In the past several years, various continuous-time robust control methods have been proposed for robot manipulators. Most of them consist of linearization by state feedback and robust compensation by switching gain. In the digital implementation of these methods, the computation time required for linearization should be taken into consideration. On the other hand, robust compensation should be implemented at a short sampling period to achieve the specified tracking error bound. However, there is little consideration of these points in the previous researches. Consequently, the systems often lead to chattering and the specified tracking error bound can't be achieved.
In this paper we propose a scheme for a hierarchical robust control system for robot manipulators which is composed of an upper level loop and a lower level loop. In the upper level loop, the input for linearizing compensation, the desired trajectory and the switching gain are computed at a low sampling frequency. In the lower level loop, the switching input are generated at a high sampling frequency. This scheme will make the computation for robust compensation very fast. The performance of this hierarchical system is analyzed in consideration of the sampling period of an upper level loop and the modeling error.