Some design guidelines for discrete-time adaptive controllers (Q1059595)

This paper presents an analysis, based upon linearization, around the equilibrium of the unsuitable behaviour (leading perhaps to instability) which appears in adaptive control systems when unmoduled dynamics are present. Several heuristical design rules around the choice of the sampling period, adaptive gains and nominal open-loop systems are given in order to dominish the possible drawbacks in the presence of unmodeled dynamics. In spite of the fact that the presented results and some of those referenced in this work of the same authors have been the first results on this topic, the analysis is often unrigorous. Linearizations of a nonlinear system around the equilibrium can lead to erroneous stability results (under purely imaginary system modes) for interpretation of the true nonlinear stability even locally around the equilibrium. The results about stability under constant reference sequences will be probably different, and non applicable to the case of other reference input types. Moreover,, adaptive systems have occasionally more than one equilibrium point. However, the investigations in this paper are interesting, since this problem (unmodeled dynamics) occurs in practice, and it is not known in classical control due to the plant filtering effects.