Adaptive Optimal Output Regulation of Time-Delay Systems via Measurement Feedback

Weinan Gao, Zhong-Ping Jiang

Research output: Contribution to journalArticle

Abstract

This brief proposes a novel solution to problems related to the measurement feedback adaptive optimal output regulation of discrete-time linear systems with input time-delay. Based on reinforcement learning and adaptive dynamic programming, an approximate optimal control policy is obtained via recursive numerical algorithms using online information. Convergence proofs for the proposed algorithms are given. Notably, the exact knowledge of the plant and the exosystem is not needed. The learned control policy is only a function of retrospective input and measurement output data. Theoretical analysis and an application to a grid-connected inverter show that the proposed methodologies serve as effective tools for solving adaptive and optimal output regulation problems.

Original languageEnglish (US)
JournalIEEE Transactions on Neural Networks and Learning Systems
DOIs
StateAccepted/In press - Jul 23 2018

Fingerprint

Time delay
Feedback
Reinforcement learning
Dynamic programming
Linear systems

Keywords

  • Adaptive systems
  • Dynamic programming
  • Feedback control
  • Learning systems
  • Linear systems
  • Measurement feedback control
  • Optimal control
  • optimal control
  • output regulation
  • Regulators
  • reinforcement learning
  • time-delay systems.

ASJC Scopus subject areas

  • Software
  • Computer Science Applications
  • Computer Networks and Communications
  • Artificial Intelligence

Cite this

@article{0f7a9221a4124eae8c7aa5aabda181f4,
title = "Adaptive Optimal Output Regulation of Time-Delay Systems via Measurement Feedback",
abstract = "This brief proposes a novel solution to problems related to the measurement feedback adaptive optimal output regulation of discrete-time linear systems with input time-delay. Based on reinforcement learning and adaptive dynamic programming, an approximate optimal control policy is obtained via recursive numerical algorithms using online information. Convergence proofs for the proposed algorithms are given. Notably, the exact knowledge of the plant and the exosystem is not needed. The learned control policy is only a function of retrospective input and measurement output data. Theoretical analysis and an application to a grid-connected inverter show that the proposed methodologies serve as effective tools for solving adaptive and optimal output regulation problems.",
keywords = "Adaptive systems, Dynamic programming, Feedback control, Learning systems, Linear systems, Measurement feedback control, Optimal control, optimal control, output regulation, Regulators, reinforcement learning, time-delay systems.",
author = "Weinan Gao and Zhong-Ping Jiang",
year = "2018",
month = "7",
day = "23",
doi = "10.1109/TNNLS.2018.2850520",
language = "English (US)",
journal = "IEEE Transactions on Neural Networks and Learning Systems",
issn = "2162-237X",
publisher = "IEEE Computational Intelligence Society",

}

TY - JOUR

T1 - Adaptive Optimal Output Regulation of Time-Delay Systems via Measurement Feedback

AU - Gao, Weinan

AU - Jiang, Zhong-Ping

PY - 2018/7/23

Y1 - 2018/7/23

N2 - This brief proposes a novel solution to problems related to the measurement feedback adaptive optimal output regulation of discrete-time linear systems with input time-delay. Based on reinforcement learning and adaptive dynamic programming, an approximate optimal control policy is obtained via recursive numerical algorithms using online information. Convergence proofs for the proposed algorithms are given. Notably, the exact knowledge of the plant and the exosystem is not needed. The learned control policy is only a function of retrospective input and measurement output data. Theoretical analysis and an application to a grid-connected inverter show that the proposed methodologies serve as effective tools for solving adaptive and optimal output regulation problems.

AB - This brief proposes a novel solution to problems related to the measurement feedback adaptive optimal output regulation of discrete-time linear systems with input time-delay. Based on reinforcement learning and adaptive dynamic programming, an approximate optimal control policy is obtained via recursive numerical algorithms using online information. Convergence proofs for the proposed algorithms are given. Notably, the exact knowledge of the plant and the exosystem is not needed. The learned control policy is only a function of retrospective input and measurement output data. Theoretical analysis and an application to a grid-connected inverter show that the proposed methodologies serve as effective tools for solving adaptive and optimal output regulation problems.

KW - Adaptive systems

KW - Dynamic programming

KW - Feedback control

KW - Learning systems

KW - Linear systems

KW - Measurement feedback control

KW - Optimal control

KW - optimal control

KW - output regulation

KW - Regulators

KW - reinforcement learning

KW - time-delay systems.

UR - http://www.scopus.com/inward/record.url?scp=85050640300&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050640300&partnerID=8YFLogxK

U2 - 10.1109/TNNLS.2018.2850520

DO - 10.1109/TNNLS.2018.2850520

M3 - Article

C2 - 30047903

AN - SCOPUS:85050640300

JO - IEEE Transactions on Neural Networks and Learning Systems

JF - IEEE Transactions on Neural Networks and Learning Systems

SN - 2162-237X

ER -