aiqbench analysis
JAMB introduces a novel diffusion policy that jointly models bimanual actions and future 3D point tracks within a shared Transformer, explicitly accounting for how one arm's movement alters the shared environment. This geometry-aware approach significantly outperforms existing baselines in both simulation and real-world robotic tasks by refining action hypotheses through mutual information during the denoising process. The framework demonstrates robust generalization to cluttered scenes, addressing a critical gap in current multi-arm manipulation systems that often ignore dynamic scene interactions.
JAMB: Joint Action-Motion Diffusion for Bimanual Manipulation
arXiv:2609.25322v1 Announce Type: new Abstract: Coordinated bimanual manipulation is challenging because the motion of either arm can alter the shared 3D scene and thereby affect the other arm. Yet most diffusion policies generate actions without explicitly modeling these future geometric consequences, while predictive variants typically use future state only as auxiliary supervision or fixed conditioning. We address this limitation by proposing JAMB, a diffusion policy that jointly denoises bimanual actions and future 3D point tracks. By allowing action and track hypotheses to evolve together within a shared Transformer, each can inform and refine the other throughout denoising. We further ground multimodal representations in a shared spatiotemporal coordinate system to facilitate geometry-aware interaction during joint denoising. We evaluate JAMB on diverse bimanual manipulation tasks in RoboTwin 2.0 and on a real-world robot, comparing it with action-only policies and alternative future-prediction approaches spanning different state representations and learning objectives. Across 16 simulation tasks, JAMB achieves an average success rate of 83.4%, outperforming the strongest baseline by 23.9 percentage points. On three real-world tasks, it outperforms the action-only and auxiliary geometry prediction methods by 50.0 and 21.2 percentage points, respectively. Beyond these performance gains, JAMB shows stronger generalization to cluttered scenes and out-of-distribution backgrounds than the evaluated baselines. Together, these results demonstrate the effectiveness of our joint action-motion modeling framework for coordinated bimanual manipulation. Our project website is available at https://jam-bimanual.github.io/
Abstract:Coordinated bimanual manipulation is challenging because the motion of either arm can alter the shared 3D scene and thereby affect the other arm. Yet most diffusion policies generate actions without explicitly modeling these future geometric consequences, while predictive variants typically use future state only as auxiliary supervision or fixed conditioning. We address this limitation by proposing JAMB, a diffusion policy that jointly denoises bimanual actions and future 3D point tracks. By allowing action and track hypotheses to evolve together within a shared Transformer, each can inform and refine the other throughout denoising. We further ground multimodal representations in a shared spatiotemporal coordinate system to facilitate geometry-aware interaction during joint denoising. We evaluate JAMB on diverse bimanual manipulation tasks in RoboTwin 2.0 and on a real-world robot, comparing it with action-only policies and alternative future-prediction approaches spanning different state representations and learning objectives. Across 16 simulation tasks, JAMB achieves an average success rate of 83.4%, outperforming the strongest baseline by 23.9 percentage points. On three real-world tasks, it outperforms the action-only and auxiliary geometry prediction methods by 50.0 and 21.2 percentage points, respectively. Beyond these performance gains, JAMB shows stronger generalization to cluttered scenes and out-of-distribution backgrounds than the evaluated baselines. Together, these results demonstrate the effectiveness of our joint action-motion modeling framework for coordinated bimanual manipulation. Our project website is available at this https URL
Robotics (cs.RO)
arXiv:2609.25322 [cs.RO]
(or arXiv:2609.25322v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.25322 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 318 words
Related coverage
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
DexRoam: Learning Mobile Bimanual Dexterous Manipulation from Egocentric Whole-Body Human Demonstrations
arXiv · Sep 29, 2026
Beyond Dexterity: Why Contact May Define the Next Era of Robotics
IEEE Spectrum · Jun 9, 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · Sep 25, 2026