Haruhisa Kawasaki, and Geng Li

Abstract:
This paper presents a gain tuning method based on the sampling period in discrete-time adaptive control for robots. Gain matrices of model-based adaptive control in a continuous-time system are allowed a high gain positive definite. The maximum of the gains depends on the sampling period, however, and gain tuning is very time-consuming. It is thus desirable to give a gain tuning rule in discrete-time adaptive control. The proposed gain tuning consists of two steps. The first is gain tuning at the basic sampling period by a skilful specialist by trial and error. The second step, executed if the sampling period changes, is a new gain calculation based on a new sampling period. The simulation and experiments with 1-dof and 3-dof robots demonstrate that the robot controller is stable at the large variance of sampling period changes and more accurate than a fixed gain controller.

Keywords: robot, adaptive control, discrete time system, gain tuning, sampling period, Lyapunov function, stability