Abstract
This paper proposes a robust control scheme for the robotic visual feedback control. The visual feedback control system is regarded as the motion control system based on the image information obtained from the camera mounted on the robot end-point. Two ways of visual feedback control shame, namely position-based and feature-based have been proposed. The feature-based scheme uses the object image feature directly without computing the object 3D position and orientation. In this paper the plant is represented in state space with the state being the object features, which is linearized around the desired state to obtain a linear time-invariant system. The modeling error due to the linearization is regarded as a multiplicative plant perturbation. The robustly stabilizing controller is obtained by solving a mixed sensitibity H∞ problem. Real-time experimental results on PUMA560 are also presented to show the effectiveness of the proposed method.
