Adaptive robust control of robot manipulators subject to input-dependent uncertainties

In most approaches to adaptive robust control design for manipulators, a constant upper bound is considered for unmodeled dynamics. However, experiments show that the nature of the control input affects the magnitude of the unmodeled dynamics. For example, in an almost rigid manipulator where the mechanical flexibilities are considered as unmodeled dynamics, high frequency components of the input torque which are resulted form non-smooth control laws, can themselves excite system flexibilities. In this paper, we consider different characterizations of the unmodeled dynamics to take into account the effect of control input signal on the unmodeled dynamics. Simulation results illustrate the performance of the proposed controllers.