VoLN: Vision-Only Long-Horizon Navigation---Paradigm, Benchmark, and Method

2026-07-23

AI Summary

This work addresses the limitations of existing vision-and-language navigation, which relies on explicit path priors in task instructions and thus fails to evaluate pure visual long-horizon navigation capabilities. We propose Vision-Only Long-Horizon Navigation (VoLN), a new paradigm that specifies the destination solely via a target view and requires agents to make online decisions based exclusively on local visual cues within the environment. To support this setting, we introduce VoLN-UAV, the first benchmark for aerial robots in pure visual long-horizon navigation, along with VoLN-MLLM, a baseline method leveraging multimodal large language models. VoLN-MLLM integrates self-supervised visual features, structured semantic alignment, historical observations, and egomotion to predict short-horizon waypoint sequences. Evaluated on a Test-Unseen split across five environments, the model achieves success rates of 7.4%, 4.5%, and 1.8% under Easy, Normal, and Hard difficulty levels, respectively, highlighting core challenges in long-term evidence integration, cross-view target matching, and loop-closure stability.

Abstract

Vision-and-Language Navigation (VLN) enables embodied agents to follow natural-language instructions. However, route-level instructions commonly encode spatial priors, such as orientation, distance, and layout, that are not explicitly available from onboard sensing at deployment in open, GPS-denied environments. Benchmark performance under such interfaces therefore jointly reflects visual navigation ability and the use of route structure explicitly supplied by the task description. As a complementary formulation, we propose Vision-Only Long-Horizon Navigation (VoLN), which shifts route-relevant information from externally supplied instructions and global guidance to locally observable in-scene cues. In VoLN, goal views specify the destination, while route-relevant information is available only through locally observable in-scene cues that the agent must detect, interpret, and select online. We instantiate VoLN for aerial navigation through VoLN-UAV, a 7,210-episode benchmark that combines long-horizon goal-directed flight, continuous 3D motion, large viewpoint changes, and context-dependent beacon selection. We further provide VoLN-MLLM as an initial reference baseline. It aligns self-supervised visual features with a structured semantic space and predicts short-horizon waypoint segments from observation history, goal views, retrieved visual--semantic tokens, and proprioception. On the five-environment Test-Unseen split, it obtains success rates of 7.4%, 4.5%, and 1.8% on Easy, Normal, and Hard episodes, respectively. These results provide an initial evaluation of VoLN and reveal substantial remaining challenges in long-horizon evidence integration, cross-view goal matching, and closed-loop stability. Project page: https://admire-ljb.github.io/VoLN-UAV/

Problem

• Vision-Only Navigation
• Long-Horizon Navigation
• Embodied AI
• In-Scene Cues
• Goal-Directed Navigation

Innovation

• Vision-Only Navigation
• Long-Horizon Navigation
• In-Scene Cues
• Self-Supervised Visual Features
• Aerial Embodied AI