FreqNav: Stage-Wise Frequency Routing for Object-Oriented Aerial Vision-Language Navigation

2026-08-01

AI Summary

This work addresses the challenges of background interference and target ambiguity in long-horizon closed-loop aerial vision-language navigation caused by fixed visual representations. To this end, we propose a lightweight frequency-routing adaptive perception framework that models navigation as a progressive shift in frequency-domain preferences—from global spatial structures to fine-grained local details. The framework employs a stage-aware frequency token router, integrating dual-view frequency selection with stage-dependent visual anchoring to dynamically allocate perceptual resources under a fixed computational budget. By incorporating a diffusion Transformer architecture with explicit supervision, our method enables efficient and continuous trajectory prediction. Experimental results demonstrate that the proposed approach achieves approximately 3× faster inference while maintaining high accuracy, with real-world robotic validation confirming its efficacy and practicality in extended autonomous aerial missions.

Abstract

Object-oriented aerial vision-and-language navigation (VLN) requires searching for a described target and landing on it precisely, under long-horizon and closed-loop control. Guided by a target-descriptive instruction during navigation, perceptual priorities dynamically evolve: early-stage exploration prioritizes low-frequency spatial layout, and then shifts to high-frequency target details. Existing VLN methods model the varying perceptual requirements across navigation stages with identical visual tokens, leading to interference from irrelevant objects and background clutter. To this end, we therefore formulate long-horizon aerial navigation as a frequencypreference shift from spatial structure to local detail and propose FreqNav, a lightweight frequency-routing adaptive perception framework. Under a fixed computational budget, FreqNav dynamically reallocates visual tokens across frequency components according to the current navigation stage. A Frequency Token Router selects stage-relevant visual representations from dual-view observations, while a Phase-dependent Grounding Module anchors visual evidence through explicit supervision. A Diffusion Transformer then predicts smooth trajectories for continuous control. Experiments show that FreqNav outperforms strong baselines while achieving approximately 3x faster inference. Real-world deployment further demonstrates its effectiveness, efficiency, and practical potential for long-horizon aerial autonomy.

Problem

• vision-language navigation
• aerial navigation
• frequency perception
• long-horizon control
• object-oriented navigation

Innovation

• frequency routing
• adaptive perception
• vision-language navigation
• diffusion transformer
• aerial autonomy