Robust Visual Tracking Control of a Nonholonomic Mobile Robot with Depth-based Camshift

Abstract

This paper mainly addresses the problem of tracking a moving target for a nonholonomic mobile robot equipped with an on-board camera. A visual tracking control system consisting of a robust visual tracking algorithm and a robust visual tracking controller is developed. The proposed visual tracking algorithm, namely the depth-based Camshift, possesses robustness against partial occlusion, image noise, deformation and similar-colored objects. Supplied with direct measurements from visual tacking, a robust controller is developed based on the idea of feedback linearization and sliding mode techniques to decouple relative motion equation in kinematic level and overcome unknown motion of the target. Simulation studies and experiments are conducted then to verify the performance of proposed controller and the whole mobile robot tracking scheme.