Convergence and logarithm laws of self-tuning regulators (Q1893000)

This paper presents an extension of the author's previous results [\textit{H. F. Chen} and the author, Identification and stochastic adaptive control (Birkhäuser, 1991; Zbl 0747.93002)], [the author, SIAM J. Control Optimization 32, 187--212 (1994; Zbl 0796.93113)] known as Chen-Guo convergence criteria for least-squares based self-tuning controllers. More precisely, the author studies in a unified way such properties like stability, optimality, consistency and the best convergence rate for such regulators and proposes some new techniques for the analysis of adaptive control systems. Among the problems solved here, the most interesting issues, devoted to establishing the logarithmic law and the best convergence rate for LS-based self-tuning controllers and the consistency of LS estimate in adaptive control systems in the absence of external excitations. The class of systems considered covers single input single output systems with unit delays and white noise disturbance but the results could be easily extended for more general MIMO systems with multiple delays and coloured noises. Moreover, the problem of convergence in the case of unknown high frequency gain is discussed. The author presents also an excellent survey of the recent results in the area of convergence of self-tuning algorithms.