Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Prompt-driven Task Planning Method for Multi-drones based on Large Language Model

Yaohua Liu

TL;DR

This paper addresses the complexity of multi-drone task planning and human–machine interaction by introducing a prompt-driven method that uses LLMs to interpret natural language prompts and generate executable drone control code. The main approach combines a drone motion function library with system and user prompts, enabling high-level task understanding and flexible coordination without extensive task-specific retraining. Key contributions include the library design, the prompt architecture separating system and user prompts, and experimental validation on a Tello drone platform for single and multi-drone scenarios, including synchronous and asynchronous control. The findings suggest that prompt-driven LLM control can deliver flexible, scalable multi-drone operations with enhanced human–drone interaction in dynamic environments, reducing the need for large task-specific datasets.

Abstract

With the rapid development of drone technology, the application of multi-drones is becoming increasingly widespread in various fields. However, the task planning technology for multi-drones still faces challenges such as the complexity of remote operation and the convenience of human-machine interaction. To address these issues, this paper proposes a prompt-driven task planning method for multi-drones based on large language models. By introducing the Prompt technique, appropriate prompt information is provided for the multi-drone system.

A Prompt-driven Task Planning Method for Multi-drones based on Large Language Model

TL;DR

This paper addresses the complexity of multi-drone task planning and human–machine interaction by introducing a prompt-driven method that uses LLMs to interpret natural language prompts and generate executable drone control code. The main approach combines a drone motion function library with system and user prompts, enabling high-level task understanding and flexible coordination without extensive task-specific retraining. Key contributions include the library design, the prompt architecture separating system and user prompts, and experimental validation on a Tello drone platform for single and multi-drone scenarios, including synchronous and asynchronous control. The findings suggest that prompt-driven LLM control can deliver flexible, scalable multi-drone operations with enhanced human–drone interaction in dynamic environments, reducing the need for large task-specific datasets.

Abstract

With the rapid development of drone technology, the application of multi-drones is becoming increasingly widespread in various fields. However, the task planning technology for multi-drones still faces challenges such as the complexity of remote operation and the convenience of human-machine interaction. To address these issues, this paper proposes a prompt-driven task planning method for multi-drones based on large language models. By introducing the Prompt technique, appropriate prompt information is provided for the multi-drone system.
Paper Structure (11 sections, 7 figures)

This paper contains 11 sections, 7 figures.

Table of Contents

  1. Introduction
  2. Drones with LLM
  3. Drone Motion Function Library
  4. Prompt Design
  5. Human-LLM interaction
  6. Drone Code Generation and Execution
  7. Experiments
  8. LLM control of a single drone
  9. Synchronized control of multiple drones
  10. Asynchronous control of multiple drones
  11. Conclusion

Figures (7)

  • Figure 1: Framework of the prompt-driven control method for multi-drones based on LLM.
  • Figure 2: The system prompt.
  • Figure 3: The user prompt.
  • Figure 4: The voice-text user interface.
  • Figure 5: The single drone tasks.
  • ...and 2 more figures