基于多智能体强化学习的物流配送路径调度策略

Logistics Distribution Route Scheduling Strategy Based on Multi-Agent Reinforcement Learning

为优化物流系统的配送路径,课题组提出了基于多智能体强化学习的配送路径调度(Multi-agent Reinforcement Learning based Routing Scheduling,MRLRS)策略。MRLRS策略将配送路径调度视为路径生成过程,采用了具有注意层的编码器-解码器框架来迭代生成物流的配送路径,并为模型训练设计了一种具有无监督辅助网络的多智能体强化学习方法。使用模拟实验对MRLRS策略进行评估,实验结果表明,提出的MRLRS策略的性能要优于现有的方法。

Aiming at the distribution route optimization problem in the logistics system,the paper proposes a Multi-agent Reinforcement Learning based Routing Scheduling(MRLRS)strategy.The MRLRS strategy regards delivery route scheduling as a route generation process,proposes an encoder-decoder framework with attention layers to iteratively generate logistics delivery routes,and designs a multi-agent reinforcement with unsupervised auxiliary network for model training study method.The MRLRS strategy is evaluated using simulation experiments,and the experimental results show that the proposed MRLRS strategy outperforms the existing methods.