Improvement and development of simulation scenarios for air traffic management using Large Language Models (LLM) Artificial intelligence (AI) in the context of safe and efficient air traffic management

Abstract

This contribution presents an innovative approach to improving the development of simulation scenarios for Air Traffic Management (ATM) through the utilization of Large Language Models (LLMs). Traditional methods of scenario creation often rely on manual scripting, which can be time-consuming and limited in scope. The use of LLMs offers a novel solution by automating scenario generation based on natural text or speech input. By analyzing large text datasets, prompt engineering, and fine-tuning, LLMs can extract relevant information and generate scenarios that reflect various operational conditions, including airspace congestion, aircraft malfunctions, and weather events. This approach not only simplifies the process of scenario development but also enables the creation of dynamic and adaptable simulations that closely mimic real-world scenarios. Furthermore, the integration of LLMs into ATM simulations facilitates scenario customization, allowing users to tailor scenarios to specific training objectives or research requirements. This contribution highlights the potential of LLMs to revolutionize the development of simulation scenarios for ATM and contribute to more effective training, research, and system optimization in the aviation domain.