Learning Stabilizable Dynamical Systems via Control Contraction Metrics
We propose a novel framework for learning stabilizable nonlinear dynamical systems for continuous control tasks in robotics. The key idea is to develop a new control-theoretic regularizer for dynamics fitting rooted in the notion of stabilizability, which guarantees that the learned system can be ac...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
Springer International Publishing
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139674 |
| _version_ | 1826201618855493632 |
|---|---|
| author | Singh, Sumeet Sindhwani, Vikas Slotine, Jean-Jacques E Pavone, Marco |
| author2 | Massachusetts Institute of Technology. Department of Mechanical Engineering |
| author_facet | Massachusetts Institute of Technology. Department of Mechanical Engineering Singh, Sumeet Sindhwani, Vikas Slotine, Jean-Jacques E Pavone, Marco |
| author_sort | Singh, Sumeet |
| collection | MIT |
| description | We propose a novel framework for learning stabilizable nonlinear dynamical systems for continuous control tasks in robotics. The key idea is to develop a new control-theoretic regularizer for dynamics fitting rooted in the notion of stabilizability, which guarantees that the learned system can be accompanied by a robust controller capable of stabilizing any open-loop trajectory that the system may generate. By leveraging tools from contraction theory, statistical learning, and convex optimization, we provide a general and tractable semi-supervised algorithm to learn stabilizable dynamics, which can be applied to complex underactuated systems. We validated the proposed algorithm on a simulated planar quadrotor system and observed notably improved trajectory generation and tracking performance with the control-theoretic regularized model over models learned using traditional regression techniques, especially when using a small number of demonstration examples. The results presented illustrate the need to infuse standard model-based reinforcement learning algorithms with concepts drawn from nonlinear control theory for improved reliability. |
| first_indexed | 2024-09-23T11:54:12Z |
| format | Article |
| id | mit-1721.1/139674 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T11:54:12Z |
| publishDate | 2022 |
| publisher | Springer International Publishing |
| record_format | dspace |
| spelling | mit-1721.1/1396742023-04-11T21:03:23Z Learning Stabilizable Dynamical Systems via Control Contraction Metrics Singh, Sumeet Sindhwani, Vikas Slotine, Jean-Jacques E Pavone, Marco Massachusetts Institute of Technology. Department of Mechanical Engineering We propose a novel framework for learning stabilizable nonlinear dynamical systems for continuous control tasks in robotics. The key idea is to develop a new control-theoretic regularizer for dynamics fitting rooted in the notion of stabilizability, which guarantees that the learned system can be accompanied by a robust controller capable of stabilizing any open-loop trajectory that the system may generate. By leveraging tools from contraction theory, statistical learning, and convex optimization, we provide a general and tractable semi-supervised algorithm to learn stabilizable dynamics, which can be applied to complex underactuated systems. We validated the proposed algorithm on a simulated planar quadrotor system and observed notably improved trajectory generation and tracking performance with the control-theoretic regularized model over models learned using traditional regression techniques, especially when using a small number of demonstration examples. The results presented illustrate the need to infuse standard model-based reinforcement learning algorithms with concepts drawn from nonlinear control theory for improved reliability. 2022-01-24T19:08:24Z 2022-01-24T19:08:24Z 2020 2022-01-24T19:04:34Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/139674 Singh, Sumeet, Sindhwani, Vikas, Slotine, Jean-Jacques E and Pavone, Marco. 2020. "Learning Stabilizable Dynamical Systems via Control Contraction Metrics." 14. en 10.1007/978-3-030-44051-0_11 Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Springer International Publishing arXiv |
| spellingShingle | Singh, Sumeet Sindhwani, Vikas Slotine, Jean-Jacques E Pavone, Marco Learning Stabilizable Dynamical Systems via Control Contraction Metrics |
| title | Learning Stabilizable Dynamical Systems via Control Contraction Metrics |
| title_full | Learning Stabilizable Dynamical Systems via Control Contraction Metrics |
| title_fullStr | Learning Stabilizable Dynamical Systems via Control Contraction Metrics |
| title_full_unstemmed | Learning Stabilizable Dynamical Systems via Control Contraction Metrics |
| title_short | Learning Stabilizable Dynamical Systems via Control Contraction Metrics |
| title_sort | learning stabilizable dynamical systems via control contraction metrics |
| url | https://hdl.handle.net/1721.1/139674 |
| work_keys_str_mv | AT singhsumeet learningstabilizabledynamicalsystemsviacontrolcontractionmetrics AT sindhwanivikas learningstabilizabledynamicalsystemsviacontrolcontractionmetrics AT slotinejeanjacquese learningstabilizabledynamicalsystemsviacontrolcontractionmetrics AT pavonemarco learningstabilizabledynamicalsystemsviacontrolcontractionmetrics |