Abstract
Extensions of the existing single-step active adaptive control algorithms are considered in this paper. Firstly, the active adaptive control of Goodwin and Payne and the innovations dual control of Milito et al. are extended to ARX models with the randomly varying coefficients consisting of white and coloured noise processes. The relationship and drawbacks of the extended algorithms are discussed. A new algorithm is proposed to improve the performances of the extended algorithms for the models with the randomly varying gain coefficient containing the white noise component. Secondly, the extension to the models with d(d≥2) input delays is considered. It is shown that predictors of the outputs up to d-1 step ahead are required for the realization of the active adaptive control algorithms. Finally, simulation results are presented to illustrate the performance levels of the algorithms.
