Multiobjective vehicle routing optimization via deep reinforcement learning for coordinated energy efficiency and emission control in urban road networks

Urban transportation is increasingly challenged by energy inefficiency and excessive emissions, hindering progress toward the Sustainable Development Goals (SDGs). While the Vehicle Routing Problem (VRP) has been extensively studied, most approaches remain limited to single-objective optimization and struggle to balance efficiency with environmental performance. This study proposes a multi-objective optimization model for energy–emission collaborative control in urban road networks, combined with an Actor-Critic deep reinforcement learning framework enhanced by an Improved Sparrow Search Algorithm (ISSA). The method integrates traffic signal coordination, energy budgets, and emission constraints, enabling adaptive hyperparameter tuning and stable convergence. Simulation results show superior performance compared with conventional intelligent optimization algorithms, effectively reducing energy use and emissions under varying vehicle scales. The findings not only advance VRP optimization theory but also provide a practical pathway for achieving sustainable cities and communities within the broader SDGs framework.