A discrete-time direct adaptive multiple-model control scheme

A new multiple-model adaptive control scheme for discrete-time systems is developed to improve the adaptive system transient response, by using a direct model reference adaptive control design. Such a scheme employs a bank of direct adaptive controllers whose parameters are initialized from different parameter subregions and directly updated on-line. A control switching mechanism is set up, by finding the minimal performance cost index which is based on estimation errors generated from the tracking error and the individual controller parameter estimates, to select the most appropriate controller to generate the control signal. Such a multiple-model adaptive scheme for systems with unknown sign of the system high frequency gain is also designed, by constructing two controllers whose adaptation is from modified robust adaptive laws with positive and negative adaption gains respectively. Simulation results are presented to show the improved system transient responses, as compared with that of a single-model adaptive control system, and the desired system performance in the system with unknown sign of the system high frequency gain.