Global Iterative Learning Control Of Feedback Linearizable Systems

Abstract

Globally feedback linearizable systems with matching uncertain nonlinearities are considered: no parametrization is required for the uncertainties. Periodic reference signals with known period T are to be tracked. Provided that known bounding functions on the uncertainties are available, a state feedback iterative learning control is designed which achieves asymptotic tracking and guarantees bounded closed loop signals from any intial condition. The unknown open loop periodic reference input is asymptotically reconstructed by the controller, so that a dynamic inversion of the uncertain nonlinear system is achieved. An extension to partial state feedback is also proposed. The novel control technique is illustrated for a single-link robot arm