Title:
Controlled interacting particle systems for estimation and sampling Abstract:
In recent years there has been a growing interest in developing controlled interacting particle systems as an alternative to stochastic Monte-Carlo methods for estimation and sampling. In numerical evaluations, it is often found that these control-based algorithms exhibit smaller simulation variance and better scaling properties with problem dimension when compared to the traditional methods. In this talk, I will give an overview of this class of algorithms to numerically solve the nonlinear filtering problem. Specifically, I will present a principled framework to design control laws for a system of particles in the mean-field limit as well as numerical methods to approximate the mean-field control law for a finite number of particles. I will conclude with applications to an attitude estimation problem and the sensory control of locomotory gaits. Bio:
Amirhossein Taghvaei is a Postdoctoral Scholar in the Department of Mechanical and Aerospace Engineering at University of California, Irvine. He obtained his Ph.D. in Mechanical Science and Engineering and the M.S. in Mathematics from University of Illinois at Urbana-Champaign. He is currently working in the areas of control theory and machine learning. |