Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910894316847104 |
|---|---|
| author | Abyaneh, Amin Boroujeni, Mahrokh G. Lin, Hsiu-Chin Ferrari-Trecate, Giancarlo |
| author_facet | Abyaneh, Amin Boroujeni, Mahrokh G. Lin, Hsiu-Chin Ferrari-Trecate, Giancarlo |
| contents | Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamical systems guarantees convergence to a desired state, it often overlooks transient behavior. We propose a framework for learning policies modeled by contractive dynamical systems, ensuring that all policy rollouts converge regardless of perturbations, and in turn, enable efficient OOS recovery. By leveraging recurrent equilibrium networks and coupling layers, the policy structure guarantees contractivity for any parameter choice, which facilitates unconstrained optimization. We also provide theoretical upper bounds for worst-case and expected loss to rigorously establish the reliability of our method in deployment. Empirically, we demonstrate substantial OOS performance improvements for simulated robotic manipulation and navigation tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_07544 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery Abyaneh, Amin Boroujeni, Mahrokh G. Lin, Hsiu-Chin Ferrari-Trecate, Giancarlo Machine Learning Robotics Imitation learning is a data-driven approach to learning policies from expert behavior, but it is prone to unreliable outcomes in out-of-sample (OOS) regions. While previous research relying on stable dynamical systems guarantees convergence to a desired state, it often overlooks transient behavior. We propose a framework for learning policies modeled by contractive dynamical systems, ensuring that all policy rollouts converge regardless of perturbations, and in turn, enable efficient OOS recovery. By leveraging recurrent equilibrium networks and coupling layers, the policy structure guarantees contractivity for any parameter choice, which facilitates unconstrained optimization. We also provide theoretical upper bounds for worst-case and expected loss to rigorously establish the reliability of our method in deployment. Empirically, we demonstrate substantial OOS performance improvements for simulated robotic manipulation and navigation tasks. |
| title | Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery |
| topic | Machine Learning Robotics |
| url | https://arxiv.org/abs/2412.07544 |