Uniformly optimal control of linear time-varying systems (Q761404): Difference between revisions

A certain control structure is considered with cost function \(\cos t=\sup (\| y\|^ 2_ 2+\| u\|^ 2_ 2)^{1/2}\) over all d in D, where \(D=\{d:\) \(d=W_ 1v\), \(v\in h^ p_ 2\), \(\| v\|_ 2=1\}\), \(W_ 1\) is a linear operator representing the filter of output, \(h^ p_ 2\) is the direct sum of p copies of \(h_ 2\), which is the Hilbert space of square-summable complex-valued sequences \(\{f_ i: i\geq 0\}\). The cost function is to be minimized over all causal controllers achieving internal stability. Zames' work [see e.g. \textit{G. Zames}, IEEE Trans. Autom. Control AC-26, 301-320 (1981; Zbl 0474.93025)] gives the following formula: \(C=Q(I-PQ)^{-1}\) for \(Q\in C^{p,m}\) where P is a linear operator representing an object, \(C^{p,m}\) is a causal operator, m, p are numbers of inputs and outputs respectively. This formula parametrizes all above mentioned controllers. In the paper Zames' cost function is reformulated as follows: \[ \cos t=\| T-RQW_ 1\|,\quad where\quad T=\left[ \begin{matrix} W_ 2,W_ 1\\ 0\end{matrix} \right]\in C^{p,p+m};\quad R=\left[ \begin{matrix} W_ 2,P\\ 0\end{matrix} \right]\in C^{p,p+m}, \] \(W_ 2\) is a linear operator representing the output filter. It is shown that an optimal controller described by Zames' formula exists and the infimal cost is given by the formula: \(v=\inf \{\| T-RQW_ 1\|:\quad Q\in C^{p,m}\}.\)