Control of a rigid robot using an uncalibrated stereo vision system

Abstract

A supervisor of the estimator-based type together with a certainty equivalence control is proposed for controlling the position of a rigid robot in a prescribed workspace using data observed by two uncalibrated video cameras. By exploiting a well known algebraic relationship from the computer vision field (the epipolar constraint), the supervisor is able to falsify or eliminate over time, unpromising candidate camera models.