基于强化学习的带软时间窗多行程绿色车辆路径优化研究
Reinforcement Learning Based Path Optimization for Multi-Trip Green Vehicles with Soft Time Window
姚利军 1王可君1
作者信息
- 1. 湖南省长株潭烟草物流有限责任公司,湖南长沙 410004
- 折叠
摘要
为了助力物流行业响应"碳达峰"和"碳中和"建设目标,提速绿色物流产业的建立与发展,首先综合考虑油耗、碳排放、人力、车辆、用户体验等因素,构建带软时间窗约束的多行程绿色车辆路径优化模型.然后综合考虑PinSAGE图网络、TRPO和GAE方法来改进Actor-Critic的深度强化学习优化算法,最后采用Actor-Critic算法对模型对绿色多行程车辆路径方案求解.实验表明,提出的求解方法能高效规划绿色车辆路径,进而显著减少物流成本,实现物流企业经济效益与环境效益的双重优化.
Abstract
In order to assist the logistics industry in achieving its goal of peak carbon dioxide emissions and carbon neu-trality,construction must be facilitated,and a green logistics industry must be rapidly established and developed.Firstly,a multi-trip green vehicle path optimization model with soft time window constraints is constructed by comprehensively consid-ering the factors of fuel consumption,carbon emission,manpower,vehicles,and user experience.Subsequently,the Pin-SAGE graph network,TRPO,and GAE methods are considered collectively to enhance the deep reinforcement learning opti-mization algorithm of Actor-Critic.Ultimately,the Actor-Critic algorithm is employed to address the model for the green multi-trip vehicle path scheme.Experimental evidence indicates that the solution method proposed in the paper is an effective means of planning green vehicle routes,which in turn has the potential to significantly reduce logistics costs and realise the dual optimisation of economic and environmental benefits for logistics enterprises.
关键词
绿色物流/软时间窗/深度强化学习/Actor-Critic框架Key words
green logistics/soft time window/deep reinforcement learning/Actor-Critic framework引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据