Research On Location And Path Planning Of Handling Robot Based On Deep Reinforcement Learning

In recent years,handling robots have been widely used in manufacturing and service industries.Handling robots gradually replace manual tasks.Positioning and path planning are the prerequisites for carrying out tasks of handling robots,and are also hot research issues in the field of handling robots at present.At the same time,there are some problems in the positioning and path planning of the handling robot,such as positioning and map deviation caused by sensor error and map distortion,low efficiency of global path planning algorithm caused by data redundancy,and dynamic obstacles affecting the optimal path selection of the robot.In view of the problems in robot positioning and path planning,this paper conducts the following research :Aiming at the problem of positioning and map matching deviation caused by sensor error and map distortion,the SLAM algorithm based on particle filter is optimized by hierarchical sampling method,and the robot track is calculated and the positioning accuracy is improved.Then,the improved Gmapping algorithm is used to construct map matching,remove the distortion during map construction,improve the accuracy of map matching and reduce the error.The experimental results show that the positioning error of the handling robot is reduced by 42.86 %,and the trajectory error is reduced by 46.53 %.Aiming at the problems of large amount of data and low efficiency of algorithm in path planning of handling robots,this paper integrates BN algorithm into Q-learning algorithm based on deep reinforcement learning,normalizes the data,reduces the amount of algorithm data and improves the efficiency of algorithm.The experimental results show that the convergence speed of the algorithm is improved by 25 %,the learning reward value is increased by 34 %,and the robot can find an optimal path.Aiming at the problem of unreachable target and local minimum points caused by dynamic obstacles in local path planning of handling robot in dynamic environment,the method of increasing obstacle information in potential field function and providing external force for robot is adopted to improve the ability of robot to avoid obstacles and get rid of local minimum points.