WARP: Training mobile manipulation policies directly from human data
danfei_xu · x · 2026-08-20
The WARP project aims to learn mobile manipulation policies directly from human data without teleoperation.
Key Breakthrough: It solves the problem of accurate whole-body kinematic retargeting. WARP converts offline human motion into replayable robot actions in closed form—precise, consistent, and faithful to whole-body intent.
Results: This enables training a wide range of mobile manipulation skills directly from human demonstrations, making direct human-to-robot policy transfer a reality.
Future Work: Improvements are needed in observation-space matching and controller optimization to achieve natural human motion and speed. This points to a broader shift: as fidelity of sensorized human data improves, human experience will become a primary source of post-training data for robot policies.
More from Embodied
- Team says no technical blockers remain for v1 self-replicating factory using off-the-shelf parts — ihorbeaver · 2026-10-16
- Author's own caveat: direct LLM robot control only suits repetitive, constrained tasks — ihorbeaver · 2026-10-16
- Skip the fancy vision: a $300 automatic screw feeder beats LLM-controlled picking — ihorbeaver · 2026-10-16
- Precise part insertion via LLM-written data collection code and a linear regression aligner — ihorbeaver · 2026-10-16
- Picking parts from a random pile with an extra flip-and-align platform — ihorbeaver · 2026-10-16
- Robot fully controlled by frontier LLMs assembles gadgets autonomously for 1.5 hours error-free — ihorbeaver · 2026-10-16