Microduck Open Sources Full RL Stack for Sim2Real Bipedal Robot
carlosdponx · x · 2026-08-28
Pollen Robotics released the reinforcement learning (RL) training environment for its Microduck bipedal robot. Built on MuJoCo Warp (mjlab) and PPO, the project provides a complete Sim2Real workflow:
- Tech Stack: Policies are trained at 50 Hz, exported to ONNX, and deployed on the real robot.
- Sim2Real Recipe: The repo encodes the full recipe, including BAM actuator physics, domain randomization, backlash simulation, and reward-design lessons distilled in AGENTS.md.
- Quickstart: Available on GitHub requiring CUDA GPU and uv, designed as the "Hello World" for legged robotics education.
More from Embodied
- Robots 'lose their lives' in a race gone wrong — TansuYegen · 2026-08-28
- Ambi releases CARGO physical AI model with 72.5% stacking density — jfiance · 2026-08-28
- Anthrobotics discusses artificial muscles and product building — IanPritchard · 2026-08-28
- Shield AI aims for 10x performance, Hivemind flies on 30+ platforms — BrettKrieger12 · 2026-08-28
- Avride (ex-Yandex self-driving): 600k deliveries & 100k autonomous rides — demian_ai · 2026-08-28
- Tesla Optimus supply chain scaled for 70k units, silence signals commitment — JOBhakdi · 2026-08-28