From Language Priors to Field Adaptation: Preference Learning for Traversability Estimation

2026-10-02

AI Summary

This study addresses the poor cross-platform generalization of robot traversability estimation models and the high annotation costs associated with target-domain adaptation. To this end, we propose a sample-efficient preference learning framework that models traversability using von Mises–Fisher mixture prototypes within the frozen feature space of a vision-language model. By incorporating common-sense priors through natural language rules and leveraging sparse image preferences, the method achieves efficient fine-tuning. Experiments on the WayFAST dataset demonstrate that the proposed framework attains accuracy comparable to end-to-end models while significantly reducing target-domain annotation requirements. Furthermore, the results validate improvements in zero-shot generalization capability and dense prediction performance, while also yielding semantically interpretable prototypes.

Abstract

Image-based traversability estimation is inherently dependent on the robot platform, deployment domain, and mission preferences, which limits the applicability of purpose-trained models. To facilitate domain adaptation, this work aims to reduce the number of required annotations in the target domain using sample-efficient preference learning. Our method represents traversability through von Mises-Fisher mixture prototypes in a frozen vision-language feature space. Relative natural-language rules provide a commonsense prior, while sparse relative image annotations adapt the prototype directions and utilities to a target domain through computationally and sample-efficient fine-tuning. Experiments on WayFAST demonstrate accuracy competitive with end-to-end trained estimators while enabling sample-efficient image-based adaptation. Qualitative experiments further demonstrate the language prior's zero shot applicability and the fine-tuned estimator's improved dense prediction on semantic maps. Evaluation is complemented via semantic interpretation of learned prototypes by dissecting semantically close natural language prompts. Code and trained estimators available at https://resireg.github.io

Problem

• traversability estimation
• domain adaptation
• preference learning
• vision-language models
• sample efficiency

Innovation

• Traversability Estimation
• Preference Learning
• Vision-Language Models
• Domain Adaptation
• von Mises-Fisher Mixture Prototypes