Low-Carbon Scheduling of Urban Logistics Electric Vehicles Using Deep Reinforcement Learning with Carbon Quota Constraints

Abstract

With the increasing emphasis on carbon reduction and green logistics, optimizing the scheduling of urban electric logistics vehicles under carbon quota constraints has become a critical challenge. This study proposes an integrated optimization framework that combines an Improved Grey Wolf Optimizer (IGWO) with an Actor-Critic deep reinforcement learning algorithm to address the electric vehicle routing problem (EVRP). The model incorporates transportation cost, energy consumption, and carbon emission penalties, while ensuring compliance with vehicle capacity, charging, and carbon quota limitations. Simulation results demonstrate that the proposed IGWO-Actor-Critic method outperforms traditional heuristic and metaheuristic approaches in both efficiency and solution quality, particularly in large-scale logistics scenarios. This research not only contributes to the advancement of low-carbon logistics strategies, but also directly supports the Sustainable Development Goals (SDGs), especially the goal of Sustainable cities and communities, by providing a scalable and intelligent solution for urban logistics decarbonization.