Pupper v3 · published recipe

velocity-tracking locomotion from joystick (default walking policy)

Record

Robot: Pupper v3 stanford/pupper-v3
Framework: MuJoCo MJX + Brax PPO (pupperv3-mjx), JSON weights run by the neural_controller ROS 2 node · MuJoCo Playground / MJX
Policy license: not stated by the package · robot software repos: AGPL-3.0 AND GPL-3.0 AND MIT
File: ros2_ws/src/neural_controller/launch/policy_latest.json
Source: https://github.com/Nate711/pupperv3-monorepo
Note: Git LFS file (~39 MB); config.yaml neural_controller.model_path points at it. Same folder also holds policy.json and two named W&B runs.

Not in the tab yet

To replay this policy in the browser we need an ONNX export of a trained checkpoint, the observation and action contract copied from the authors' own sim2sim or deploy script, and a MuJoCo model of the robot that the contract matches. Then it must walk 60 s in our runtime without a fall. Until then the source link is the way in.

Other Pupper v3 policies · 1

three-legged walkingMuJoCo MJX + Brax PPO (pupperv3-mjx)