Active Localization for Mobile Service Robots in Symmetrical and Open Environments

Abstract

As mobile service robots are increasingly being introduced into everyday human activities, it is essential to get service robots moving securely. Improve robot localization will lead to more secure and efficient robots. Passive localization systems for indoor environments are a mature area and have a good performance in environments with enough features; however, robots still could get lost in symmetrical or open areas. On the other hand, active localization systems improve localization by letting the robot choose where to move or point its sensors to improve the robot pose. In this paper, we address the problem of active localization in the context of the indoor position tracking problem in very symmetrical and open environments. Our active localization system’s main contributions are: (i) introduce two cluster algorithms to group pose hypotheses, (ii) use a reduced map for searching informative points, and (iii) perform only rotational actions to point the sensor to the selected point without drift the robot from the navigation path. The system demonstrated a significant decrease in ground truth error and localization uncertainty in both simulated and real robot experiments.