Humanoid recovery paper trains balance-aware stand-up policy on Unitree H1
Evidence notes
- The paper embeds classical balance metrics such as capture point, center-of-mass state, and centroidal momentum into reinforcement-learning training.
- The recovery policy is trained on Unitree H1-2 in Isaac Lab and reports a 93.4% recovery rate across randomized initial poses.
Company context
Develops quadruped and humanoid robots, with strength in dynamic locomotion, vertically integrated hardware, and relatively low-cost commercial deployment. Unitree is one of the clearest examples of a legged robotics company moving from research visibility into real productisation and broader market distribution.