Robot Diffusion Policy Learns from Low-Quality Data
giannis_daras · x · 2026-07-13
The author will present Ambient Diffusion Policy in two spotlight talks at RSS 2026 workshops.
According to the cited abstract, this method tackles the common issue of "low-quality demonstration data" in robotics. Instead of aggressively filtering data or co-training good and bad data together, it selectively learns useful features through a noise-varying data usage strategy.
The author positions this as a simple, principled approach for real-world scenarios filled with suboptimal data:
- No data filtering, preventing waste
- No simple co-training, avoiding learning bad features alongside good ones
- Using a diffusion/noise mechanism to help the model focus on effective information during training
The post focuses on the core concepts of this robotic learning method rather than the event itself.
More from Embodied
- Tesla expands Robotaxi rides to seven areas, including new Orlando and Tampa zones — elonmusk · 2026-07-22
- Hands-on robotics workshop on Saturday may be the last in-person session before August — StewartalsopIII · 2026-07-22
- NVIDIA pitches World Foundation Models as a way to scale physical AI data generation — MonaJalal_ · 2026-07-22
- RoboMME Podcast Preview: Benchmarking Memory for Robotic Policies — chris_j_paxton · 2026-07-21
- Gritt says an 8-person crew now installs 3,000 to 4,000 solar panels a day — HaktanSuren · 2026-07-21
- A helium-powered flying robot whale aims to be a quiet companion pet — chris_j_paxton · 2026-07-21