Robust controller design with reduced conservatism for calibration-free visual feedback systems and its experimental verification

This paper proposes a design method of robust controllers for systems containing uncertainties in the measurement equation. For our specific problem of visual feedback control, the effects of image distortion are reflected as uncertainties represented in an affine manner. In our previous design method, it is implicitly assumed that the parameter, which is used for deriving the state-space matrices of the controller to be designed, is chosen as the nominal value of the uncertainty parameter. This assumption causes some conservatism. We thus propose a new design method by replacing the nominal value of the uncertainty parameter in our previous method as a newly introduced fictitious parameter whose value should be a priori set but can be chosen freely. A controller for a visual feedback system which allows measurement errors due to large lens distortion is designed by using the proposed method. The validity of our controller is confirmed by a lab-level experimental setup.