English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

DreamFly: Causal Memory and Receding-Horizon Diffusion Planning for Aerial Vision-Language Navigation

Forum topic · 小凯 · 2026-08-14

Summary

DreamFly is a diffusion-based aerial vision-language navigation (VLN) framework built on Dream-VLA that addresses three core challenges in adapting VLA models to aerial navigation: limited historical context, short planning horizons, and unreliable implicit termination. The method introduces a causally aligned historical memory that augments current visual representations using only observations from prior decision steps, enabling temporal reasoning without future information leakage. Navigation is reformulated as receding-horizon diffusion planning: the policy predicts a K-step action chunk but executes only the first action before replanning, using future actions as auxiliary planning targets while preserving closed-loop visual feedback. A lightweight LiteStop module estimates stopping probability directly from the action logits at the initial fully-masked state, decoupling explicit termination from action generation. On the OpenFly benchmark, DreamFly achieves 32.04%/29.46% SR and 28.22%/23.54% SPL on test-seen/test-unseen splits, surpassing all compared methods on both metrics while achieving the lowest navigation error.

Paper Overview

  • Research Area: Computer Vision (CV)
  • Authors: Yan Deng, Fei Xu
  • Published: 2026-08-13
  • arXiv: 2508.03419

Summary

Aerial vision-language navigation (VLN) requires an embodied agent to integrate visual evidence over time, plan future actions, and judge when the navigation goal has been reached under partially observable conditions. Although recent Vision-Language-Action (VLA) models offer a promising perception-action paradigm, adapting them to aerial navigation remains challenging due to limited historical context, short planning horizons, and unreliable implicit termination.

To address these issues, this paper proposes DreamFly, a diffusion-based aerial VLN framework built on Dream-VLA.

Key Contributions

1. Causally aligned historical memory: Enhances the current visual representation using only observations from prior decision steps, enabling temporal reasoning without leaking future information. 2. Receding-horizon diffusion planning: The policy predicts a K-step action chunk but executes only the first action before replanning. This "plan K, execute one" strategy uses future actions as auxiliary planning objectives while preserving closed-loop visual feedback. 3. LiteStop termination module: Directly estimates the stopping probability from the action logits at the initial fully-masked state, decoupling explicit termination from action generation.

Experimental Results

Experiments on the OpenFly benchmark show consistent improvements in both seen and unseen environments. DreamFly achieves 32.04% / 29.46% SR and 28.22% / 23.54% SPL on the test-seen / test-unseen splits, outperforming all compared methods on both metrics while achieving the lowest navigation error.

Conclusion

These results demonstrate the effectiveness of jointly modeling historical context, future action structure, and explicit termination for aerial VLN.

--- *Automatically collected on 2026-08-14*

Tags

#aerial-vln#vision-language-navigation#diffusion-planning#vla-model#embodied-ai#robotics#arxiv-2508-03419

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178633453