GPC Opensources Flow-Matching Robot Policies Trained via Sampling-Based Predictive Control
rsasaki0109 · x · 2026-09-05
Vince Kurtz and Joel Burdick released the code for 'Generative Predictive Control: Flow Matching Policies for Dynamic and Difficult-to-Demonstrate Tasks' (79 GitHub stars). GPC targets robot tasks that are hard to demonstrate but easy to simulate: it alternates between generating training data with sampling-based predictive control, fitting a generative flow-matching model to that data, and using the model to improve the sampling distribution. The repo includes conda setup and training/testing examples across multiple robot systems, offering a reproducible alternative to expensive teleoperated demonstrations.
More from Embodied
- MolmoACT2 policy runs on a YAM robot for the first time, impressively smooth — DJiafei · 2026-09-06
- Neuralink recipient Audrey now plays Mario Kart entirely with her mind — XFreeze · 2026-09-06
- Dev couldn't resist: Microduck desktop robot gets ROS 2 support via hand-built URDF, now on GitHub — kamathsblog · 2026-09-06
- Solo dev builds autonomous Android agent with 30 tools, on-device Gemma 3 1B, and a hard policy gate — Adventurous-Win6029 · 2026-09-06
- AI glasses cheating spreads worldwide as expert warns of 'checkmate' for traditional exams — _akpiper · 2026-09-05
- Home Robot Companion Wars: Cute Bots from France, China and Japan Compared — alexcovo_eth · 2026-09-05