Research On Encirclement Control For A Group Of Targets By Multi-robot System

The multi-robot system has the characteristics of high reliability,strong extensibility and less communication cost.It is now widely used in military,manufacturing industry,warehouse logistics,science,education and entertainment.Multi-target circumnavigation control based on multi-robot system is a comprehensive multi-robot collaborative control problem.It is required that all robots establish cooperative strategies including top-level decision-making,task allocation and formation control.During the circumnavigation,it needs to ensure that there is no conflict between the robots.In this paper,the task allocation and formation control of multi-target circumnavigation are studied.Firstly,the task allocation algorithm of multi-robot multi-target circumnavigation is designed based on hybrid architecture,and then the classical control method and deep reinforcement learning method are used to realize the encirclement.The research contents and innovations of each part are as follows:Firstly,a hybrid architecture-based task allocation algorithm(Hybrid Dynamic Task Allocation,HDTA)is designed for multi-robot multi-objective task allocation.In reality,there are some differences of motion performance of each robots,and the target's ability to escape is not exactly the same.Therefore,this paper designs a dynamic task allocation(HDTA)algorithm based on hybrid architecture,which can assign multiple targets to each robot group efficiently without conflict,and it provides the possibility for multiple robots to surround targets.Then,this paper designs an improved circumnavigation formation control algorithm for the encirclement control problem.This paper uses graph theory and classical control theory to model and analyze the encirclement formation problem,and then designs a distributed formation control algorithm a totally.At the same time,in view of the real-time path planning problem of each robot in the formation control process,this paper combines two classical real-time path planning algorithms,sub-objective method and artificial potential field method,to meet real-time planning requirements and guarantee collision-free.Finally,the proposed method will insure that the robots can avoid the obstacle during the circumnavigation.Finally,based on the Multi-Agent Deep Deterministic Policy Gradient(MADDPG)reinforcement learning algorithm,the encirclement formation control of multi-robot target circumnavigation is realized.The multi-robot target circumnavigation is regarded as a multi-agent game problem.The problem is modeled and analyzed by Markov game theory.The MADDPG algorithm is then used to train the multi-robot reinforcement learning and verified by simulation experiments.The simulation results show that the proposed method can effectively accomplish the multi-robot target circumnavigation task and provide a new idea for solving the multi-robot multi-target circumnavigation problem.