摘要
“货到人”拣选系统背景下,对AGV的路径进行优化是提高拣选效率、降低运行成本的关键。首先,针对仓储环境引入路径表概念,存储相似路径,减少重复搜索同一路径的工作;其次,针对传统Q学习收敛速度慢及容易陷入局部最优解问题,借助模拟退火算法中的动态探索因子思想以一定概率跳出局部最优解。最后,采用栅格法建立仿真环境,将改进Q学习算法与A*算法、传统Q学习算法对比进行仿真实验,实验结果验证了本方法的有效性。
AVG path optimization is the key to improving efficiency and reducing operating cost of a cargo-to-man picking system.In this paper,we introduced the concept of path table into the storage environment to store similar paths,so as to eliminate repeated searches for the same paths.Next,target at the slow convergence and propensity toward local optimization of the traditional Q learning process,we leveraged the dynamic exploration factor in the simulated annealing algorithm to break out of local optimization at certain probability.At the end,we used the grid method to establish the simulation environment,and compared the improved Q learning algorithm with the traditional Q learning algorithm by way of a simulation experiment,the result of which verified the effectiveness of the improved learning algorithm.
作者
杜卓颖
李金禧
张祥来
朱琳
Du Zhuoying;Li Jinxi;Zhang Xianglai;Zhu Lin(School of Management,Harbin University of Commerce,Harbin 150000,China)
出处
《物流技术》
2020年第12期88-92,共5页
Logistics Technology
基金
2019年国家级大学创新创业训练项目“基于强化学习的AGV路网规划研究”(201910240005)。
