Research On Multi-Robot Cooperative Pursuit Problem

The multi-robot cooperative pursuit problem is a problem in which a team of autonomous mobile robots act as pursuers to pursue multiple moving targets cooperatively. In the course of pursuit-evasion, the pursuer team must act cooperatively and collaboratively to complete the pursuit task. The multi-robot cooperative pursuit problem is a typical problem for the research of distributed multi-agent robotics cooperation and coordination. The key technologies of the problem are used widely in military and industry, such as search and rescue in battlefield, capture of prisoner of war, enclose/capture the invader by military robot troop, aerial warfare, etc.This paper divides the multi-robot pursuit problem into three problems: searching target, allocating pursuit task and organizing pursuer team, pursuing target to researches. The main content of this paper are followed.Firstly, the background knowledge and the domestic and foreign research status of the pursuit problem are summarized, and the objective and significance of the multi-robot pursuit problem is introduced.Secondly, searching target and building map in the condition that environment and the position of the target are unknown are studied. As for the pursuit in unknown environment, the usual approach is to transform the problem into a problem in known environment by building a map of the terrain firstly. But in this paper, the pursuers search target and build map simultaneously. This dissertation firstly discusses map building based on probabilistic framework. Then, with the use of probabilistic map, several heuristic target-searching strategies are proposed.Thirdly, how to organize the pursuer team when there are multiple targets is discussed. We treat the multiple targets pursuit problem as multiple tasks allocation problem. The traditional Contract Net Protocol is improved by using Case-based reasoning to reduce the scope of inviting bidding and dynamic alliance. Then a multiple tasks allocation algorithm based on auction is proposed. Fourthly, multi-robot pursuit problem in continuous environment is investigated from the view of behavior-based robotics methodology. Five basic behaviors used for the problem are designed.Fifthly, the pursuit simulation platform is developed, and several experiments are carried out on the platform in order to verify the feasibility and validity of the proposed methods and models.