Loading [MathJax]/extensions/tex2jax.js

Iterative Model Learning and Dual Iterative Learning Control: A Unified Framework for Data-Driven Iterative Learning Control

Research output: Contribution to journalArticleResearchpeer review

Authors

Research Organisations

External Research Organisations

  • Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU Erlangen-Nürnberg)

Details

Original languageEnglish
JournalIEEE Transactions on Automatic Control
Publication statusE-pub ahead of print - 6 Jun 2025

Abstract

Accurate reference tracking is essential in control tasks, and, in repetitive systems, Model-Based Iterative Learning Control (MB-ILC) is a standard solution. However, MB-ILC suffers from two downsides: MB-ILC not only requires prior model information but also learning parameters that have to be manually tuned, which poses an inherent design effort. To overcome the requirement of model information, Data-Driven ILC (DD-ILC) methods have been proposed which, nonetheless, still require manual parameter tuning and also do not preserve the modularity and theoretical guarantees of MB-ILC. To overcome these issues, we propose the two frameworks of Iterative Model Learning (IML) and Dual Iterative Learning Control (DILC). The IML framework enables iterative learning of unknown dynamics in repetitive systems using input/output trajectory pairs, and we formally prove the duality of IML and ILC, i.e., an IML system is equivalent to an ILC system with a trial-varying reference and trial-varying but known dynamics. Hence, existing MB-ILC methods can be utilized within the IML framework to learn models of unknown dynamics. The proposed DILC framework combines IML and MB-ILC to modularly employ various MB-ILC methods and to relieve them of requiring prior model information. To overcome the need for manual parameter tuning, we propose systematic self-parametrization schemes that enable the proposed methods to self-reliantly determine necessary learning parameters. We formally investigate the convergence of the proposed methods, and both IML and DILC are validated in extensive simulations and real-world experiments. A comparison using simulations demonstrates that by means of self-parametrization the proposed DILC framework significantly outperforms two state-of-the-art DD-ILC methods.

Keywords

    Autonomous systems, iterative learning control, monotonic convergence, reference tracking

ASJC Scopus subject areas

Cite this

Download
@article{371e27709c6e4f079a2912f16189f183,
title = "Iterative Model Learning and Dual Iterative Learning Control: A Unified Framework for Data-Driven Iterative Learning Control",
abstract = "Accurate reference tracking is essential in control tasks, and, in repetitive systems, Model-Based Iterative Learning Control (MB-ILC) is a standard solution. However, MB-ILC suffers from two downsides: MB-ILC not only requires prior model information but also learning parameters that have to be manually tuned, which poses an inherent design effort. To overcome the requirement of model information, Data-Driven ILC (DD-ILC) methods have been proposed which, nonetheless, still require manual parameter tuning and also do not preserve the modularity and theoretical guarantees of MB-ILC. To overcome these issues, we propose the two frameworks of Iterative Model Learning (IML) and Dual Iterative Learning Control (DILC). The IML framework enables iterative learning of unknown dynamics in repetitive systems using input/output trajectory pairs, and we formally prove the duality of IML and ILC, i.e., an IML system is equivalent to an ILC system with a trial-varying reference and trial-varying but known dynamics. Hence, existing MB-ILC methods can be utilized within the IML framework to learn models of unknown dynamics. The proposed DILC framework combines IML and MB-ILC to modularly employ various MB-ILC methods and to relieve them of requiring prior model information. To overcome the need for manual parameter tuning, we propose systematic self-parametrization schemes that enable the proposed methods to self-reliantly determine necessary learning parameters. We formally investigate the convergence of the proposed methods, and both IML and DILC are validated in extensive simulations and real-world experiments. A comparison using simulations demonstrates that by means of self-parametrization the proposed DILC framework significantly outperforms two state-of-the-art DD-ILC methods.",
keywords = "Autonomous systems, iterative learning control, monotonic convergence, reference tracking",
author = "Michael Meindl and Simon Bachhuber and Thomas Seel",
note = "Publisher Copyright: {\textcopyright} 1963-2012 IEEE.",
year = "2025",
month = jun,
day = "6",
doi = "10.1109/TAC.2025.3577958",
language = "English",
journal = "IEEE Transactions on Automatic Control",
issn = "0018-9286",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

Download

TY - JOUR

T1 - Iterative Model Learning and Dual Iterative Learning Control

T2 - A Unified Framework for Data-Driven Iterative Learning Control

AU - Meindl, Michael

AU - Bachhuber, Simon

AU - Seel, Thomas

N1 - Publisher Copyright: © 1963-2012 IEEE.

