Robust neural control of robot-camera visual tracking

Abstract

In this paper, we propose a new method to control a robot-camera visual tracking system to track a moving target so that the image feature of the target can match some desired one. In particular, we develop a new control algorithm to calculate the necessary joint torques. To deal with the dynamics and Jacobian uncertainty of the problem, an on-line learning neural network (NN) is used to approximate uncertain components and tune the control scheme to ensure the mismatch of the image feature vanishing to 0. We also prove the asymptotical stability of the proposed tracking method by using Lyapunov stability method.