A Decoupled Motion Estimation for Visual Regulation of Nonholonomic Mobile Robot

Abstract

A visual regulation strategy based on a decoupled ego-motion estimation technique, is presented for a nonholonomic mobile robot. Ego-motion in a static environment can be robustly estimated by planar region alignment, which initially detects the 2D planar motion between two frames, and the 2D motion is used to align corresponding image regions. Such a 2D registration removes all effects of the camera rotation, and the resulting residual displacement between the two aligned images is an epipolar field centered at the FOE (Focus of Expansion). Then 3D camera translation is recovered from the epipolar field. The 3D camera rotation is then derived from the recovered 3D translation and the detected 2D motion. By this way, the ego-motion estimation is decoupled into a 2D parametric motion and residual epipolar parallax displacements, which avoids many of the inherent ambiguities and instabilities associated with decomposing the image motion into its rotational and translational components, and hence makes the computation of ego-motion or 3D structure estimation more robust. Based on the ego-motion estimation, an adaptive control law for visual regulation of nonholonomic mobile robot is presented and the stability of the close loop system is analyzed in the sense of Lyapunov stability theory. Experiments show that the convergence of the proposed visual regulation.