Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

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

Source

arXiv

arxiv.org/abs/2607.21400v1

Read the source

Source

Plain-Language Brief

The researchers propose a new benchmark for vision-based navigation called Vision-Only Long-Horizon Navigation (VoLN). This benchmark is designed for aerial navigation and requires agents to detect, interpret, and select online route-relevant information from locally observable in-scene cues. The authors introduce VoLN-UAV, a 7,210-episode benchmark, and provide an initial reference baseline called VoLN-MLLM.

What the Document Says

  • The paper introduces Vision-Only Long-Horizon Navigation (VoLN), a new benchmark for vision-based navigation.
  • VoLN shifts route-relevant information from externally supplied instructions and global guidance to locally observable in-scene cues.
  • Goal views specify the destination in VoLN.
  • The authors present VoLN-UAV, a 7,210-episode benchmark for aerial navigation.
  • VoLN-MLLM is an initial reference baseline provided by the authors.

Key Concepts Mentioned

  • Vision-Only Long-Horizon Navigation (VoLN)
  • Vision-and-Language Navigation (VLN)
  • aerial navigation
  • goal views
  • VoLN-UAV
  • VoLN-MLLM

Claims Made by the Document

  • VoLN provides an initial evaluation of Vision-Only Long-Horizon Navigation.
  • The success rates of VoLN-MLLM on the Test-Unseen split reveal substantial remaining challenges in long-horizon evidence integration, cross-view goal matching, and closed-loop stability.

Evidence Used

  • 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.
  • Project page: https://admire-ljb.github.io/VoLN-UAV/

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

  • The authors present initial results and identify substantial remaining challenges in long-horizon evidence integration, cross-view goal matching, and closed-loop stability.

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • It is unknown if the authors have addressed these identified limitations in subsequent work or improvements.
  • Further research may be necessary to fully understand the challenges mentioned by the authors and potential solutions.

Open Questions

  • What are the specific challenges in long-horizon evidence integration, cross-view goal matching, and closed-loop stability mentioned by the authors?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The paper focuses on the evaluation of a new benchmark for vision-based navigation. The success rates of the initial reference baseline are presented as an evaluation but do not provide detailed insights into the challenges.
  • The success rates provided in the text might vary from actual performance due to factors like testing conditions, evaluation metrics, and other implementation details.

Raw Abstract / Excerpt

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/