Variable Neighborhood Search Based on Reinforcement Learning for Green Vehicle Routing Problem

The green vehicle routing problem (GVRP) is a trendy variant of the well-known vehicle routing problem that incorporates environmental considerations such as minimized fuel consumption or emissions. This study introduces a new hybrid approach combining variable neighborhood search (VNS) with the reinforcement learning (RL) paradigm to effectively resolve the GVRP. VNS is a metaheuristic optimization technique that explores numerous neighborhoods of a solution to improve it, whereas RL is an agent-based machine learning algorithm. Thus, integrated with VNS, the RL may find competitive solutions for the GVRP. Our numerical results and analysis prove the effectiveness of the proposed hybrid methodology for resolving the GVRP. This new hybrid strategy benefits from the combination of VNS as an evolutionary optimizer and RL as a machine learning methodology to effectively resolve the GVRP.