Simulated hand with 39 muscles masters 9 poses after 426M RL training steps
MaziyarPanahi · x · 2026-11-11
A demo of stroke-rehab-inspired reinforcement learning: a simulated hand with 39 muscles and 23 joints learns hand control one movement at a time, mimicking how patients relearn after a stroke.
- Untrained brain reached 0 of 9 target poses; after 426M training steps, 9 of 9
- The number under each hand is the radian error of its 23 joints vs the target: 3.4–4.2 rad untrained, converging to 0.63–0.70 rad, just under the 0.7 pass line
- The author notes it's a simulation, not a real patient
More from Embodied
- NVIDIA open-sources GR00T 2 robot model running real dexterous chem lab demos — EricBuess · 2026-11-11
- Russ Tedrake's CoRL 2026 keynote is recommended viewing for robotics founders — adnothing · 2026-11-11
- Homebrew robot arm build log Day 7: geartrain CAD, wiring and gripper sizing — _Stocko_ · 2026-11-11
- Perceptron AI launches EgoHome, an open benchmark for sub-second egocentric hand understanding — rohanpaul_ai · 2026-11-11
- Runway open-sources Praxis-1, turning video pretraining into real robot control — runwayml · 2026-11-11
- Unitree G1 humanoid wearing VR goggles pilots an A1 drone in land-air robot demo — CyberRobooo · 2026-11-11