A supervised discrete adaptive control scheme with multiestimation (Q1776795)

A linear and time-invariant discrete plant is decomposed into a modeled part plus an unmodeled one that includes different amounts of unmodeled dynamics through the use of different models. \(N_{\varepsilon}\) adaptive estimators are considered depending on the different decompositions of the plant into a modeled part plus an unmodeled one. The closed-loop stability is guaranteed if the contribution of the unmodeled dynamics to the output is sufficiently small compared to that of the modeled part of the plant. Each adaptive estimator, which can be indexed by an integer \(i\in N_{\varepsilon}\), has a different dimension according to the number of poles and zeros of the modeled part. All the estimators are running at the same time and each of them parameterizes its corresponding adaptive controller. Then, a supervisory quality index related to the associated identification error and control effort is used to decide online which estimator/adaptive controller pair has to be selected at each time interval to control the plant. The main objective of this work is to show how the transient response and the performance in general of an unknown plant can be improved by using a multiestimation-based adaptive controller while guaranteeing the closed-loop stability. The main novelty of the proposed approach compared to previous ones is that the various estimation models are of different orders while the corresponding plant unmodeled dynamics is modified accordingly from an estimation model to another one. This permits the design accommodation to the dominant plant frequencies