Research On Multi-robot Task Planning Method

In recent years,with the development of AI technology,Robot technology has attracted wide attention from scholars at home and abroad.At present,great progress has been made in the research of single robot,which can accomplish such tasks as speech recognition,image recognition,natural language processing,automatic control and so on.But in many special situations,such as outer space exploration,seabed exploration,military anti-terrorism,industrial mining,post-earthquake search and rescue,it is difficult to accomplish the task efficiently only by a single robot.Because multirobot system has obvious advantages over single-robot system in computing ability,resources and spatial distribution,it can accomplish tasks through coordination and cooperation that single robot can not accomplish.Therefore,the research of multi-robot cooperation is of great significance.This thesis mainly researches the multi-robot task planning methods from three aspects: multi-robot task allocation,multi-objective multi-robot task allocation and multi-robot path planning.Firstly,a multi-robot task allocation model is established,and a novel swarm intelligence optimization algorithm,firework algorithm,is introduced.The firework algorithm is applied to solve the task allocation problem of multirobot.And an improved firework algorithm is proposed by optimizing the formula of explosive spark's number,explosive amplitude and selection strategy.The improved firework algorithm is compared with the original firework algorithm and several commonly used heuristic algorithms to verify its performance.Then,a multi-objective multi-robot task allocation model is established,and several typical multi-objective optimization algorithms,such as NSGA-II,SPEA2 and PESA,are introduced.Because they are all multi-objective optimization algorithms based on genetic algorithm,this thesis proposes a multi-objective optimization algorithm based on firework algorithm,and it is applied to solve multi-objective multi-robot task allocation problem.And the S-metric is used as the assessment criteria of Pareto solution set.The four algorithms are compared to verify the advantages of the proposed algorithm.Finally,the multirobot path planning problem is considered as multi-traveling salesman problem,and a multi-robot path planning model is established.This thesis introduces the detailed steps of ant colony algorithm to solve traveling salesman problem,and proposes a new ant colony algorithm which does not need to add virtual nodes and transforms multi-traveling salesman problem into traveling salesman problem.And it is used to solve the multi-robot path planning problem.Experiments show that the ant colony algorithm has better optimization performance than particle optimization algorithm and genetic algorithm.The performance of iterative algorithms such as swarm intelligence optimization algorithms is affected by various factors,such as the setting of algorithm parameters and termination conditions.In the part of experimental,the advantages and disadvantages of various termination conditions are analyzed in detail.Finally,the runtime and evaluations are selected as the termination conditions for comparing algorithms effectively.