Physics-informed generative methods of aircraft trajectory for terminal airspace design evaluation

The terminal airspace surrounding an airport is characterized by high flight density and complex structures. While aircraft are generally assigned standard arrival and departure routes, deviations frequently occur due to factors such as air traffic control directives, pilot decisions, surveillance limitations, and varying flight performance. Accurately capturing these trajectory uncertainties and generating realistic trajectories is pivotal when revising established routes and guiding the design of future terminal airspace. Traditional simulation-based methods often involve complex, time-consuming, scenariospecific setups and lack the ability to consider actual operational factors. AI generative methods offer promise for rapid trajectory generation but face critical limitations. A key challenge is integrating real-world physical factors, such as aircraft dynamics and operational constraints, into generative models. Many current methods oversimplify or omit these essential factors, focusing on replicating historical data patterns without incorporating the real-world dynamics governing aircraft behavior. Additionally, these models often focus on historical distribution modeling, limiting theirflexibility to adapt to novel designs or changing operational scenarios. This gap highlights the need for a more comprehensive approach that incorporates physical laws and operational realities, enabling accurate simulations of both current and novel airspace configurations. In this project, we propose developing physics-informed generative methods to generate realistic aircraft trajectories for terminal airspace design evaluation.The investigation will focus on three main areas: (1) Physics-informed neural network modeling, embedding aircraft dynamics and design constraints into the generative process to ensure adherence to real-world physical laws; (2) Data-driven modeling of trajectory uncertainties, capturing flight behaviors and operational factors and replicating historical patterns; and (3) Transferable design-deviated pattern discovery across various airspace designs, ensuring the model can be applied to diverse airport terminalairspace configurations, enhancing flexibility and transferability. By embedding physical constraints into the generative process, our approach bridges the gap between rapid, efficient modeling and the complex realities of real-world operations, offering a robust framework for comprehensive airspace design assessments. This framework will facilitate fast, flexible, and accurate evaluations of airspace redesigns, improving decision-making in airspace management, while enhancing efficiency, reducing delays, and increasing safety across airport operations.