← Back to Portfolio
RoboticsAIJune 22, 2026· Chaotic Curiosity
Teaching a Humanoid to Move (Unitree G1 RL)
Reinforcement learning on a simulated Unitree G1 — walking, a cartwheel, a backflip and getting up from a fall — trained on our NVIDIA DGX Spark and written up as a 15-chapter guide.
mjlabMuJoCo-WarpReinforcement Learning (PPO)Motion ImitationUnitree G1 (simulated)NVIDIA DGX Spark
We trained a simulated Unitree G1 humanoid with reinforcement learning in mjlab (MuJoCo-Warp) on our NVIDIA DGX Spark. The walking policy learned velocity tracking at about 1 m/s, reaching a mean reward of 50.5 and lasting 995 of 1,000 steps per episode. Motion imitation produced a cartwheel and a backflip, and getting up from a fall is a custom task we built from scratch. We tried running too — it didn't work, and we kept it in as a lesson in reward hacking. The whole arc is published as a 15-chapter, zero-background guide. Simulation only; no real hardware.