Tracking and pursuing persons with a mobile robot

In this paper we address the problem of detecting, tracking, and pursuing persons with a mobile robot in real-time. Our system consists of two cooperatively interacting parts: A visual gaze holding process using an active stereo vision system and a component controlling the trajectory of the robot. The detection process is based on an approach which combines color, motion, and contour information. Additionally, facial features and depth information are integrated to increase the robustness of detection. For navigation a potential field method is used taking into account the path of the person going ahead as well as a local environment map built by means of laser range sensor data. Consequently, stationary and dynamic obstacles can be avoided during the process of pursuit. The system is able to operate at approximately human walking rates