PY - 2025/6/6

Y1 - 2025/6/6

N2 - Accurate reference tracking is essential in control tasks, and, in repetitive systems, Model-Based Iterative Learning Control (MB-ILC) is a standard solution. However, MB-ILC suffers from two downsides: MB-ILC not only requires prior model information but also learning parameters that have to be manually tuned, which poses an inherent design effort. To overcome the requirement of model information, Data-Driven ILC (DD-ILC) methods have been proposed which, nonetheless, still require manual parameter tuning and also do not preserve the modularity and theoretical guarantees of MB-ILC. To overcome these issues, we propose the two frameworks of Iterative Model Learning (IML) and Dual Iterative Learning Control (DILC). The IML framework enables iterative learning of unknown dynamics in repetitive systems using input/output trajectory pairs, and we formally prove the duality of IML and ILC, i.e., an IML system is equivalent to an ILC system with a trial-varying reference and trial-varying but known dynamics. Hence, existing MB-ILC methods can be utilized within the IML framework to learn models of unknown dynamics. The proposed DILC framework combines IML and MB-ILC to modularly employ various MB-ILC methods and to relieve them of requiring prior model information. To overcome the need for manual parameter tuning, we propose systematic self-parametrization schemes that enable the proposed methods to self-reliantly determine necessary learning parameters. We formally investigate the convergence of the proposed methods, and both IML and DILC are validated in extensive simulations and real-world experiments. A comparison using simulations demonstrates that by means of self-parametrization the proposed DILC framework significantly outperforms two state-of-the-art DD-ILC methods.

AB - Accurate reference tracking is essential in control tasks, and, in repetitive systems, Model-Based Iterative Learning Control (MB-ILC) is a standard solution. However, MB-ILC suffers from two downsides: MB-ILC not only requires prior model information but also learning parameters that have to be manually tuned, which poses an inherent design effort. To overcome the requirement of model information, Data-Driven ILC (DD-ILC) methods have been proposed which, nonetheless, still require manual parameter tuning and also do not preserve the modularity and theoretical guarantees of MB-ILC. To overcome these issues, we propose the two frameworks of Iterative Model Learning (IML) and Dual Iterative Learning Control (DILC). The IML framework enables iterative learning of unknown dynamics in repetitive systems using input/output trajectory pairs, and we formally prove the duality of IML and ILC, i.e., an IML system is equivalent to an ILC system with a trial-varying reference and trial-varying but known dynamics. Hence, existing MB-ILC methods can be utilized within the IML framework to learn models of unknown dynamics. The proposed DILC framework combines IML and MB-ILC to modularly employ various MB-ILC methods and to relieve them of requiring prior model information. To overcome the need for manual parameter tuning, we propose systematic self-parametrization schemes that enable the proposed methods to self-reliantly determine necessary learning parameters. We formally investigate the convergence of the proposed methods, and both IML and DILC are validated in extensive simulations and real-world experiments. A comparison using simulations demonstrates that by means of self-parametrization the proposed DILC framework significantly outperforms two state-of-the-art DD-ILC methods.

KW - Autonomous systems

KW - iterative learning control

KW - monotonic convergence

KW - reference tracking

UR - http://www.scopus.com/inward/record.url?scp=105007645940&partnerID=8YFLogxK

U2 - 10.1109/TAC.2025.3577958

DO - 10.1109/TAC.2025.3577958

M3 - Article

AN - SCOPUS:105007645940

JO - IEEE Transactions on Automatic Control

JF - IEEE Transactions on Automatic Control

SN - 0018-9286

ER -

By the same author(s)