RoboAuto: Robot Program Search With Multi-Level Physical Understanding
DJiafei · x · 2026-11-01
A new paper tackles the bottleneck of LLM-based robot self-improvement: existing Code-as-Policy methods rely mostly on execution feedback without truly understanding the physical world. RoboAuto organizes physical understanding at three levels via hierarchical knowledge bases (affordances, spatial relations, action effects), scene grounding, and multi-stage planning that runs self-designed experiments to verify physical effects and test competing hypotheses, reporting superior performance on manipulation tasks.
More from Embodied
- mink 1.4.0 released with new Elastic Constraints feature for robot control — kevin_zakka · 2026-11-01
- Naval predicts 4-passenger drones will eventually end commercial air travel, with a long interregnum ahead — naval · 2026-11-01
- Meta Quest game scans your real room and turns it into a haunted house — Scobleizer · 2026-11-01
- Developer dresses robot 'Bob' as a ghost to hand out Halloween candy — chrismatthieu · 2026-11-01
- GRASP authors compare with SuperMimic: extracting more optimization signal from sim experience — hankyang94 · 2026-11-01
- Harvard's GRASP refines human demos in physics sim for dexterous hands and Unitree G1 — hankyang94 · 2026-11-01