Skip to Main Content
A semi-definite programming (SDP) approach to compute the controllability radius is proposed in this paper. The initial nonconvex optimisation problem is transformed into the minimisation of the smallest eigenvalue of a bivariate real or trigonometric polynomial with matrix coefficients. A sum-of-squares relaxation leads to the SDP formulation. A similar technique is used for the computation of the stabilisability radius. The approach is extended to the computation of the worst-case controllability radius for systems that depend polynomially on a small number of parameters. Experimental results show that the proposed methods compete well with previous ones in a complexity/accuracy trade-off.