Discrete-time model reference learning control: Theory and experiment

Many modern industrial installations include digital computers as an integral part of the operations. Digital computers are extensively used to implement complex control algorithms to control the functioning of the system. The discretization of the nonlinear dynamic equations like robot dynamics results in an extremely complicated discrete dynamic equations. Therefore, it will be difficult to design a discrete-time controller to give good tracking performances in the presence of certain uncertainties. In this paper, a discrete-time Model Reference Learning Control (MRLC) algorithm is presented for a class of nonlinear and lime varying discrete-time system. Sufficient conditions for guaranteeing the convergence of the discrete-time MRLC system are derived. The robustness of the learning system to measurement noise, dynamics fluctuation and re-initialization error is studied. Experimental results of an industrial robot SEIKO TT3000 are presented to verify the theoretical analysis.