Cross-Controller Transfer via In-Context Behavior Adaptation
Cross-Controller Transfer via In-Context Behavior Adaptation
This research paper introduces a method to transfer robot data across diverse embodiments and controllers without conflicting action interpretations. The proposed framework uses a transformer context encoder to summarize behavioral interactions, enabling a single policy to adapt in context. Experiments show the method improves average success by 33 percentage points over naive data pooling, achieves 50% success on unseen controller modes, and 52% on unseen controller gains. The approach is demonstrated on contact-rich tasks like USB insertion and surface wiping, with results available at https://controller-transfer.github.io.
The authors present a framework for transferring robot policies across different controllers and embodiments using in-context behavior adaptation, reporting specific success metrics on unseen controller modes and gains based on real-world experiments.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
robots
6
View original evidence
data across both embodiments and controllers without allowing conflicting action interpretations to interfere with learning. Our setting spans six controller types as well as variations in controller gains and compliance. We propose a framework that encodes behavioral interaction under a controller through examples of commanded actions and the resulting robot motions. A transformer context encoder summarizes these in
Open source S3controller types
6
View original evidence
data across both embodiments and controllers without allowing conflicting action interpretations to interfere with learning. Our setting spans six controller types as well as variations in controller gains and compliance. We propose a framework that encodes behavioral interaction under a controller through examples of commanded actions and the resulting robot motions. A transformer context encoder summarizes these in
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
teractions, allowing a single policy to adapt in context. Across real-world experiments, our method improves average success by 33 percentage points over naive data pooling, achieves 50\% success on unseen controller modes and 52\% on unseen controller gains, and transfers to tasks with limited controller-specific data. On contact-rich tasks such as USB insertion and surface wiping, behavioral interaction context ena
Open source S4
data across both embodiments and controllers without allowing conflicting action interpretations to interfere with learning. Our setting spans six controller types as well as variations in controller gains and compliance. We propose a framework that encodes behavioral interaction under a controller through examples of commanded actions and the resulting robot motions. A transformer context encoder summarizes these in
Open source S3
arXiv:2610.39280v2 Announce Type: replace Abstract: Generalist robot policies increasingly learn from data collected across diverse tasks and embodiments, yet typically assume a single controller for each robot. This assumption limits transfer because different controllers can assign different physical meanings to the same action and may require different strategies to complete a task. We study how to transfer robot
Open source S2
bles the policy to adapt not only its actions but also its strategy. Our results provide a path toward generalist robot policies that can leverage data collected across diverse action interpretations. For more information, visit \href{https://controller-transfer.github.io}{https://controller-transfer.github.io}.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org