DeepKoCo: Efficient latent planning with a task-relevant Koopman representation

Research output: Chapter in Book/Conference proceedings/Edited volumeConference contributionScientificpeer-review

Abstract

This paper presents DeepKoCo, a novel modelbased agent that learns a latent Koopman representation from images. This representation allows DeepKoCo to plan efficiently using linear control methods, such as linear model predictive control. Compared to traditional agents, DeepKoCo learns taskrelevant dynamics, thanks to the use of a tailored lossy autoencoder network that allows DeepKoCo to learn latent dynamics that reconstruct and predict only observed costs, rather than all observed dynamics. As our results show, DeepKoCo achieves a similar final performance as traditional model-free methods on complex control tasks, while being considerably more robust to distractor dynamics, making the proposed agent more amenable for real-life applications.
Original languageEnglish
Title of host publicationProceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2021)
PublisherIEEE
Pages183-189
ISBN (Electronic)978-1-6654-1714-3
ISBN (Print)978-1-6654-1715-0
DOIs
Publication statusPublished - 2021
Event2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) - Online at Prague, Czech Republic
Duration: 27 Sep 20211 Oct 2021

Conference

Conference2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
Country/TerritoryCzech Republic
CityOnline at Prague
Period27/09/211/10/21

Bibliographical note

Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care
Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Keywords

  • Model-based reinforcement learning
  • Koopman theory
  • model-predictive control

Fingerprint

Dive into the research topics of 'DeepKoCo: Efficient latent planning with a task-relevant Koopman representation'. Together they form a unique fingerprint.

Cite this