Karma: an open-source robot control stack for VR teleop and policy inference
k7agar · x · 2026-09-23
Developers open-sourced Karma (SRA-VJTI/karma), the full control stack behind their cloth-folding bimanual robot: Meta Quest VR teleoperation with demonstration recording, single/dual-arm YAM and SO100/SO101 control, MolmoAct2 policy inference, human-in-the-loop DAgger data collection, and LeRobot v3 dataset recording, built on Physical Intelligence's openpi-control interface. Everything runs via a single uv run karma executable.
More from Embodied
- New paper: Transferring the Intelligence of VLMs to Robotic Control — _akhaliq · 2026-09-23
- CoRL 2026 Workshop Will Probe Why Controllers and Hardware Beat Algorithm Tweaks in Robot Learning — pulkitology · 2026-09-23
- Cognex Acquires RealSense for $500M, 439 Days After Intel Spinout — lukas_m_ziegler · 2026-09-23
- Wayve-Powered Mercedes Impresses AV Veterans: 'Much Better Than My FSD' — Sethwinterroth · 2026-09-23
- LEAP@CoRL late submissions open, Dieter Fox joins program committee — tomssilver · 2026-09-23
- rei_labs unveils Adapt-1 Machina: RL learns continuous control sequences without demos or critic — burny_tech · 2026-09-23