An efficient algorithm for checking the robust stability of a polytope of polynomials (Q809043)

The author studies the family of polynomials \[ p(s;\delta)=:s^ n+a_ n(\delta)s^{n-1}+...+a_ 2(\delta)s+a_ 1(\delta), \] where \(\delta \equiv [\delta_ 1...\delta_ m]\in {\mathcal P}\) and \({\mathcal P}\) is a convex polytope in \({\mathbb{R}}^ n\), and \(a_ i(\delta)=:\sum^{m}_{i=1}y_{ji}\delta_ i+y_{i0},\) \(j=1,...,n\). Let \(\Omega (j\omega)=:\{p(j\omega;\delta)/\delta \in {\mathcal P}\}\) be the image of \({\mathcal P}\) in the complex plane \({\mathbb{C}}\) under p(j\(\omega\) ;\(\delta\)). Then the polytope of polynomials \({\mathcal I}\) is robustly stable if and only if \(p(j\omega;\delta^ 0)\) is strictly Hurwitz for some \(\delta^ 0\in {\mathcal P}\) which corresponds to the nominal plant, and \(0\not\in \Omega (j\omega)\) for all \(\omega\in [0,\infty)\). The author proves that the condition (*) needs to be checked only at a finite number of frequence points, namely, at the positive zeros of polynomials corresponding to the exposed edges in the parameter space. Then the problem is reduced to a parametric linear program.