Adaptive control by predictive identification and optimization

A model-reference adaptive control system is described where extrapolation techniques are used for identification and for error-prediction at discrete time intervals. The system employs rectangular adaptation pulses of finite duration to minimize a cost-functional of predicted square errors. Weighted squares of the error rate-of-change are included in the cost-functional to be minimized and a number of constraints are considered. Simulation resuits for systems consisting of linear, time-varying, and nonlinear second- to fifth-order processes with linear second-order reference-models are given, where satisfactory adaptation is accomplished.