The Use of Negative Detection in Cooperative Localization in a Team of Four-legged Robots

In multirobot localization, the pose beliefs of two robots are updated whenever one robot detects another one and measures their relative distance. This paper proposes the use of a recently proposed localization algorithm that uses negative detection information to localize a group of four-legged Aibo robots in the domain of the RoboCup Standard Platform Four-Legged Competition. Our aim is to test the performance of this algorithm in a real world environment, where robots must use vision capabilities to detect the presence of other robots. Experiments were made with two robots and with three robots, using a cascade of detections to update the position of more than one of the robots. Experimental results show that negative detection localization allows us to localize the robots in situations where other algorithms would fail, leading to an improvement of the localization of real robots, working in a real world domain. Keywords— Cooperative Robot Localization, Markov Localization, SIFT Algorithm.