LogisticsVLN: Vision-Language Navigation For Low-Altitude Terminal Delivery Based on Agentic UAVs

2025-05-06

AI Summary

Existing vision-language navigation (VLN) methods for UAVs focus on coarse-grained, long-range targets and fail to meet the precision requirements of low-altitude last-mile delivery. Method: This paper formally defines and addresses the fine-grained aerial last-mile delivery VLN task. We propose a modular multimodal large model architecture that integrates a lightweight large language model (LLM) with a vision-language model (VLM), enabling joint natural language understanding, floor-level localization, fine-grained object detection, and autonomous action decision-making. We further construct CARLA-based Vision-Language Delivery (VLD), the first benchmark dataset tailored for aerial last-mile delivery. Contribution/Results: Extensive end-to-end evaluation and ablation studies on VLD demonstrate significant improvements in navigation accuracy and environmental robustness over baseline approaches.

Technology Category

• Computer Vision: Language and Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The growing demand for intelligent logistics, particularly fine-grained terminal delivery, underscores the need for autonomous UAV (Unmanned Aerial Vehicle)-based delivery systems. However, most existing last-mile delivery studies rely on ground robots, while current UAV-based Vision-Language Navigation (VLN) tasks primarily focus on coarse-grained, long-range goals, making them unsuitable for precise terminal delivery. To bridge this gap, we propose LogisticsVLN, a scalable aerial delivery system built on multimodal large language models (MLLMs) for autonomous terminal delivery. LogisticsVLN integrates lightweight Large Language Models (LLMs) and Visual-Language Models (VLMs) in a modular pipeline for request understanding, floor localization, object detection, and action-decision making. To support research and evaluation in this new setting, we construct the Vision-Language Delivery (VLD) dataset within the CARLA simulator. Experimental results on the VLD dataset showcase the feasibility of the LogisticsVLN system. In addition, we conduct subtask-level evaluations of each module of our system, offering valuable insights for improving the robustness and real-world deployment of foundation model-based vision-language delivery systems.

Problem

• Autonomous UAV delivery for fine-grained terminal logistics
• Bridging gap in Vision-Language Navigation for precise delivery
• Developing multimodal LLM system for low-altitude aerial tasks

Innovation

• Uses MLLMs for autonomous terminal delivery
• Integrates LLMs and VLMs in modular pipeline
• Constructs VLD dataset in CARLA simulator