CATNAV: Cached Vision-Language Traversability for Efficient Zero-Shot Robot Navigation

TL;DR

CATNAV is a novel framework that uses multimodal large language models and a visuosemantic cache to enable efficient, embodiment-aware zero-shot robot navigation in unstructured environments, significantly reducing computational costs and improving safety.

Contribution

It introduces a cost-aware, zero-shot navigation method leveraging multimodal LLMs and a visuosemantic cache for risk assessment and path selection without task-specific training.

Findings

Achieves 10% higher goal-reaching rate

Reduces behavioral violations by 33%

Cuts online VLM queries by 85.7%

Abstract

Navigating unstructured environments requires assessing traversal risk relative to a robot's physical capabilities, a challenge that varies across embodiments. We present CATNAV, a cost-aware traversability navigation framework that leverages multimodal LLMs for zero-shot, embodiment-aware costmap generation without task-specific training. We introduce a visuosemantic caching mechanism that detects scene novelty and reuses prior risk assessments for semantically similar frames, reducing online VLM queries by 85.7%. Furthermore, we introduce a VLM-based trajectory selection module that evaluates proposals through visual reasoning to choose the safest path given behavioral constraints. We evaluate CATNAV on a quadruped robot across indoor and outdoor unstructured environments, comparing against state-of-the-art vision-language-action baselines. Across five navigation tasks, CATNAV achieves 10 percentage point higher average goal-reaching rate and 33% fewer behavioral constraint violations.