Finetuning Offline World Models in the Real World

Yunhai Feng, Nicklas Hansen, Ziyan Xiong, Chandramouli Rajagopalan, Xiaolong Wang
Proceedings of The 7th Conference on Robot Learning, PMLR 229:425-445, 2023.

Abstract

Reinforcement Learning (RL) is notoriously data-inefficient, which makes training on a real robot difficult. While model-based RL algorithms (world models) improve data-efficiency to some extent, they still require hours or days of interaction to learn skills. Recently, offline RL has been proposed as a framework for training RL policies on pre-existing datasets without any online interaction. However, constraining an algorithm to a fixed dataset induces a state-action distribution shift between training and inference, and limits its applicability to new tasks. In this work, we seek to get the best of both worlds: we consider the problem of pretraining a world model with offline data collected on a real robot, and then finetuning the model on online data collected by planning with the learned model. To mitigate extrapolation errors during online interaction, we propose to regularize the planner at test-time by balancing estimated returns and (epistemic) model uncertainty. We evaluate our method on a variety of visuo-motor control tasks in simulation and on a real robot, and find that our method enables few-shot finetuning to seen and unseen tasks even when offline data is limited. Videos are available at https://yunhaifeng.com/FOWM

Cite this Paper

BibTeX
@InProceedings{pmlr-v229-feng23a, title = {Finetuning Offline World Models in the Real World}, author = {Feng, Yunhai and Hansen, Nicklas and Xiong, Ziyan and Rajagopalan, Chandramouli and Wang, Xiaolong}, booktitle = {Proceedings of The 7th Conference on Robot Learning}, pages = {425--445}, year = {2023}, editor = {Tan, Jie and Toussaint, Marc and Darvish, Kourosh}, volume = {229}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v229/feng23a/feng23a.pdf}, url = {https://proceedings.mlr.press/v229/feng23a.html}, abstract = {Reinforcement Learning (RL) is notoriously data-inefficient, which makes training on a real robot difficult. While model-based RL algorithms (world models) improve data-efficiency to some extent, they still require hours or days of interaction to learn skills. Recently, offline RL has been proposed as a framework for training RL policies on pre-existing datasets without any online interaction. However, constraining an algorithm to a fixed dataset induces a state-action distribution shift between training and inference, and limits its applicability to new tasks. In this work, we seek to get the best of both worlds: we consider the problem of pretraining a world model with offline data collected on a real robot, and then finetuning the model on online data collected by planning with the learned model. To mitigate extrapolation errors during online interaction, we propose to regularize the planner at test-time by balancing estimated returns and (epistemic) model uncertainty. We evaluate our method on a variety of visuo-motor control tasks in simulation and on a real robot, and find that our method enables few-shot finetuning to seen and unseen tasks even when offline data is limited. Videos are available at https://yunhaifeng.com/FOWM} }
Endnote
%0 Conference Paper %T Finetuning Offline World Models in the Real World %A Yunhai Feng %A Nicklas Hansen %A Ziyan Xiong %A Chandramouli Rajagopalan %A Xiaolong Wang %B Proceedings of The 7th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Jie Tan %E Marc Toussaint %E Kourosh Darvish %F pmlr-v229-feng23a %I PMLR %P 425--445 %U https://proceedings.mlr.press/v229/feng23a.html %V 229 %X Reinforcement Learning (RL) is notoriously data-inefficient, which makes training on a real robot difficult. While model-based RL algorithms (world models) improve data-efficiency to some extent, they still require hours or days of interaction to learn skills. Recently, offline RL has been proposed as a framework for training RL policies on pre-existing datasets without any online interaction. However, constraining an algorithm to a fixed dataset induces a state-action distribution shift between training and inference, and limits its applicability to new tasks. In this work, we seek to get the best of both worlds: we consider the problem of pretraining a world model with offline data collected on a real robot, and then finetuning the model on online data collected by planning with the learned model. To mitigate extrapolation errors during online interaction, we propose to regularize the planner at test-time by balancing estimated returns and (epistemic) model uncertainty. We evaluate our method on a variety of visuo-motor control tasks in simulation and on a real robot, and find that our method enables few-shot finetuning to seen and unseen tasks even when offline data is limited. Videos are available at https://yunhaifeng.com/FOWM
APA
Feng, Y., Hansen, N., Xiong, Z., Rajagopalan, C. & Wang, X.. (2023). Finetuning Offline World Models in the Real World. Proceedings of The 7th Conference on Robot Learning, in Proceedings of Machine Learning Research 229:425-445 Available from https://proceedings.mlr.press/v229/feng23a.html.

Related Material