Jorge Cortés
Professor
Cymer Corporation Endowed Chair
Online stochastic optimization for unknown linear systems: data-driven controller synthesis and analysis
G. Bianchin, M. Vaquero, J. Cortés, E. Dall'Anese
IEEE Transactions on Automatic Control 69 (7) (2024), to appear
Abstract
This paper proposes a data-driven control framework to steer an
unknown, stochastic linear system towards the solution of a
(stochastic) convex optimization problem. Despite the centrality of
this problem in many modern engineering and scientific applications,
most available methods critically rely on a precise knowledge of the
system dynamics, thus requiring prior phases for system identification
and model refinement before they can be applied. Instead, here we
show that the transfer function of a linear system can be computed
from control experiments without knowledge or estimation of the system
model. Such direct computation allows us to avoid any explicit model
identification and, numerically, it is as accurate as the model-based
computation. Inspired by stochastic gradient descent, we leverage this
data-driven representation to design a controller that drives the
system to the solution of the prescribed optimization problem without
requiring any knowledge of the time-varying disturbances affecting the
model equation. Our technical analysis combines concepts and tools
from behavioral theory, stochastic optimization with
decision-dependent distributions, and online optimization with
measurement feedback. We illustrate the applicability of the framework
on a case study for ride service scheduling.
pdf
Mechanical and Aerospace Engineering,
University of California, San Diego
9500 Gilman Dr,
La Jolla, California, 92093-0411
Ph: 1-858-822-7930
Fax: 1-858-822-3107
cortes at ucsd.edu
Skype id:
jorgilliyo