Multi-robot Search Algorithm Based on Experiential Learning

Multi-robot search is a hot issue in the field of multi-robot research and has great application potential in the fields such as disaster search and rescue, geological exploration, indoor search, battlefield exploration, etc. This paper considers a cooperative multi-robot search problem in which the target is movable in a limited space with various obstacles, and designs a search method based on probability and voronoi diagram partition to repeatedly search for the movable target until it is found. In the method, the search space is divided into different regions and the probability for each region is updated with the search times of each robot, so that the robots can obtain experience from the search. According to the region probability and voronoi diagram, a subtask region segmentation algorithm and a region repeat traversal algorithm with obstacle avoidance strategy are designed for the robots to search in the whole space as quickly as possible. The experiments indicate that the performance of the proposed cooperative search algorithm is much more significant than that of other compared approachs and is gradually improved when the experience becomes more.