The AAstrom-Wittenmark self-tuning regulator revisited and ELS-based adaptive trackers (Q3987145)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 30886
Language Label Description Also known as
default for all languages
No label defined
    English
    The AAstrom-Wittenmark self-tuning regulator revisited and ELS-based adaptive trackers
    scientific article; zbMATH DE number 30886

      Statements

      The AAstrom-Wittenmark self-tuning regulator revisited and ELS-based adaptive trackers (English)
      0 references
      0 references
      0 references
      28 June 1992
      0 references
      least-squares identification
      0 references
      ARMAX model
      0 references
      stochastic adaptive control
      0 references
      extended least-squares algorithm
      0 references
      stochastic discrete-time multidimensional closed-loop system
      0 references
      A rigorous proof is given for an answer to the open problem on the convergence of the original \(\mathring A-W\) self-tuning regulator in the basic problem of designing an adaptive control purely based on the extended least-squares (ELS) algorithm in order that the stochastic discrete-time multidimensional closed-loop system \(A(z)y_ n=B(z)u_{n- 1}+C(z)w_ n\) is globally stable, the tracking error is minimized and, by the algorithm, the obtained estimates \(\theta_ n\) of the sequence \(\theta\) of matrix coefficients of \(A\), \(B\), \(C\) are strongly consistent [\textit{K. J. Åström} and \textit{B. Wittenmark}, Automatica 9, 185-199 (1973; Zbl 0249.93049)]. In the case when the matrix coefficient \(B_ 1\) of \(B\) is known the authors have introduced the self-tuning tracker with the reference signal \(\{y^*_ n\}\) which is arbitrary a.s. bounded sequence not necessarily equal to zero and proved its stability and optimality. When \(B_ 1\) is unknown the ELS-based adaptive tracker with such properties is established. Under the assumption that \(A\), \(B\), \(C\) with known upper bounds \(p\), \(q\), \(r\) for orders have no common left factor and \([A_ p B_ q C_ r]\) is of full-row rank, the convergence of ELS-based adaptive trackers with paying attention to the consistence issue of the estimates is discussed and a solution to the basic problem is given in the last theorem by using the diminishing excitation technique [the authors, SIAM J. Control Optimization 25, 558-575 (1987; Zbl 0619.93048)]. In this case, the excitation source is defined by the sequence of \(m\)-dimensional i.i.d. random vectors independent of \(\{w_ i,y^*_ i\}\). The proofs of stability and optimality of trackers are based on two lemmas on estimations \(\|\varphi_{s,k}\|^ 2\), \(s=0,1\), of the sequences \[ \varphi_{s,k}=\left[y^ T_ k\cdots y^ T_{k-p+1} u^ T_{k-2}\cdots u^ T_{k-q+1}\hat w^ T_{s,k}\cdots\hat w^ T_{s,k-r+1}\right]^ T, \] where \(B_ 0\equiv 0\), \(\hat w_{s,n}=y_ n-B_ s u_{n-1}-\theta^ T_ n\varphi_{n- 1}\), \(n\geq 0\). The convergence rate of the estimates in the ELS recursion is devoted to Lemma 3 under the above mentioned assumptions. In concluding remarks, a question and a conjecture are formulated.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references