Booster T1

Source

BoosterRobotics/booster_gym @ da396a06d6ee

Checkpoint

deploy/models/T1.ptbooster_gym (Isaac Gym) PPO, MLP 47-256-128-128-12, ELU

License

Apache-2.0

Replicates

play_mujoco.py + envs/T1.yaml

Model

Booster T1 12-DoF · BoosterRobotics/booster_gym/resources/T1/T1_locomotion.xml

Export

ONNX opset 17, weights unchanged. max |a_torch - a_onnx| < 3e-6 over 200 random inputs (onnxruntime 1.x CPU).

Timing

physics 2 ms · policy every 10 steps (50 Hz) · network runs before the step

Obs

47 floats: projected_gravity[3] (sensor orientation) · base_ang_vel[3] (sensor angular-velocity) · command[3] x[1, 1, 1] · gait_clock[2] f=1.5Hz · joint_pos_rel[12] · joint_vel[12] x0.1 · last_action[12]

Action

12 joint targets = action (clip ±1) x 1 + default · PD kp/kd 200/5, 50/1 · torque to motor ctrl

Joints

Left_Hip_Pitch, Left_Hip_Roll, Left_Hip_Yaw, Left_Knee_Pitch, Left_Ankle_Pitch, Left_Ankle_Roll, Right_Hip_Pitch, Right_Hip_Roll, Right_Hip_Yaw, Right_Knee_Pitch, Right_Ankle_Pitch, Right_Ankle_Roll

Commands

trained vx [-1, 1] vy [-1, 1] yaw [-1, 1] · starts at [0.5, 0, 0]

Tested

60 s at [0.5, 0, 0] in MuJoCo 3.14.0 WASM + onnxruntime-web 1.30 (wasm, 1 thread), Chromium, real time: no fall, mean speed along heading 0.44 m/s, base height never below 0.64 m (2026-09-25)

Notes

• Loop order follows play_mujoco.py: at every 10th step (from step 0) the policy runs on the current state, then PD, then mj_step, then the gait clock advances by dt * f.
• Gravity and angular velocity come from the MJCF's framequat 'orientation' and gyro 'angular-velocity' sensors on the trunk IMU site, read from sensordata exactly as the script does.
• PD torque is clipped to each motor's ctrlrange (MuJoCo clamps limited ctrl). Actions are clipped to +-1 before use and fed back as the last-action term.
• Gait frequency is the mean of the trained range [1.0, 2.0] Hz while any command is non-zero, and 0 (clock terms zeroed) when all commands are zero, as in play_mujoco.py.
• The sensors declare noise in the MJCF; MuJoCo does not apply sensor noise, so the replay is deterministic.
• Replay check: this runtime (MuJoCo WASM + onnxruntime-web in Node) and the reference script (MuJoCo 3.14 Python + TorchScript) give the same base trajectory to the centimetre over 30 s at [0.5, 0, 0] (x 10.09 m, y -7.19 m). At that command the policy curves to the right in the reference script as well; steer with yaw.

Contract file: packages/sim/policies/booster_t1_walk/contract.json · served at /policies/booster_t1_walk/contract.json

mjlab ships no stock task for Booster T1. The training page gives you the pinned model and the steps to write a robot config.