Neural Stochastic Contraction Metrics for Learning-based Control and Estimation
We present Neural Stochastic Contraction Metrics (NSCM), a new design framework for provably-stable learning-based control and estimation for a class of stochastic nonlinear systems. It uses a spectrally-normalized deep neural network to construct a contraction metric and its differential Lyapunov f...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
Institute of Electrical and Electronics Engineers (IEEE)
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139679 |
| _version_ | 1826216085244870656 |
|---|---|
| author | Tsukamoto, Hiroyasu Chung, Soon-Jo Slotine, Jean-Jacques E |
| author2 | Massachusetts Institute of Technology. Nonlinear Systems Laboratory |
| author_facet | Massachusetts Institute of Technology. Nonlinear Systems Laboratory Tsukamoto, Hiroyasu Chung, Soon-Jo Slotine, Jean-Jacques E |
| author_sort | Tsukamoto, Hiroyasu |
| collection | MIT |
| description | We present Neural Stochastic Contraction Metrics (NSCM), a new design framework for provably-stable learning-based control and estimation for a class of stochastic nonlinear systems. It uses a spectrally-normalized deep neural network to construct a contraction metric and its differential Lyapunov function, sampled via simplified convex optimization in the stochastic setting. Spectral normalization constrains the state-derivatives of the metric to be Lipschitz continuous, thereby ensuring exponential boundedness of the mean squared distance of system trajectories under stochastic disturbances. The trained NSCM model allows autonomous systems to approximate optimal stable control and estimation policies in real-time, and outperforms existing nonlinear control and estimation techniques including the state-dependent Riccati equation, iterative LQR, EKF, and the deterministic NCM, as shown in simulation results. |
| first_indexed | 2024-09-23T16:42:05Z |
| format | Article |
| id | mit-1721.1/139679 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T16:42:05Z |
| publishDate | 2022 |
| publisher | Institute of Electrical and Electronics Engineers (IEEE) |
| record_format | dspace |
| spelling | mit-1721.1/1396792023-01-10T15:20:35Z Neural Stochastic Contraction Metrics for Learning-based Control and Estimation Tsukamoto, Hiroyasu Chung, Soon-Jo Slotine, Jean-Jacques E Massachusetts Institute of Technology. Nonlinear Systems Laboratory We present Neural Stochastic Contraction Metrics (NSCM), a new design framework for provably-stable learning-based control and estimation for a class of stochastic nonlinear systems. It uses a spectrally-normalized deep neural network to construct a contraction metric and its differential Lyapunov function, sampled via simplified convex optimization in the stochastic setting. Spectral normalization constrains the state-derivatives of the metric to be Lipschitz continuous, thereby ensuring exponential boundedness of the mean squared distance of system trajectories under stochastic disturbances. The trained NSCM model allows autonomous systems to approximate optimal stable control and estimation policies in real-time, and outperforms existing nonlinear control and estimation techniques including the state-dependent Riccati equation, iterative LQR, EKF, and the deterministic NCM, as shown in simulation results. 2022-01-24T19:48:55Z 2022-01-24T19:48:55Z 2021 2022-01-24T19:36:39Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/139679 Tsukamoto, Hiroyasu, Chung, Soon-Jo and Slotine, Jean-Jacques E. 2021. "Neural Stochastic Contraction Metrics for Learning-based Control and Estimation." IEEE Control Systems Letters, 5 (5). en 10.1109/LCSYS.2020.3046529 IEEE Control Systems Letters Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Institute of Electrical and Electronics Engineers (IEEE) arXiv |
| spellingShingle | Tsukamoto, Hiroyasu Chung, Soon-Jo Slotine, Jean-Jacques E Neural Stochastic Contraction Metrics for Learning-based Control and Estimation |
| title | Neural Stochastic Contraction Metrics for Learning-based Control and Estimation |
| title_full | Neural Stochastic Contraction Metrics for Learning-based Control and Estimation |
| title_fullStr | Neural Stochastic Contraction Metrics for Learning-based Control and Estimation |
| title_full_unstemmed | Neural Stochastic Contraction Metrics for Learning-based Control and Estimation |
| title_short | Neural Stochastic Contraction Metrics for Learning-based Control and Estimation |
| title_sort | neural stochastic contraction metrics for learning based control and estimation |
| url | https://hdl.handle.net/1721.1/139679 |
| work_keys_str_mv | AT tsukamotohiroyasu neuralstochasticcontractionmetricsforlearningbasedcontrolandestimation AT chungsoonjo neuralstochasticcontractionmetricsforlearningbasedcontrolandestimation AT slotinejeanjacquese neuralstochasticcontractionmetricsforlearningbasedcontrolandestimation |