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

In this paper, the authors introduce DreamFly, a new framework for aerial vision-language navigation that addresses the challenges of limited historical context, short planning horizons, and unreliable implicit termination in previous models. DreamFly uses causally aligned historical memory to enable temporal reasoning without future information leakage, receding-horizon diffusion planning to plan actions and execute only the first action before replanning, and LiteStop to estimate stop probability directly from action logits. The experiments on OpenFly benchmark show that DreamFly outperforms other methods in both seen and unseen environments.

What the Document Says

• The paper presents a new framework called DreamFly for aerial vision-language navigation.
• DreamFly addresses the issues of limited historical context, short planning horizons, and unreliable implicit termination in previous models.
• It uses causally aligned historical memory for temporal reasoning.
• The paper introduces receding-horizon diffusion planning where only the first action is executed before replanning.
• LiteStop is used to estimate stop probability directly from action logits.
• Experiments on OpenFly benchmark demonstrate improvements in seen and unseen environments.
• DreamFly outperforms all compared methods on both SR (Success Rate) and SPL (Success-in-Place-Likelihood)

Key Concepts Mentioned

• aerial vision-language navigation
• DreamFly
• Dream-VLA
• causally aligned historical memory
• receding-horizon diffusion planning
• LiteStop
• Success Rate (SR)
• Success-in-Place-Likelihood (SPL)

Claims Made by the Document

• DreamFly addresses the challenges of limited historical context, short planning horizons, and unreliable implicit termination in previous models.

Evidence Used

• The paper discusses these issues and presents DreamFly as a solution.

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.