{"version":1,"type":"story","url":"https://digestai.news/story/microduck-can-be-trained-with-isaac-lab-3-0-and-deployed-via-onnx","json":"https://digestai.news/story/microduck-can-be-trained-with-isaac-lab-3-0-and-deployed-via-onnx.json","markdown":"https://digestai.news/story/microduck-can-be-trained-with-isaac-lab-3-0-and-deployed-via-onnx.md","slug":"microduck-can-be-trained-with-isaac-lab-3-0-and-deployed-via-onnx","headline":"Microduck can be trained with Isaac Lab 3.0 and deployed via ONNX","summary":"Microduck, an affordable open‑source bipedal robot co‑released by Pollen Robotics and Hugging Face, can now be taught to walk using NVIDIA’s Isaac Lab 3.0. The tutorial shows how to run thousands of parallel GPU simulations, apply Newton‑based actuator modeling and domain randomization, and train a walking policy that observes 61 values and outputs 14 joint positions.\n\nThe guide walks users through installing Isaac Lab on a Linux workstation, launching the rough‑terrain training task with up to 1.25 M steps per second on 16 GPUs, and exporting the resulting policy as both TorchScript and ONNX files. The ONNX model (max 64 MiB) can be loaded onto the robot’s onboard runtime, tested in the Hugging Face‑hosted Microduck Arena, or shared on the Hub for community reuse. The article also lists the hardware requirements, command‑line steps, and how to handle servo backlash and other real‑world effects during simulation.\n\nBy combining open‑source hardware, a high‑performance simulation stack, and a straightforward deployment pipeline, developers can iterate quickly from virtual training to a physical robot that recovers from falls and walks over uneven terrain.","keyPoints":["Microduck is an affordable open‑source bipedal robot released by Pollen Robotics and Hugging Face.","Isaac Lab 3.0 can train walking policies at up to 1.25 M steps per second using 16 GPUs.","Trained policies are exported as ONNX (max 64 MiB) and loaded onto the robot’s onboard runtime."],"whyItMatters":"The workflow lowers the barrier for researchers and makers to develop real‑world robot locomotion using simulation‑to‑real transfer, accelerating open‑source robotics innovation.","category":{"slug":"robotics","name":"Robotics & Physical AI","url":"https://digestai.news/category/robotics"},"entities":{"companies":["Pollen Robotics","Hugging Face","NVIDIA"],"models":[],"people":["Antoine Pirrone","Augustin Crampette","Matthieu Lapeyre","Coralie Deplanne","Anne‑Charlotte Passanisi","Thomas Wolf"]},"firstPublishedAt":"2026-10-21T17:58:42Z","updatedAt":"2026-10-21T17:58:42Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"Hugging Face","title":"Train in Simulation, Deploy on a Real Robot: Microduck with Isaac Lab 3.0","url":"https://huggingface.co/blog/nvidia/microduck-isaac-lab-sim-to-real","publishedAt":"2026-10-21T17:58:42Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Microduck can be trained with Isaac Lab 3.0 and deployed via ONNX\", 21 October 2026, https://digestai.news/story/microduck-can-be-trained-with-isaac-lab-3-0-and-deployed-via-onnx","publisher":"Digest AI","title":"Microduck can be trained with Isaac Lab 3.0 and deployed via ONNX","datePublished":"2026-10-21T17:58:42Z","url":"https://digestai.news/story/microduck-can-be-trained-with-isaac-lab-3-0-and-deployed-via-onnx"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}