Does Peer Observation Help? Vision-Sharing Collaboration for Vision-Language Navigation
For VLN agents operating in shared environments, this work shows that simple peer observation sharing can significantly boost performance, addressing the partial observability bottleneck.
Co-VLN introduces a framework where multiple VLN agents share visual observations from common locations, improving navigation performance on the R2R benchmark by expanding each agent's receptive field without extra exploration. The method yields substantial gains in both learning-based (DUET) and zero-shot (MapGPT) paradigms.
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.