aiqbench analysis
The paper addresses a real limitation of Cartesian end-effector policies: they ignore the robot's full body, so they can collide with the arm or other links. Its training-free method applies whole-body, joint-space collision guidance during diffusion sampling, without retraining the policy. The reported gains are large but come from specific constrained scenarios and simulated benchmarks, so readers should look at the evaluation setup before generalizing.
EmbodiSteer: Steering Embodiment-Agnostic Visuomotor Policies with Joint-Space Guidance for Zero-Shot Cross-Embodiment Deployment
arXiv:2606.12965v2 Announce Type: replace Abstract: Scalable robot imitation learning relies on large-scale heterogeneous data from diverse robots or body-free data, making Cartesian end-effector actions a key interface for embodiment-agnostic policy learning. However, end-effector-only abstraction leaves Cartesian policies unaware of the deployed robot body, making them brittle under robot-specific constraints such as whole-body collision avoidance. To overcome this limitation, we present EmbodiSteer, a training-free framework that steers embodiment-agnostic visuomotor policies toward zero-shot, embodiment-aware deployment. EmbodiSteer keeps policy learning in Cartesian space while efficiently lifting inference-time diffusion sampling into the target robot's joint space via forward kinematics and Jacobian-based updates. With whole-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 improvements are 90.0 and 36.7 percentage points. Our project page is at https://frankwang67.github.io/EmbodiSteer-Page.
Abstract:Scalable robot imitation learning relies on large-scale heterogeneous data from diverse robots or body-free data, making Cartesian end-effector actions a key interface for embodiment-agnostic policy learning. However, end-effector-only abstraction leaves Cartesian policies unaware of the deployed robot body, making them brittle under robot-specific constraints such as whole-body collision avoidance. To overcome this limitation, we present EmbodiSteer, a training-free framework that steers embodiment-agnostic visuomotor policies toward zero-shot, embodiment-aware deployment. EmbodiSteer keeps policy learning in Cartesian space while efficiently lifting inference-time diffusion sampling into the target robot's joint space via forward kinematics and Jacobian-based updates. With whole-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 improvements are 90.0 and 36.7 percentage points. Our project page is at this https URL.
Robotics (cs.RO)
arXiv:2606.12965 [cs.RO]
(or arXiv:2606.12965v2 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2606.12965 Focus to learn more arXiv-issued DOI via DataCite
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 · 256 words
Related coverage
EmbodiSteer: Steering Embodiment-Agnostic Visuomotor Policies with Joint-Space Guidance for Zero-Shot Cross-Embodiment Deployment
arXiv · Jun 12, 2026
Latent Policy Steering: An Efficient and Flexible Framework for Cross-Embodiment Transfer
arXiv · Sep 22, 2026
Building a Foundation Stack for General-Purpose Robots
IEEE Spectrum · Jul 13, 2026
Game-Theoretic Disempowerment: Loss of Control to Agencyless AI
LessWrong · Sep 27, 2026