Does Peer Observation Help? Vision-Sharing Collaboration for Vision-Language Navigation

TL;DR

This paper introduces Co-VLN, a framework where multiple agents share visual observations during navigation, significantly improving their performance in vision-language tasks by expanding perceptual awareness without extra exploration.

Contribution

The paper presents a novel, model-agnostic framework for peer observation sharing in VLN, demonstrating its effectiveness across different paradigms and establishing a foundation for collaborative navigation research.

Findings

Vision-sharing improves VLN performance substantially.

Peer observation sharing expands agents' perceptual fields.

Framework is validated on R2R benchmark with two paradigms.

Abstract

Vision-Language Navigation (VLN) systems are fundamentally constrained by partial observability, as an agent can only accumulate knowledge from locations it has personally visited. As multiple robots increasingly coexist in shared environments, a natural question arises: can agents navigating the same space benefit from each other's observations? In this work, we introduce Co-VLN, a minimalist, model-agnostic framework for systematically investigating whether and how peer observations from concurrently navigating agents can benefit VLN. When independently navigating agents identify common traversed locations, they exchange structured perceptual memory, effectively expanding each agent's receptive field at no additional exploration cost. We validate our framework on the R2R benchmark under two representative paradigms (the learning-based DUET and the zero-shot MapGPT), and conduct extensive analytical experiments to systematically reveal the underlying dynamics of peer observation sharing in VLN. Results demonstrate that vision-sharing enabled model yields substantial performance improvements across both paradigms, establishing a strong foundation for future research in collaborative embodied navigation.