Research On Multiple AUV Path Planning And Formation Control

In the field of marine resources development and marine military development,underwater robots have a very broad application prospect and have been a research hotspot in recent years.Research on underwater robots in various countries is also becoming more and more in-depth.Underwater robots combine a variety of intelligent control technology theories on traditional carriers and can accomplish complex tasks underwater.In underwater robot technology,path planning and formation control is one of the most important technologies and has always been the focus of research.Therefore,this paper analyzes the current mainstream path planning and formation control algorithms at home and abroad.Based on its existing problems,the following studies have been conducted:In order to solve the problem that the reinforcement learning can not calculate the continuous action and the convergence is poor in the large-scale continuous environment,this paper proposes an AFsarsa(?)algorithm,introduces fuzzy control theory into reinforcement learning,and designs two-level fuzzy sarsa(?)learning.Firstly,use the MRMI model to build the fuzzy reasoning rules base on the state value function,generalize the large-scale continuous state,and calculate the continuous action.Secondly,use the FMP model to build the fuzzy reasoning rules base on the reward value function.By adding additional bonus values,the convergence of the algorithm is improved and the path planning task is completed more efficiently.For the current multi-AUV formation using the traditional leader-virtual structure method to control the formation movement,there are some problems: centralized control problems,difficult problems of rigid structure obstacle avoidance.This paper proposes an improved leader-virtual structure method.The formation shrinkage based on three-dimensional space was proposed,and the formation team made efficient contraction and obstacle avoidance according to the formation contraction degree,solved the problem of poor obstacle avoidance of rigid structure formations.At the same time,a dynamic switching pilot strategy based on heuristic evaluation function is proposed to solve the centralized control problem and to avoid obstacles while keeping the formation system stable.Finally,we established a communication interface between the control platform and the simulation software,conducted a visual simulation experiment,performed data processing and algorithm analysis and comparison,verified the superiority of the proposed AFsarsa(?)algorithm and the improved leader-virtual structure method.