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

2026-03-24

AI Summary

This work addresses the challenge of efficiently evaluating path risk according to a robot’s physical constraints during zero-shot navigation in unstructured environments. The authors propose an embodied navigation framework that requires no task-specific training, leveraging a vision-language model (VLM) to generate an embodiment-aware cost map. To minimize online VLM queries, they introduce a visual-semantic caching mechanism that reuses historical risk assessments. Furthermore, a VLM-driven trajectory selection module is designed to ensure safe path planning under behavioral constraints. Experiments on a quadrupedal robot demonstrate that the proposed approach improves target arrival success by 10 percentage points, reduces behavioral violations by 33%, and decreases online VLM invocations by 85.7%.

Technology Category

• Intelligent Robots: Embodied AI
• Natural Language Processing: Safety and Robustness
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

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.

Problem

• robot navigation
• traversability assessment
• zero-shot learning
• embodiment-aware
• unstructured environments

Innovation

• zero-shot navigation
• visuosemantic caching
• embodiment-aware traversability
• vision-language models
• costmap generation