Robust Object Tracking for Robot Manipulation and Navigation

Applications for visual control of a robot manipulator often require the pose (position and orientation) of an object as input for the control loop of the robot. On the other hand robot navigation usually needs the pose of the robot with respect to a world co-ordinate system. In both cases the same task of determining the pose of an object with respect to the camera has to be carried out. This object can be a tool to grasp or the environment to navigate in. We present a system that is able to robustly determine object pose in real-time from features in the image such as points, straight edges and ellipses using a 3D model of the object. As an example of a robot navigation task we present a walking robot which is to be used for inspection and welding of sections of large ships at a shipyard .