Robustness analysis of polynomials with polynomial parameter dependency using Bernstein expansion

From MaRDI portal





Let the parameter set \(Q\) be an \(l\)-dimensional box \(Q = [\underline q_1,\overline q_1]\times\dots\times[\underline q_{ l}, \overline q_l],\) and let a family of polynomials be given by \[ p(s,\mathbf q) = a_0(\mathbf q)s^m +\dots +a_{m-1}(\mathbf q)s+ a_m(\mathbf q), \] \(\mathbf q = (q_1,\dots,q_l),\) where the coefficients \(a_k(\mathbf q)\) are depending polynomially on the parameters \(q_i\), \(a_k(\mathbf q)=\sum_{i_1,\dots,i_l=0}^da_{i_1\dots i_l}^{(k)}q_1^{i_1}\dots q_l^{i_l}.\) The robust stability problem is considered: Are the polynomials \(p(s,\mathbf q)\) robustly stable for all \(\mathbf q\in Q\)? Here stability is meant in the sense of Hurwitz or asymptotic stability. Two algorithms for checking robust stability are presented. The first one is based on the expansion of the Hurwitz determinant into Bernstein polynomials. This leads to a fast algorithm. The second one avoids the blowing up of the problem caused by using the Hurwitz determinant and profits from the convex hull property of the Bernstein expansion. The underlying idea of this algorithm is to watch for zero crossing over the imaginary axis by inspecting the so-called value set. Numerical results to real-world control problems are presented showing the efficiency of both algorithms.




Cited in
(29)








This page was built for publication: Robustness analysis of polynomials with polynomial parameter dependency using Bernstein expansion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4396076)