Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (17): 318-327.DOI: 10.3778/j.issn.1002-8331.2204-0491

• Engineering and Applications • Previous Articles     Next Articles

AGV Path Planning Based on Memristor Reinforcement Learning in Warehouse Environment

YANG Hailan, QI Yongqiang, RONG Dan   

  1. School of Mathematics, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China
  • Online:2023-09-01 Published:2023-09-01

仓储环境下基于忆阻强化学习的AGV路径规划

杨海兰,祁永强,荣丹   

  1. 中国矿业大学 数学学院,江苏 徐州 221116

Abstract: In order to solve the AGV path planning problem in the dynamic storage environment, the grid method is used to model the warehouse environment. And the path planning tasks in the static environment is completed by improving the probability transfer function and the ant colony algorithm pheromone. Since the memristor characteristics is similar to biological synapse, which is used as the neural network synapse structure. And the DQN algorithm based on memristor array is used for dynamic local obstacle avoidance. The path planning method is switched in real time according to whether there are dynamic obstacles within the sensing range of the AGV to achieve efficient AGV handling. Simulation experiments are carried out on the MATLAB, and the results show that the path planning method can effectively and real-time plan a safe and collision-free optimal path for the AGV.

Key words: automated guided vehicle(AGV), dynamic environment, deep Q-network(DQN), memristor, path planning

摘要: 针对动态仓储环境下的AGV路径规划,采用栅格法对仓储环境进行建模,通过改进了概率转移函数及信息素的蚁群算法完成静态环境下的路径规划;利用忆阻器和生物神经突触类似的特性,将其作为神经网络突触结构,改进传统的DQN算法,并利用基于忆阻器阵列的DQN算法进行动态局部避障;依据AGV感知范围内是否存在动态障碍物实时地切换路径规划机制,以实现高效的AGV搬运工作。在MATLAB仿真平台进行实验,结果表明该路径规划方法可有效、实时地为AGV规划出一条安全无碰撞的最优路径。

关键词: 自动引导车(AGV), 动态环境, 深度Q网络(DQN), 忆阻器, 路径规划