Data-Based Modeling and Control of the Nonlinear Aircraft System Using Extended Implicit Sparse Identification

Abstract

Modeling and controlling aircraft can be particularly challenging when the system is highly nonlinear or only partially understood. While data-driven approaches can be promising in this regard, they are often bottlenecked by the requirement for extensive training data and may struggle to generalize beyond the training set. To this end, we propose an extended implicit version of the sparse identification of nonlinear dynamics with control (EISINDYc) to identify rational nonlinear aircraft dynamics. The proposed approach integrates prior flight mechanics knowledge and variable correlations to automate model selection using an information criterion. As a result, EISINDYc requires a reduced size and number of candidate functions while being able to balance accuracy with interpretability. Moreover, in this work, it is combined with nonlinear model predictive control (NMPC) to perform maneuver control. Simulation studies show that EISINDYc-NMPC has improved prediction accuracy, control performance, and generalization capability under low data conditions.