OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulation
OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulation
This research presents OASIS, a visuomotor policy that aligns intermediate representations with the action space through SE(3) end-effector trajectory prediction. By coupling a 3D-aware feature encoder with an SE(3) trajectory predictor, OASIS generates action chunks consistent with rigid-body motion. Experiments show it outperforms VLA and WAM baselines in success rate and out-of-distribution generalization across both simulation and real-world settings.
The paper introduces OASIS, a visuomotor policy that aligns intermediate representations with the action space using SE(3) end-effector trajectory prediction, demonstrating improved success rates and out-of-distribution generalization over VLA and WAM baselines in both simulation and real-world experiments.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org