MessyNav teaches robots to push obstacles aside for zero-shot navigation

MessyNav, accepted at CoRL 2026 by Stanford and UC Berkeley researchers, is a VLM-based framework enabling zero-shot navigation: when no collision-free path exists, robots reason about which objects are safe to move and push obstacles aside.

2026-11-10 ~ 2026-11-10 · 2 related posts

• MessyNav at CoRL 2026: robots clear clutter out of the way with zero-shot VLM navigation — leto__jean · 2026-11-10
• MessyNav at CoRL 2026: zero-shot robot navigation by safely moving obstacles aside — leto__jean · 2026-11-10