MOTO: Offline Pre-training to Online Fine-tuning for Model-based Robot Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909063476936704 |
|---|---|
| author | Rafailov, Rafael Hatch, Kyle Kolev, Victor Martin, John D. Phielipp, Mariano Finn, Chelsea |
| author_facet | Rafailov, Rafael Hatch, Kyle Kolev, Victor Martin, John D. Phielipp, Mariano Finn, Chelsea |
| contents | We study the problem of offline pre-training and online fine-tuning for reinforcement learning from high-dimensional observations in the context of realistic robot tasks. Recent offline model-free approaches successfully use online fine-tuning to either improve the performance of the agent over the data collection policy or adapt to novel tasks. At the same time, model-based RL algorithms have achieved significant progress in sample efficiency and the complexity of the tasks they can solve, yet remain under-utilized in the fine-tuning setting. In this work, we argue that existing model-based offline RL methods are not suitable for offline-to-online fine-tuning in high-dimensional domains due to issues with distribution shifts, off-dynamics data, and non-stationary rewards. We propose an on-policy model-based method that can efficiently reuse prior data through model-based value expansion and policy regularization, while preventing model exploitation by controlling epistemic uncertainty. We find that our approach successfully solves tasks from the MetaWorld benchmark, as well as the Franka Kitchen robot manipulation environment completely from images. To the best of our knowledge, MOTO is the first method to solve this environment from pixels. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_03306 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | MOTO: Offline Pre-training to Online Fine-tuning for Model-based Robot Learning Rafailov, Rafael Hatch, Kyle Kolev, Victor Martin, John D. Phielipp, Mariano Finn, Chelsea Machine Learning Artificial Intelligence Robotics We study the problem of offline pre-training and online fine-tuning for reinforcement learning from high-dimensional observations in the context of realistic robot tasks. Recent offline model-free approaches successfully use online fine-tuning to either improve the performance of the agent over the data collection policy or adapt to novel tasks. At the same time, model-based RL algorithms have achieved significant progress in sample efficiency and the complexity of the tasks they can solve, yet remain under-utilized in the fine-tuning setting. In this work, we argue that existing model-based offline RL methods are not suitable for offline-to-online fine-tuning in high-dimensional domains due to issues with distribution shifts, off-dynamics data, and non-stationary rewards. We propose an on-policy model-based method that can efficiently reuse prior data through model-based value expansion and policy regularization, while preventing model exploitation by controlling epistemic uncertainty. We find that our approach successfully solves tasks from the MetaWorld benchmark, as well as the Franka Kitchen robot manipulation environment completely from images. To the best of our knowledge, MOTO is the first method to solve this environment from pixels. |
| title | MOTO: Offline Pre-training to Online Fine-tuning for Model-based Robot Learning |
| topic | Machine Learning Artificial Intelligence Robotics |
| url | https://arxiv.org/abs/2401.03306 |