Porteous Meets Walsh: Counting Critical Points for $\mathcal{H}_2$-Norm Model Order Reduction Using Algebraic Geometry Behind Multiparameter Eigenvalue Problems

Abstract

Several polynomial systems arising in systems and control have been shown to exhibit a partially linear structure, which allows them to be reformulated as a multiparameter eigenvalue problem (MEP). In this work, we show how such MEP formulations can be leveraged to obtain refined information about the solution set of the polynomial system. We demonstrate, in the setting of single-input single-output $\mathcal{H}_2$-norm model order reduction, how the Thom--Porteous formula can be used to derive an exact count of the complex solutions of the critical equations derived from Walsh's theorem. We prove, under suitable genericity assumptions, that their set of complex solutions is finite and that all solutions occur without multiplicity. Moreover, we derive an explicit formula for the total number of complex solutions as a function of the model orders. Throughout, we provide a self-contained and application-oriented exposition of the algebro-geometric tools required for our analysis, with the aim of making them accessible to the broader engineering community.