When Large Language Models Meet UAVs: How Far Are We?

2025-09-16

AI Summary

A significant gap exists between academic research and industrial deployment in integrating large language models (LLMs) with unmanned aerial vehicles (UAVs): academia emphasizes theoretical modeling and task generalization, whereas industry demands robust flight control, real-time mission planning, and reliable human–UAV interaction. Method: Through a mixed-methods empirical analysis—encompassing 74 peer-reviewed papers, 56 open-source projects, and 52 valid practitioner surveys—we systematically construct a novel taxonomy of nine UAV-LLM task categories and a bottleneck analysis framework. Contribution/Results: Our study quantifies application distribution and technology readiness gaps across domains. We find that 40.4% of practitioners have initiated LLM integration efforts, yet safety certification, real-time performance constraints, and system-level integration remain critical barriers. We propose a co-evolutionary roadmap balancing theoretical rigor and engineering feasibility, establishing the first cross-disciplinary empirical benchmark and methodological foundation for LLM-augmented autonomous UAV systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: (Large) Language Models

Application Category

• 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
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

The integration of unmanned aerial vehicles (UAVs) and large language models (LLMs) has emerged as a research direction of growing interest, with the potential to address challenges in autonomous decision-making, human-UAV interaction, and real-time adaptability. However, existing studies have remained largely in preliminary exploration with a limited understanding of real-world practice, risking a misalignment between academic research and practical needs and hindering the translation of results. To examine and address these potential challenges, we conducted an empirical study of 74 selected papers and 56 public GitHub projects, identified nine task types for LLMs in UAV systems, and quantified their distribution. Our findings show that academic research emphasizes theoretical modeling and task optimization with dispersed attention across tasks. In contrast, industrial projects focus on flight control, task planning, and human-machine interaction, prioritizing operability and efficiency. To further capture industry perspectives, we distributed an online questionnaire. We obtained 52 valid responses: 40.4% of practitioners have attempted to apply LLMs to UAV tasks. We further identify factors that impede real-world integration, including technological maturity, performance, safety, cost, and other considerations. Finally, we highlight challenges for future development and provide recommendations.

Problem

• Investigating UAV and LLM integration challenges
• Identifying gaps between academic and industrial applications
• Assessing real-world barriers to LLM-UAV implementation

Innovation

• Empirical study of papers and GitHub projects
• Online questionnaire gathering industry perspectives
• Identified nine LLM task types in UAV systems