Learning finite-horizon optimal control with unknown control-affine dynamics

Yuqing Chen; Yangzhi Li; David J. Braun

doi:10.1016/j.sysconle.2025.106161

Learning finite-horizon optimal control with unknown control-affine dynamics

Yuqing Chen
, Yangzhi Li
, David J. Braun^*

^*Corresponding author for this work

Xi'an Jiaotong-Liverpool University

Singapore University of Technology and Design
Adaptive Machines LLC

Research output: Contribution to journal › Article › peer-review

1 Citation (Scopus)

Abstract

This paper introduces a method for learning finite-horizon optimal control for systems with control-affine dynamics without using the model of the system dynamics. We approximate the time- and state-dependent optimal control policy using model-free relearning of simple linear control policies. Assuming persistent excitation, we prove the convergence and optimality of the proposed learning method and demonstrate its use through a numerical example.

Original language	English
Article number	106161
Journal	Systems and Control Letters
Volume	203
DOIs	https://doi.org/10.1016/j.sysconle.2025.106161
Publication status	Published - Sept 2025

Keywords

Control-affine systems
Learning
Model-free optimal control

Access to Document

10.1016/j.sysconle.2025.106161

Cite this

@article{10dc57fd6ed14564ac649a7e96ff6878,

title = "Learning finite-horizon optimal control with unknown control-affine dynamics",

abstract = "This paper introduces a method for learning finite-horizon optimal control for systems with control-affine dynamics without using the model of the system dynamics. We approximate the time- and state-dependent optimal control policy using model-free relearning of simple linear control policies. Assuming persistent excitation, we prove the convergence and optimality of the proposed learning method and demonstrate its use through a numerical example.",

keywords = "Control-affine systems, Learning, Model-free optimal control",

author = "Yuqing Chen and Yangzhi Li and Braun, \{David J.\}",

note = "Publisher Copyright: {\textcopyright} 2025 Elsevier B.V.",

year = "2025",

month = sep,

doi = "10.1016/j.sysconle.2025.106161",

language = "English",

volume = "203",

journal = "Systems and Control Letters",

issn = "0167-6911",

}

TY - JOUR

T1 - Learning finite-horizon optimal control with unknown control-affine dynamics

AU - Chen, Yuqing

AU - Li, Yangzhi

AU - Braun, David J.

PY - 2025/9

Y1 - 2025/9

N2 - This paper introduces a method for learning finite-horizon optimal control for systems with control-affine dynamics without using the model of the system dynamics. We approximate the time- and state-dependent optimal control policy using model-free relearning of simple linear control policies. Assuming persistent excitation, we prove the convergence and optimality of the proposed learning method and demonstrate its use through a numerical example.

AB - This paper introduces a method for learning finite-horizon optimal control for systems with control-affine dynamics without using the model of the system dynamics. We approximate the time- and state-dependent optimal control policy using model-free relearning of simple linear control policies. Assuming persistent excitation, we prove the convergence and optimality of the proposed learning method and demonstrate its use through a numerical example.

KW - Control-affine systems

KW - Learning

KW - Model-free optimal control

UR - https://www.scopus.com/pages/publications/105008174130

U2 - 10.1016/j.sysconle.2025.106161

DO - 10.1016/j.sysconle.2025.106161

M3 - Article

AN - SCOPUS:105008174130

SN - 0167-6911

VL - 203

JO - Systems and Control Letters

JF - Systems and Control Letters

M1 - 106161

ER -

Learning finite-horizon optimal control with unknown control-affine dynamics

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this