TY - JOUR
T1 - Towards Data-driven LQR with Koopmanizing Flows
AU - Bevanda, Petar
AU - Beier, Max
AU - Heshmati-Alamdari, Shahab
AU - Sosnowski, Stefan
AU - Hirche, Sandra
N1 - Publisher Copyright:
Copyright © 2022 The Authors.
PY - 2022/7/1
Y1 - 2022/7/1
N2 - We propose a novel framework for learning linear time-invariant (LTI) models for a class of continuous-time non-autonomous nonlinear dynamics based on a representation of Koopman operators. In general, the operator is infinite-dimensional but, crucially, linear. To utilize it for efficient LTI control design, we learn a finite representation of the Koopman operator that is linear in controls while concurrently learning meaningful lifting coordinates. For the latter, we rely on Koopmanizing Flows - a diffeomorphism-based representation of Koopman operators and extend it to systems with linear control entry. With such a learned model, we can replace the nonlinear optimal control problem with quadratic cost to that of a linear quadratic regulator (LQR), facilitating efficacious optimal control for nonlinear systems. The superior control performance of the proposed method is demonstrated on simulation examples.
AB - We propose a novel framework for learning linear time-invariant (LTI) models for a class of continuous-time non-autonomous nonlinear dynamics based on a representation of Koopman operators. In general, the operator is infinite-dimensional but, crucially, linear. To utilize it for efficient LTI control design, we learn a finite representation of the Koopman operator that is linear in controls while concurrently learning meaningful lifting coordinates. For the latter, we rely on Koopmanizing Flows - a diffeomorphism-based representation of Koopman operators and extend it to systems with linear control entry. With such a learned model, we can replace the nonlinear optimal control problem with quadratic cost to that of a linear quadratic regulator (LQR), facilitating efficacious optimal control for nonlinear systems. The superior control performance of the proposed method is demonstrated on simulation examples.
KW - Koopman operators
KW - Learning Systems
KW - Learning for control
KW - Machine learning
KW - Neural networks
KW - Representation Learning
UR - http://www.scopus.com/inward/record.url?scp=85142237508&partnerID=8YFLogxK
U2 - 10.1016/j.ifacol.2022.07.601
DO - 10.1016/j.ifacol.2022.07.601
M3 - Conference article
AN - SCOPUS:85142237508
SN - 1474-6670
VL - 55
SP - 13
EP - 18
JO - IFAC Proceedings Volumes (IFAC-PapersOnline)
JF - IFAC Proceedings Volumes (IFAC-PapersOnline)
IS - 15
T2 - 6th IFAC Conference on Intelligent Control and Automation Sciences, ICONS 2022
Y2 - 13 July 2022 through 15 July 2022
ER -