English

On computing the distance to stability for matrices using linear dissipative Hamiltonian systems

Optimization and Control 2017-08-22 v2

Abstract

In this paper, we consider the problem of computing the nearest stable matrix to an unstable one. We propose new algorithms to solve this problem based on a reformulation using linear dissipative Hamiltonian systems: we show that a matrix AA is stable if and only if it can be written as A=(JR)QA = (J-R)Q, where J=JTJ=-J^T, R0R \succeq 0 and Q0Q \succ 0 (that is, RR is positive semidefinite and QQ is positive definite). This reformulation results in an equivalent optimization problem with a simple convex feasible set. We propose three strategies to solve the problem in variables (J,R,Q)(J,R,Q): (i) a block coordinate descent method, (ii) a projected gradient descent method, and (iii) a fast gradient method inspired from smooth convex optimization. These methods require O(n3)\mathcal{O}(n^3) operations per iteration, where nn is the size of AA. We show the effectiveness of the fast gradient method compared to the other approaches and to several state-of-the-art algorithms.

Keywords

Cite

@article{arxiv.1611.00595,
  title  = {On computing the distance to stability for matrices using linear dissipative Hamiltonian systems},
  author = {Nicolas Gillis and Punit Sharma},
  journal= {arXiv preprint arXiv:1611.00595},
  year   = {2017}
}

Comments

21 pages, 5 figures, 4 tables. Some typos and errors fixed, new remark on uniqueness issues

R2 v1 2026-06-22T16:39:42.887Z