Experimental studies of position control of humanoid robot arms using a simplified time-delayed controller: A practical approach

Abstract

A time-delayed control (TDC) method is one of robust control methods that reject disturbance for nonlinear multi-input multi-output systems. The TDC uses the previous information of control input and inertia in association with acceleration to cancel out uncertainties of the system. Performance of the TDC depends upon how the acceleration term and inertia model are accurately estimated. However, the estimation of acceleration terms is quite noisy since the second derivative of the time difference between two sequential encoder measurements is required. Thus, the estimation of acceleration terms in the TDC without sensors is a burden for real-time control applications. Since the TDC without an acceleration term may have minimal effects on cases with constant low velocities, our proposal is to ignore the estimation of the acceleration term. Ignorance may be compensated by increasing feedback gains. Experimental studies of joint control of humanoid robot arms are conducted to demonstrate and compare performances of the TDC method and the TDC without an acceleration term.