulated using an energy consumption model, considering factors like vehicle dynamics, road information, and charging losses. The problem is solved using a transformer-based DRL method, with a policy network designed following the Transformer structure. Experiments show the proposed DRL method outperforms existing learning-based methods and conventional methods in solving both EM-EVRP and Distance Minimization EVRP (DM-EVRP) , with the EM-EVRP achieving greater cost reduction than the traditional DM-EVRP.