Adapt On-the-Go: Behavior Modulation for Single-Life Robot Deployment

Annie S Chen, Govind Chada, Laura Smith, Archit Sharma, Zipeng Fu, Sergey Levine, Chelsea Finn
Proceedings of The 4th Conference on Lifelong Learning Agents, PMLR 330:307-323, 2026.

Abstract

To succeed in the real world, robots must cope with situations that differ from those seen during training. We study the problem of adapting on-the-fly to such novel scenarios during deployment, by drawing upon a diverse repertoire of previously-learned behaviors. Our approach, Robust Autonomous Modulation (ROAM), introduces a mechanism based on the perceived value of pretrained behaviors to select and adapt pre-trained behaviors to the situation at hand. Crucially, this adaptation process all happens within a single episode at test time, without any human supervision. We provide theoretical analysis of our selection mechanism and demonstrate that ROAM enables a robot to adapt rapidly to changes in dynamics both in simulation and on a real Go1 quadruped, even successfully moving forward with roller skates on its feet. Our approach adapts over 2x as efficiently compared to existing methods when facing a variety of out-of-distribution situations during deployment by effectively choosing and adapting relevant behaviors on-the-fly.

Cite this Paper

BibTeX
@InProceedings{pmlr-v330-chen26a, title = {Adapt On-the-Go: Behavior Modulation for Single-Life Robot Deployment}, author = {Chen, Annie S and Chada, Govind and Smith, Laura and Sharma, Archit and Fu, Zipeng and Levine, Sergey and Finn, Chelsea}, booktitle = {Proceedings of The 4th Conference on Lifelong Learning Agents}, pages = {307--323}, year = {2026}, editor = {Chandar, Sarath and Pascanu, Razvan and Eaton, Eric and Liu, Bing and Mahmood, Rupam and Rannen-Triki, Amal}, volume = {330}, series = {Proceedings of Machine Learning Research}, month = {11--14 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v330/main/assets/chen26a/chen26a.pdf}, url = {https://proceedings.mlr.press/v330/chen26a.html}, abstract = {To succeed in the real world, robots must cope with situations that differ from those seen during training. We study the problem of adapting on-the-fly to such novel scenarios during deployment, by drawing upon a diverse repertoire of previously-learned behaviors. Our approach, Robust Autonomous Modulation (ROAM), introduces a mechanism based on the perceived value of pretrained behaviors to select and adapt pre-trained behaviors to the situation at hand. Crucially, this adaptation process all happens within a single episode at test time, without any human supervision. We provide theoretical analysis of our selection mechanism and demonstrate that ROAM enables a robot to adapt rapidly to changes in dynamics both in simulation and on a real Go1 quadruped, even successfully moving forward with roller skates on its feet. Our approach adapts over 2x as efficiently compared to existing methods when facing a variety of out-of-distribution situations during deployment by effectively choosing and adapting relevant behaviors on-the-fly.} }
Endnote
%0 Conference Paper %T Adapt On-the-Go: Behavior Modulation for Single-Life Robot Deployment %A Annie S Chen %A Govind Chada %A Laura Smith %A Archit Sharma %A Zipeng Fu %A Sergey Levine %A Chelsea Finn %B Proceedings of The 4th Conference on Lifelong Learning Agents %C Proceedings of Machine Learning Research %D 2026 %E Sarath Chandar %E Razvan Pascanu %E Eric Eaton %E Bing Liu %E Rupam Mahmood %E Amal Rannen-Triki %F pmlr-v330-chen26a %I PMLR %P 307--323 %U https://proceedings.mlr.press/v330/chen26a.html %V 330 %X To succeed in the real world, robots must cope with situations that differ from those seen during training. We study the problem of adapting on-the-fly to such novel scenarios during deployment, by drawing upon a diverse repertoire of previously-learned behaviors. Our approach, Robust Autonomous Modulation (ROAM), introduces a mechanism based on the perceived value of pretrained behaviors to select and adapt pre-trained behaviors to the situation at hand. Crucially, this adaptation process all happens within a single episode at test time, without any human supervision. We provide theoretical analysis of our selection mechanism and demonstrate that ROAM enables a robot to adapt rapidly to changes in dynamics both in simulation and on a real Go1 quadruped, even successfully moving forward with roller skates on its feet. Our approach adapts over 2x as efficiently compared to existing methods when facing a variety of out-of-distribution situations during deployment by effectively choosing and adapting relevant behaviors on-the-fly.
APA
Chen, A.S., Chada, G., Smith, L., Sharma, A., Fu, Z., Levine, S. & Finn, C.. (2026). Adapt On-the-Go: Behavior Modulation for Single-Life Robot Deployment. Proceedings of The 4th Conference on Lifelong Learning Agents, in Proceedings of Machine Learning Research 330:307-323 Available from https://proceedings.mlr.press/v330/chen26a.html.

Related Material