Maneuver-Based Cross-Validation Approach for Angle-of-Attack Estimation

The estimation of the Angle of Attack (AOA) and Angle of Sideslip (AOS) is crucial for flight monitoring and control. However, a gap has been identified on the data selection technique for the class of estimators based on data-driven methods, such as the synthetic sensor based on Neural Network (NN). This paper proposes a Cross Validation (CV) technique applied on a manoeuver-based partitioning method to provide evidence that a given selection of data can lead to better estimation performance, with the final aim of providing a list of manoeuvers suitable for the training phase of the estimator. Results are shown using simulated data related to the CleanSky 2 project MIDAS.