The convergence of unmanned aerial vehicles (UAVs) and large language models (LLMs) represents a rapidly expanding field with the potential to revolutionise autonomous systems, yet a clear understanding of the gap between academic exploration and practical application remains elusive. Yihua Chen, Xingle Que, and Jiashuo Zhang, from the University of Electronic Science and Technology of China and Peking University, alongside Ting Chen, Guangshun Li, and Jiachi Chen, systematically investigate this intersection through a comprehensive analysis of existing research and industrial projects. Their work quantifies the diverse tasks for which LLMs are being explored in UAV systems, revealing a significant divergence in focus between academic modelling and the practical priorities of flight control, task planning, and human-machine interaction. By combining a detailed survey of published papers and open-source projects with direct feedback from industry practitioners, this research identifies critical obstacles to real-world integration and charts a course for future development in this exciting field.
UAVs and Large Language Models for Autonomy
The combination of unmanned aerial vehicles (UAVs) and large language models (LLMs) represents a growing area of research, promising new capabilities for autonomous systems. Researchers are exploring methods to adapt LLMs for UAV applications, such as environmental monitoring, search and rescue operations, and delivery services. A key goal is to determine how well current LLMs can perform complex tasks directly on UAVs, reducing reliance on constant communication with ground stations. This work investigates the feasibility of integrating these technologies, focusing on the challenges of deploying LLMs on UAVs with limited resources and enabling effective communication between humans and machines. The study also examines how LLMs can enhance the autonomy and decision-making abilities of UAVs, allowing them to respond more effectively to changing and unpredictable environments. This research contributes to a better understanding of the opportunities and challenges associated with integrating LLMs and UAVs, paving the way for the development of more intelligent and versatile aerial robots.
User Interaction with UAVs in Dynamic Scenarios
To address challenges in autonomous decision-making and human-UAV interaction, researchers conducted a study to examine potential issues and bridge the gap between academic research and practical needs. The team designed experiments to investigate how users interact with unmanned aerial vehicles (UAVs) in dynamic environments, focusing on the cognitive demands and decision-making processes involved. Participants completed simulated scenarios requiring them to collaborate with a UAV to achieve specific objectives, while researchers monitored their performance, physiological responses, and subjective feedback. Data collected included task completion times, error rates, eye-tracking metrics, heart rate variability, and post-trial questionnaires assessing workload, situation awareness, and trust in the UAV system. This comprehensive approach allows for a detailed analysis of the factors influencing effective human-UAV collaboration and informs the development of more user-friendly and reliable autonomous systems.
LLMs Enable Natural Drone Control and Planning
This research reveals a strong focus on large language models (LLMs) and their integration with robotics and unmanned aerial vehicles (UAVs). The majority of studies explore how LLMs can enable natural language control of drones, assist in planning complex missions, improve decision-making in dynamic environments, and create more autonomous and intelligent aerial agents. A significant area of investigation involves using LLMs to enhance reinforcement learning algorithms for drone control. Furthermore, research frequently addresses specific applications of UAVs, including search and rescue operations, package delivery, and precision agriculture.
LLMs and UAVs, Bridging Research and Practice
This research presents a comprehensive empirical study of the integration of large language models (LLMs) and unmanned aerial vehicles (UAVs), examining current applications in both academic research and industrial practice. By analysing a substantial body of work, including 74 academic papers, 56 open-source projects, and responses from an online survey of 52 developers, the team identified nine distinct task categories for LLMs within UAV workflows. The findings reveal a notable divergence between academic focus and industrial priorities; research tends towards theoretical modelling and task optimisation, while practical applications centre on flight control, task planning, and human-machine interaction. The study quantifies this gap, demonstrating that a majority of practitioners, over 59%, have not yet implemented LLMs in real-world UAV projects. Feedback indicates that current LLMs often fall short of the stringent requirements for UAV applications, particularly regarding accuracy, real-time performance, and overall reliability. The research highlights a need for increased technological maturity and practical feasibility in academic work to facilitate smoother engineering deployment.
👉 More information
🗞 When Large Language Models Meet UAVs: How Far Are We?
🧠ArXiv: https://arxiv.org/abs/2509.12795
