Robust IMC based impedance control of robot manipulators

Abstract

This thesis deals with robust impedance control of robot manipulators based on IMC (internal model control) and its application to the human arm impedance estimation. The goal of impedance control is to deal with all parts of a robot task - free motion, constrained motion and the transition between them - by realizing desired impedance dynamics. However, accurate realization of desired impedance dynamics is not easy because of coupled nonlinear robot dynamics. Moreover, recently, it has been reported that, in impedance control, there exists a dilemma between impedance accuracy and robustness against modeling error, which includes unknown payload and the uncompensated coupled robot dynamics. Specifically, the first impedance control method, suggested by N. Hogan, is sensitive to modeling error, because the form of the controller is similar to that of computed torque method. For the improvement of robustness of Hogan’s method, the position based impedance control, one of the most common implementation of impedance control, is suggested. However, the position based impedance control inaccurately realizes desired impedance due to the inner position control loop dynamics. Many research works have been carried out with the aim of obtaining robustness of impedance control. Notable approaches include variable structure control, adaptive control, and iterative learning control. Although these approaches succeeded in enhancing robustness, they either use a robot dynamics model on real-time basis - the parameters of which is hard to obtain and evaluation of which is computationally demanding- or involves complex algorithms, which are not simple to put into practice. Therefore, in addition to the robustness and accuracy, simplicity of the control law is preferable. With these considerations in mind, to improve robustness and accuracy of impedance control, a robust IMC based impedance control technique, which is named IMBIC (internal model based impedance control), ...