A New Adaptive Visual Tracking Scheme for Robotic System Without Image-Space Velocity Information
Yu Zhang, Changchun Hua
Abstract
In this article, we investigate the adaptive visual tracking problem for uncertain robotic system with time-varying depth, in the case of no image-space velocity information. Existing works dealing with this problem either assume the estimated time-varying depth is invertible or require the priori information of the system kinematics. This case potentially leads to the singularity of the control scheme and requirement for a high gain feedback. In this brief, a new image-space observer which does not involve the inversion of the estimated depth is proposed, upon which, we design an adaptive visual tracking controller without using the image-space velocity information. More importantly, the observer employs a passive image Jacobian matrix-based constant feedback instead of high gain feedback and the priori information is no longer demanded. Additionally, the proposed control scheme enjoys a desirable separation property for the robotic dynamics and kinematics which makes it applicable to most industrial/commercial robots having a closed torque control loop and only permitting the joint velocity command. By using the Lyapunov stability analysis, the asymptotical convergence of both image-space tracking errors and observation errors are obtained. Numerical simulations and implementation issues concerning the application to industrial/commercial robots are presented to verify the efficacy of the control scheme.