OpenVLN: Open-world aerial Vision-Language Navigation

2025-11-09

AI Summary

To address data scarcity and long-horizon trajectory planning challenges in vision-language navigation (VLN) for complex outdoor aerial environments, this paper proposes a data-efficient open-world framework. Our method leverages reconstruction-based reinforcement learning to guide fine-tuning of vision-language models (VLMs) with rule-based policies and introduces a value-function-driven dynamic long-distance trajectory synthesis mechanism, enabling end-to-end mapping from natural language instructions to robust flight policies. Evaluated on the TravelUAV benchmark under multi-reward settings, our approach achieves absolute improvements of 4.34%, 6.19%, and 4.07% in success rate, oracle success rate, and path-weighted success rate over baseline methods, respectively—demonstrating enhanced accuracy and generalization for long-range aerial navigation. The core contribution is a lightweight, rule-value-coordinated training paradigm that jointly optimizes data efficiency and long-horizon planning capability.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Language and Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Vision-language models (VLMs) have been widely-applied in ground-based vision-language navigation (VLN). However, the vast complexity of outdoor aerial environments compounds data acquisition challenges and imposes long-horizon trajectory planning requirements on Unmanned Aerial Vehicles (UAVs), introducing novel complexities for aerial VLN. To address these challenges, we propose a data-efficient Open-world aerial Vision-Language Navigation (i.e., OpenVLN) framework, which could execute language-guided flight with limited data constraints and enhance long-horizon trajectory planning capabilities in complex aerial environments. Specifically, we reconfigure a reinforcement learning framework to optimize the VLM for UAV navigation tasks, which can efficiently fine-tune VLM by using rule-based policies under limited training data. Concurrently, we introduce a long-horizon planner for trajectory synthesis that dynamically generates precise UAV actions via value-based rewards. To the end, we conduct sufficient navigation experiments on the TravelUAV benchmark with dataset scaling across diverse reward settings. Our method demonstrates consistent performance gains of up to 4.34% in Success Rate, 6.19% in Oracle Success Rate, and 4.07% in Success weighted by Path Length over baseline methods, validating its deployment efficacy for long-horizon UAV navigation in complex aerial environments.

Problem

• Addressing data scarcity in aerial vision-language navigation tasks
• Enhancing long-horizon trajectory planning for UAV navigation
• Optimizing VLMs for complex outdoor aerial environments

Innovation

• Reconfigure reinforcement learning for VLM optimization
• Fine-tune VLM using rule-based policies with limited data
• Introduce long-horizon planner for dynamic trajectory synthesis