Sandris Dubovs V L Nav Neka May 2026

Uses a CVL (Curiosity-driven Vision-Language) score to prioritize exploring unknown areas that align with human descriptions.

Leverages a 3D scene graph and image memory to help Vision Language Models (VLMs) replan tasks in real-time.

Proven to navigate successfully across different floors and transitions (e.g., using elevators or stairs) in complex building layouts. 3. Performance Summary (Good for Validation) Sandris Dubovs V L Nav Neka

You can find the full technical details on arXiv: VL-Nav .

View demonstrations on robots like the Unitree G1 and Go2 at the SAIR Lab Project Page . "In rigorous testing, including the , VL-Nav achieved

"In rigorous testing, including the , VL-Nav achieved a 75–83% success rate across indoor and outdoor settings. In real-world deployments, it maintained an 86.3% success rate , demonstrating reliability over long-range trajectories of up to 483 meters." Resources for Further Development

For related open-source frameworks, check repositories like oobvlm on GitHub. "In rigorous testing

is an advanced robotic navigation framework that combines neural reasoning (the "brain") with symbolic guidance (the "logic") to help robots navigate complex environments. Unlike traditional methods that might lead to aimless wandering, VL-Nav uses a NeSy (Neuro-Symbolic) Task Planner and an Exploration System to understand abstract human instructions. Useful Text Blocks 1. The "Problem & Solution" Pitch (Good for Intros)