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

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Yan Deng, Fei Xu
• Venue: arXiv: cs.CV
• Published: 2026-08-12
• Collected: 2026-08-13
• URL: http://arxiv.org/abs/2608.12308v1
• PDF: https://arxiv.org/pdf/2608.12308v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

Researchers propose DreamFly, a framework for aerial vision-language navigation that addresses challenges like limited historical context, short planning horizons, and unreliable implicit termination. The framework uses causally aligned historical memory to augment current visual representation and receding-horizon diffusion planning for action prediction and execution. Experiments show improved performance on OpenFly benchmark compared to other methods.

What the Document Says

• DreamFly is a proposed framework for aerial vision-language navigation that integrates visual evidence over time, plans future actions, and determines when it has reached a goal under partial observability.
• It is based on Dream-VLA and introduces a causally aligned historical memory for temporal reasoning without future information leakage.
• Navigation is formulated as receding-horizon diffusion planning where the policy predicts a K-step action chunk but executes only the first action before replanning.
• It uses future actions as auxiliary planning targets and preserves closed-loop visual feedback.
• LiteStop estimates the stop probability directly from action logits at the initial all-mask state, decoupling explicit termination from action generation.
• Experiments on OpenFly benchmark show improved performance in both seen and unseen environments.

Key Concepts Mentioned

• aerial vision-language navigation
• DreamFly
• Dream-VLA
• historical memory
• receding-horizon diffusion planning
• action chunk
• auxiliary planning targets
• closed-loop visual feedback
• LiteStop

Claims Made by the Document

• DreamFly improves aerial vision-language navigation

Evidence Used

• Experiments on OpenFly benchmark show improved performance in both seen and unseen environments.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Aerial vision-language navigation (VLN) requires an embodied agent to integrate visual evidence over time, plan future actions, and determine when it has reached a navigation goal under partial observability. Although recent VLA models offer a promising perception-to-action paradigm, adapting them to aerial navigation remains challenging due to limited historical context, short planning horizons, and unreliable implicit termination. To address these challenges, we propose DreamFly, a diffusion-based aerial VLN framework built on Dream-VLA. DreamFly introduces a causally aligned historical memory that augments the current visual representation using only observations preceding the current decision step, enabling temporal reasoning without future information leakage. We further formulate navigation as receding-horizon diffusion planning, where 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 targets while preserving closed-loop visual feedback. Finally, LiteStop estimates the stop probability directly from action logits at the initial all-mask state, decoupling explicit termination from action generation. Experiments on the OpenFly benchmark demonstrate consistent improvements in seen and unseen environments. DreamFly achieves 32.04%/29.46% SR and 28.22%/23.54% SPL on the test-seen/test-unseen splits, respectively, outperforming all compared methods on both metrics while attaining the lowest navigation error. These results demonstrate the effectiveness of jointly modeling historical context, future action structure, and explicit termination for aerial VLN.