Action evaluation for mobile robot global localization in cooperative environments

This work is about solving the global localization issue for mobile robots, operating in large and cooperative environments. It tackles the problem of estimating the pose of a robot, or team of robots in a map reference frame, given the map, the real-time data from the robot onboard sensors, and the real-time data coming from other robots or sensors in the environment. After a first step of position hypotheses generation, an efficient probabilistic active strategy selects an action, for a single lost robot case, or two joint actions when two lost robots are in a line of sight, so that the hypotheses set is best disambiguated. The action set is adapted to the multi-hypothesis situation, and action evaluation takes into account remote observations available in robot network systems. This paper presents the theoretical formulation for both non-cooperative, and cooperative cases. An implementation of the proposed strategy is discussed, and simulation results presented.