EmbodiSteer: Steering Embodiment-Agnostic Visuomotor Policies with Joint-Space Guidance for Zero-Shot Cross-Embodiment Deployment
EmbodiSteer: Steering Embodiment-Agnostic Visuomotor Policies with Joint-Space Guidance for Zero-Shot Cross-Embodiment Deployment
EmbodiSteer is a training-free framework that steers embodiment-agnostic visuomotor policies toward zero-shot, embodiment-aware deployment. It operates in Cartesian space during policy learning and lifts inference-time diffusion sampling into the target robot's joint space using forward kinematics and Jacobian-based updates. The method improves task success and reduces collisions across simulated and physical robots.
EmbodiSteer introduces a training-free framework that enables zero-shot cross-embodiment deployment by steering embodiment-agnostic visuomotor policies using joint-space guidance. It improves task success and reduces collisions in both simulation and physical robots.
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.
- Environment
- SimulationOpen source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ole-body collision-aware guidance over joint trajectories after each denoising step, the arm can be steered away from collisions while preserving learned end-effector behavior. Across 9 simulated robots, EmbodiSteer reduces collision rate by 46.1 percentage points and increases task success by 28.5 percentage points over Cartesian-only execution; on two physical robots in highly constrained scenarios, the respective
Open source S4
improvements are 90.0 and 36.7 percentage points. Our project page is at https://frankwang67.github.io/EmbodiSteer-Page.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org