We introduce a novel framework to automate flight planning processes using large language models (LLMs) to identify flight operator's preferences. Our framework integrates the recent advancements in LLM and prompt engineering, low-altitude wind hazard forecasts, flight mission energy estimation, and pre-departure strategic deconfliction. First, our approach begins with the forecast of wind hazard polygons to ensure safety in flight planning. Second, we generate a diverse set of candidate flight plans to avoid these wind hazard polygons. The flight plan features include total flight distance, cruising altitude, flight mission energy consumption, and number of waypoints. Third, human flight operator specifies their preferences through natural language prompts or plain words, which are fed to the LLM to extract and prioritize these features. Our framework then evaluates and scores each flight plan based on extracted user-defined preferences, recommending the flight plan that best matches the flight operator's needs. For the purpose of demonstration, we focus on a flight planning use case for an electric vehicle take-off and landing (eVTOL) aircraft in an advanced air mobility (AAM) mission in Dallas-Fort Worth area. Our simulation experiments show the effectiveness of this approach in generating personalized, safe, and efficient flight plans. To our best knowledge, this work is among the first attempts using LLMs to enable a human-centric flight planning automation for AAM operations.
Using Large Language Models to Automate Flight Planning Under Wind Hazards
29.09.2024
4444713 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Using ASPEN to Automate EO-1 Activity Planning
NTRS | 1998
|Using ASPEN to automate EO-1 activity planning
IEEE | 1998
|Using ASPEN to Automate EO-1 Activity Planning
British Library Conference Proceedings | 1998
|Emerald Group Publishing | 2018
|