Shared Modular Recurrence in Contextual MDPs for Universal Morphology Control
Shared Modular Recurrence in Contextual MDPs for Universal Morphology Control
This paper introduces a universal controller architecture leveraging shared modular recurrence within contextual Markov decision processes to enhance data and computational efficiency. The authors address the challenge of controlling robots with highly dissimilar morphologies, particularly when generalizing to unseen agents where contextual features are partially available. By recovering these features through modular interactions, the proposed transformer-based approach demonstrates substantial improvements in zero-shot generalization performance across four different environments using a large set of MuJoCo robots with varying dynamics, kinematics, and topologies.
The authors propose a transformer-based architecture with shared modular recurrence to improve zero-shot generalization of controllers across robots with unseen dynamics, kinematics, and topologies, evaluated on a large set of MuJoCo environments.
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.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
, we implement a transformer-based architecture with shared modular recurrence and evaluate its (generalization) performance on a large set of MuJoCo robots. The results show a substantial improvement in zero-shot generalization performance on robots with unseen dynamics, kinematics, and topologies, in four different environments.
Open source S4
arXiv:2506.08630v4 Announce Type: replace-cross Abstract: A universal controller for any robot morphology would greatly improve computational and data efficiency. Steps have been made towards such multi-robot control by utilizing contextual information about the properties of individual robots and exploiting their modular structure in the architecture of deep reinforcement learning agents. When the robots have highly
Open source S2
dissimilar morphologies, however, this becomes a challenging problem, especially when the agent must generalize to new, unseen robots. In this paper, we consider the fact that contextual features are often only partially available, but that they can be recovered through modular interactions. This can allow for better multi-robot control and generalization to contexts that are not seen during training. To this extent
Open source S3
Shared Modular Recurrence in Contextual MDPs for Universal Morphology Control
Open source S1
Source：arXiv Robotics — research abstracts · arxiv.org