VISUAL BEHAVIORS FOR DOCKING

This paper describes visual-based behaviors for docking operations in mobile robotics. Two different situations are presented: in the ego-do cking, each robot is equipped with a camera, and the motion is control led when docking to a surface, whereas in the eco-docking, the camera and all the necessary computational resources are placed in a single e xternal docking station, which may serve several robots. In both situa tions, the goal consists in controlling both the orientation, aligning the camera optical axis with the surface normal, and the approaching speed (slowing down during the maneuver). These goals are accomplished without any effort to perform 3D reconstruction of the environment or any need to calibrate the setup, in contrast with traditional approac hes. Instead, we use image measurements directly to close the control loop of the mobile robot. In the approach we propose, the robot motion is directly driven by the first-order time-space image derivatives, w hich can be estimated robustly and fast. The docking system is operati ng in real time and the performance is robust both in the ego-docking and eco-docking paradigms. Experiments are described. (C) 1997 Academi c Press.