Free-standing 29-DoF Unitree G1 in MuJoCo (no fixed base, no balance assistance) returning forehand feeds. Upper body: scripted real-time controller (ball forecast with drag, Magnus and a fitted bounce; Ruckig-planned stroke). Legs: Unitree's pretrained locomotion network plus a balance residual we trained. Our evaluation on 50 unseen feeds: 44/50 legal first returns (25/25 standard, 19/25 wide; 5 of the 6 misses were feeds into the net). Simulation only, self-reported, not hardware-validated.

Alpslee·Unitree G1

Tiny robot. Big air. Booster K1 performs a simulated tuck jump and recovery in phyz. Full real-time run, followed by labeled slow-motion and freeze-frame replay. Editorial sound design; simulation only, not hardware-validated. should be for the other challenge, can't edit?

Cam Pedersen·Booster K1

Policy-driven fixed-base MuJoCo demo for a 29-DoF Unitree G1 on one qualified center-lane feed, followed by measured 1,000-feed simulation metrics. Simulation only; physical commissioning and the >=50 human-rally goal remain unverified.

Ethan A·Unitree G1

Fixed-base 29-DoF Unitree G1, MuJoCo, 50 Hz. Scripted model-based IK hitter + a per-serve 13->64->64->20 residual MLP (NumPy runtime, no torch), trained with PPO at a 0.10 m offset stance. Measured 1,000-serve deterministic evaluation (seed 2000, mean action): 78.6% legal returns, 47.0% target hits, 0.288 m mean landing error, 88.3% paddle contact. Video: 10 serves, same policy. Simulation only.

Akbar Sherzod Khamidov·Unitree G1