PoseDiff: A Unified Diffusion Model Bridging Robot Pose Estimation and Video-to-Action Control
It addresses the problem of integrating perception and control in embodied AI, offering a scalable and efficient solution, though it appears incremental by building on existing diffusion models and world models.
The paper tackles robot pose estimation and video-to-action control by introducing PoseDiff, a unified diffusion model that maps visual observations to robot states and generates action sequences, achieving state-of-the-art accuracy on the DREAM dataset and improved success rates on Libero-Object tasks.
We present PoseDiff, a conditional diffusion model that unifies robot state estimation and control within a single framework. At its core, PoseDiff maps raw visual observations into structured robot states-such as 3D keypoints or joint angles-from a single RGB image, eliminating the need for multi-stage pipelines or auxiliary modalities. Building upon this foundation, PoseDiff extends naturally to video-to-action inverse dynamics: by conditioning on sparse video keyframes generated by world models, it produces smooth and continuous long-horizon action sequences through an overlap-averaging strategy. This unified design enables scalable and efficient integration of perception and control. On the DREAM dataset, PoseDiff achieves state-of-the-art accuracy and real-time performance for pose estimation. On Libero-Object manipulation tasks, it substantially improves success rates over existing inverse dynamics modules, even under strict offline settings. Together, these results show that PoseDiff provides a scalable, accurate, and efficient bridge between perception, planning, and control in embodied AI. The video visualization results can be found on the project page: https://haozhuo-zhang.github.io/PoseDiff-project-page/.