A robot arm is neurally controlled using monocular feedback
Patrick van der Smagt · 1996
We generalise previous time-to-contact based robot arm guidance methods to generate 3D motion trajectories from optic flow, and use this to construct a model free self-learning robot arm controller (using the visual position, velocity, acceleration, etc. of a single white object to be grasped against a black background). Of importance is the fact that no model of the robot arm or the observed world is necessary to obtain depth information with monocular vision. (3 pages)