Microduck Reveals Full Robot RL Pipeline
markjeffrey · x · 2026-08-29
Microduck demonstrates a full reinforcement learning pipeline where robots train in simulation using MuJoCo and PPO, export policies to ONNX, and deploy on hardware. It tackles Sim-to-Real by modeling actuator behavior, voltage, and gear backlash. Policies include walking, recovery, skating, and more.
Related event: Pollen Robotics open-sources Microduck bipedal robot RL training stack(5 posts)→
More from Embodied
- Musk teases next-gen Tesla Roadster unveil, hints at sub-1-second 0-60 mph — elonmusk · 2026-10-15
- Robotics papers on arXiv have grown ~6x in five years, peaking at 676 in a single week — Rewkang · 2026-10-15
- Bantam Tools launches HandWriter plotter inspired by legendary HP7475A — Merzmensch · 2026-10-15
- Amazon's Bee launches $74.99 Frontier wearable and $99.99 Eclipse smartphone recorder — The Verge AI · 2026-10-15
- Tesla plans to double Giga Berlin with 536,000 sqm hall; Cybercab and Optimus tipped for 2028 EU production — JOBhakdi · 2026-10-15
- China's AgiBot Reportedly Raising $210M at $7B Valuation, IPO May Slip to 2027 — xmercury_one · 2026-10-15