Print Email Facebook Twitter DeepKoCo Title DeepKoCo: Efficient latent planning with a task-relevant Koopman representation Author van der Heijden, D.S. (TU Delft Learning & Autonomous Control) Ferranti, L. (TU Delft Learning & Autonomous Control) Kober, J. (TU Delft Learning & Autonomous Control) Babuska, R. (TU Delft Learning & Autonomous Control) Date 2021 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. Subject Model-based reinforcement learningKoopman theorymodel-predictive control To reference this document use: http://resolver.tudelft.nl/uuid:cc306fd6-b16a-4756-b2ba-c2fb2de5290c DOI https://doi.org/10.1109/IROS51168.2021.9636408 Publisher IEEE Embargo date 2022-06-16 ISBN 978-1-6654-1715-0 Source Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2021) Event 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021-09-27 → 2021-10-01, Online at Prague, Czech Republic 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. Part of collection Institutional Repository Document type conference paper Rights © 2021 D.S. van der Heijden, L. Ferranti, J. Kober, R. Babuska Files PDF DeepKoCo_Efficient_latent ... tation.pdf 1.97 MB Close viewer /islandora/object/uuid:cc306fd6-b16a-4756-b2ba-c2fb2de5290c/datastream/OBJ/view