Speech-Guided Drone Control System Based on Large Language Model

Abstract

The growing use of drones in logistics, agriculture, surveillance, and search-and-rescue operations has highlighted the need for more intuitive and adaptable control systems. Traditional control methods, such as joysticks and mobile applications, often necessitate technical expertise, thereby limiting their accessibility in hands-free or time-sensitive scenarios. This paper introduces a Speech-guided Drone Control System that utilizes Large Language Model (LLM) to process and execute natural language commands. The system is composed of three key components: speech-to-text conversion, natural language command interpretation, and drone control execution. The proposed system enables users to issue complex commands through natural speech, which are then processed and broken down into executable drone actions in real time. The system underwent rigorous testing across various levels of command complexity, demonstrating high accuracy and responsiveness even in handling ambiguous or incomplete commands. The results suggest that this approach not only simplifies drone operation but also represents a transformative step toward making advanced drone control universally accessible and highly efficient in critical sectors such as disaster response and precision agriculture. © 2025 IEEE.