Experiments in Real-Time Vision-Based Point Stabilization of a Nonholonomic Mobile Manipulator

The stabilization to a desired pose of a nonholonomic mobile robot carrying a manipulator arm, based on sensory data provided by a camera mounted on the end-effector of the arm, is considered. Instances of this problem occur in practice during docking or parallel parking maneuvers of such vehicles. The visual data obtained as the camera tracks a target of known geometry are used to implement a continuous time-varying state feedback for the stabilization of the mobile robot. We focus on the case where the camera moves parallel to the plane supporting the mobile robot. The experimental evaluation of the proposed techniques uses a mobile manipulator prototype developed in our laboratory and dedicated multiprocessor real-time image processing and control systems. The real-time programming aspects of the experiments are handled in the context of the Orccad control architecture development environment.