Font Size: a A A

Research On Multi-robot Task Allocation And Path Planning Technology Based On Warehouse Environment

Posted on:2024-01-17Degree:MasterType:Thesis
Country:ChinaCandidate:J KangFull Text:PDF
GTID:2568307055987529Subject:Engineering
Abstract/Summary:PDF Full Text Request
The multi-robot system has good autonomy,sensitivity and strong robustness,can cooperate to complete the task,and overcome the difficulties faced by a single robot during operation.Therefore,the multi-robot system has been widely used in various fields of industrial production.Its main research The content includes multi-robot task allocation and path planning.This paper conducts related research on task allocation and path planning in a multi-robot system in a warehouse environment.The main research contents are:(1)Aiming at the problem of multi-robot task assignment,this paper designs a policy network suitable for multi-robot task assignment,uses Graph SAGE network to encode and update feature nodes,updates through multi-head attention layer,decoder uses multi-head selfattention mechanism to decode context,output the decision probability of the robot.In this paper,the REINFORCE algorithm with baseline is used for training.This paper conducts simulation experiments with different numbers of robots completing various tasks,and compares the results of different algorithms.The results show that the strategy network proposed in this paper has a significant advantage in solving time compared with other algorithms under the same number of robots.The first advantage not only performs well in small-scale problems,but also is more prominent in large-scale problems,especially in largescale computing instances MRTA_1000_5 and MRTA_1000_10.Compared with algorithms such as SOM,K-means,ACO,and OR-tools,the solution distance is reduced by more than10%,and the solution time is shortened by more than 50% compared with the SOM algorithm,and only 0.32% of the OR-tools algorithm.(2)Aiming at the multi-robot path planning problem,this paper proposes a constraintbased prioritization search algorithm,which uses the constraint relationship between multiple robots to reduce the search space.First,the LPA* algorithm is used to find the The path gets the corresponding constraint relationship,explores the sorting space and determines the priority order exchange through the sum of the paths of the multi-robot system to find the priority order with the shortest overall execution path,and uses this order to plan the path of each robot.By simulating different numbers of robots completing tasks in a storage environment,the results prove that the algorithm proposed in this paper is also superior to the comparison algorithm in terms of path length.In an environment where the number of robots is 15,the total distance planned by the comparison with the traditional algorithm is reduced by more than 8%.(3)Design and realize intelligent warehouse multi-robot simulation system.By importing environment information,number of robots,task nodes and other parameters,the system can realize multiple robots to process multiple tasks in parallel in the warehouse environment,and display the output in real time,which verifies the Propose the feasibility and effectiveness of the algorithm.
Keywords/Search Tags:multi-robot systems task assignment, path planning, deep reinforcement learning, search algorithms
PDF Full Text Request
Related items