ON THE FEASIBILITY PROBLEM OF LINEAR MATRIX INEQUALITIES

Abstract

Various problems in system and control theory can be numerically solved by translating them into linear matrix inequality problems. Although there are numerous software packages that solve LMIs, they provide us only one solution if the feasibility range is nonempty. The objective of this research is to develop a methodology for partial characterization of the feasibility region of a given LMI, and to define this region via conic combinations of a finite number of vectors. Towards the end, we first introduce an inner cone for a given LMI with at least one strictly definite element. Then, we define an outer cone for a given second order LMI with nondefinite elements, and propose a procedure for refining this cone so that it will intersect the feasibility region of the given LMI.