Fast numerical methods for multivariate polynomial optimization
Reference persons GIUSEPPE CARLO CALAFIORE
Research Groups SYSTEMS AND DATA SCIENCE - SDS
Thesis type RESEARCH THESIS
Description The minimization of a multivariate polynomial constitutes a key step in several important engineering problems, such as the synthesis of the control signal for discrete-time nonlinear dynamical systems. Exact and efficient computation of a global minimum is unaffordable, in general, and current state-of-the-art approximation techniques include the Sum of Squares (SOS) relaxation, which amounts to solving a convex semidefinite programming problem (SDP). SDPs resulting from SOS are, however, very large in practice, and their numerical solution may still be unviable, in real-world applications. In this thesis, we shall explore the use of fast coordinate minimization techniques for large-scale polynomial optimization, comparing their performance with state-of-the-art SOS methods and discussing control applications.
Required skills Matlab, optimization
Deadline 04/04/2018 PROPONI LA TUA CANDIDATURA