ImitationFlow : learning deep stable stochastic dynamic systems by normalizing flows
Urain, Julen ; Ginesi, Michele ; Tateo, Davide LU
and Peters, Jan
(2020)
2020 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2020
In IEEE International Conference on Intelligent Robots and Systems
p.5231-5237
- Abstract
We introduce ImitationFlow, a novel Deep generative model that allows learning complex globally stable, stochastic, nonlinear dynamics. Our approach extends the Normalizing Flows framework to learn stable Stochastic Differential Equations. We prove the Lyapunov stability for a class of Stochastic Differential Equations and we propose a learning algorithm to learn them from a set of demonstrated trajectories. Our model extends the set of stable dynamical systems that can be represented by state-of-the-art approaches, eliminates the Gaussian assumption on the demonstrations, and outperforms the previous algorithms in terms of representation accuracy. We show the effectiveness of our method with both standard datasets and a real robot... (More)
We introduce ImitationFlow, a novel Deep generative model that allows learning complex globally stable, stochastic, nonlinear dynamics. Our approach extends the Normalizing Flows framework to learn stable Stochastic Differential Equations. We prove the Lyapunov stability for a class of Stochastic Differential Equations and we propose a learning algorithm to learn them from a set of demonstrated trajectories. Our model extends the set of stable dynamical systems that can be represented by state-of-the-art approaches, eliminates the Gaussian assumption on the demonstrations, and outperforms the previous algorithms in terms of representation accuracy. We show the effectiveness of our method with both standard datasets and a real robot experiment.
(Less)
- author
- Urain, Julen
; Ginesi, Michele
; Tateo, Davide
LU
and Peters, Jan
- publishing date
- 2020-10-24
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2020
- series title
- IEEE International Conference on Intelligent Robots and Systems
- article number
- 9341035
- pages
- 7 pages
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- conference name
- 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2020
- conference location
- Las Vegas, United States
- conference dates
- 2020-10-24 - 2021-01-24
- external identifiers
-
- scopus:85100282176
- ISSN
- 2153-0858
- 2153-0866
- ISBN
- 9781728162126
- DOI
- 10.1109/IROS45743.2020.9341035
- language
- English
- LU publication?
- no
- id
- 9cccd7d8-b83f-4442-bd13-e12a48f5ed2a
- date added to LUP
- 2025-10-16 14:39:54
- date last changed
- 2025-10-23 03:43:30
@inproceedings{9cccd7d8-b83f-4442-bd13-e12a48f5ed2a,
abstract = {{<p>We introduce ImitationFlow, a novel Deep generative model that allows learning complex globally stable, stochastic, nonlinear dynamics. Our approach extends the Normalizing Flows framework to learn stable Stochastic Differential Equations. We prove the Lyapunov stability for a class of Stochastic Differential Equations and we propose a learning algorithm to learn them from a set of demonstrated trajectories. Our model extends the set of stable dynamical systems that can be represented by state-of-the-art approaches, eliminates the Gaussian assumption on the demonstrations, and outperforms the previous algorithms in terms of representation accuracy. We show the effectiveness of our method with both standard datasets and a real robot experiment.</p>}},
author = {{Urain, Julen and Ginesi, Michele and Tateo, Davide and Peters, Jan}},
booktitle = {{2020 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2020}},
isbn = {{9781728162126}},
issn = {{2153-0858}},
language = {{eng}},
month = {{10}},
pages = {{5231--5237}},
publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
series = {{IEEE International Conference on Intelligent Robots and Systems}},
title = {{ImitationFlow : learning deep stable stochastic dynamic systems by normalizing flows}},
url = {{http://dx.doi.org/10.1109/IROS45743.2020.9341035}},
doi = {{10.1109/IROS45743.2020.9341035}},
year = {{2020}},
}