Towards Zero-Shot Terrain Traversability Estimation: Challenges and Opportunities

TL;DR

This paper investigates the potential of vision-language models for zero-shot terrain traversability estimation in robotics, highlighting challenges, introducing a new dataset, and proposing a simple integration pipeline.

Contribution

It introduces a human-annotated water traversability dataset and a pipeline integrating VLMs for zero-shot estimation, providing insights into current model limitations.

Findings

Estimations are subjective but show some human consensus.

Current foundation models are not yet practical for deployment.

The proposed pipeline offers a baseline for future research.

Abstract

Terrain traversability estimation is crucial for autonomous robots, especially in unstructured environments where visual cues and reasoning play a key role. While vision-language models (VLMs) offer potential for zero-shot estimation, the problem remains inherently ill-posed. To explore this, we introduce a small dataset of human-annotated water traversability ratings, revealing that while estimations are subjective, human raters still show some consensus. Additionally, we propose a simple pipeline that integrates VLMs for zero-shot traversability estimation. Our experiments reveal mixed results, suggesting that current foundation models are not yet suitable for practical deployment but provide valuable insights for further research.