A Flight-Data Pipeline for Performance Model Identification of a Distributed Electric Propulsion Subscale Vehicle

Abstract

Performance models are needed for trajectory optimization and guidance design but can be difficult to obtain for small research aircraft. This paper uses onboard data from six remotely piloted flights to identify a nonlinear point-mass performance model for a 9-ft-wingspan fixed-wing research aircraft. The data were synchronized, filtered, segmented by flight phase, and fitted using least squares. The identified functions map airspeed, pitch, bank angle, and motor command to vertical speed, curvature, and airspeed rate. Leave-one-flight-out cross-validation produced test R2 values of 0.93 and 0.93 for the climb and curvature models, respectively. The identified curvature coefficient was cκ = 0.78. The airspeed-rate model produced a test R2 of 0.61 for the gravity-corrected response and captured the dominant longitudinal behavior. An in-sample forward simulation reproduced the measured airspeed over a representative interval without divergence. The resulting model supports preliminary three-dimensional trajectory generation within conservative, flight-supported design bounds and can be reidentified as the vehicle configuration changes.