PUB - Publikationen an der Universität Bielefeld

Mobile service robots are recently gaining increased attention from industry as they are envisaged as a future market. Such robots need natural interaction capabilities to allow unexperienced users to make use of these robots in home and office environments. In order to enable the interaction between humans and a robot, the detection and tracking of persons in the vicinity of the robot is necessary. In this paper we present a person tracking system for a mobile robot which enables the robot to track several people simultaneously. Our approach is based on a variety of input cues that are fused using a multi-modal anchoring framework. The sensors providing input data are two microphones for sound source localization, a pan-tilt camera for face and torso recognition, and a laser range finder for leg detection. Through processing camera images to extract the torso position, our robot can follow a person guiding the robot that is not oriented towards the robot and that is not speaking. In this case the torso information is especially important for robust tracking during temporary failure of the leg detection.