Hugging Face and NVIDIA publish Microduck walking policy tutorial for Isaac Lab
Overview
Hugging Face and NVIDIA publish a tutorial for training and evaluating a walking policy for the Microduck bipedal robot in Isaac Lab and Newton.
The guide covers domain randomization and backlash modeling, then exporting the trained policy as policy.onnx for deployment on a physical robot.
Microduck is an affordable open-source biped released by Pollen Robotics and Hugging Face. The tutorial says Arena accepts models with 61 observation values, 14 action values, and a maximum size of 64 MiB.
Written by AI from the articles below · updated Oct 21, 2:17 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Hugging Face BlogOfficialHugging Face and NVIDIA publish a guide to training and deploying Microduck walking policies
AIHugging Face and NVIDIA publish a tutorial for training a walking policy for the open-source Microduck biped in Isaac Lab and Newton. The guide covers domain randomization and backlash modeling, exporting the policy as policy.onnx, and loading it onto a physical robot.
Heat Trend
Not enough continuous observations to show a trend yet.