Adaptive predictive controllers based on orthonormal series representation

This work presents and compares a set of active and passive adaptive predic tive controllers based on orthonormal series representations. The controlle rs are designed according to the receding thorizon principle, where the par ameters of the predictive model are considered as random variables. This ge neral framework allows the design of active and passive adaptive controller s, according to the assumptions made over the prediction of the posterior d ensities of the random variables. Several simulated examples illustrate the characteristics for each algorithm and the performance improvement obtaine d by active adaptive controllers. (C) 1999 John Wiley & Sons Ltd.