Flight dynamic modeling and considerations for fixed wing aircraft in realistic urban wind environments

A modular simulation framework has been developed in Simulink to evaluate aircraft response within urban wind fields. The framework allows users to modify various aspects of the environment to support targeted studies relevant to aircraft operations in urban settings. These customizable elements include arbitrary wind field domains, control law architectures, and aerodynamic models such as vortex lattice methods or stability and control derivatives obtained from flight testing. The studies that informed the development of this environment, along with its application to urban wind field research, are presented below. The aerodynamic solvers employed in this work include an unsteady vortex lattice method (UVLM), implemented via the aeroelastic toolset SHARPy, and a compact vortex lattice method (CVLM) chosen for its computational efficiency. The UVLM was utilized in the initial two studies to investigate how aircraft with varying wing loadings respond to wind fields generated using Large Eddy Simulations (LES) and reduced-order models derived from Proper Orthogonal Decomposition (POD). These studies provided valuable insights into the level of wind modeling fidelity required for aircraft operating in urban environments. Due to the high computational cost of wake modeling with UVLM and the limited control law capabilities within SHARPy, a modular Simulink environment was developed to offer greater flexibility for these analyses. The CVLM was selected as the primary aerodynamic model in this environment, enabling efficient exploration of broad spatial domains within urban wind fields. Furthermore, the simulation was integrated with ArduPilot control laws to support waypoint tracking and disturbance rejection. The 4-point method was also integrated into the state formulation to more accurately account for asymmetric loadings encountered as the aircraft navigates through wind structures. For validation, the results for the CVLM are compared to a flight-derived model created using system identification techniques. Additionally, a machine-learning algorithm known as a Convolutional Autoencoder (CAE) is employed to represent the urban wind field used in previous studies. This approach achieved an order-of-magnitude reduction in storage requirements compared to raw look-up tables, without compromising the accuracy of the reconstructed flight dynamic time histories. Moreover, it demonstrated noticeable improvements over the POD models developed in earlier work.