Iterative and Neural Network Based Methods to Solve a Model-Free Scheme for Flow Angle Estimation

Evaluation of aerodynamic, or flow, angles has always been a crucial topic since they are necessary to pilot and control the aircraft. These angles are usually measured using different probes attached to the vehicle fuselage surface which are surrounded with the flow field. However, new solutions have been explored in order to reduce this number of probes so that a better stealthiness and a less heavy impact on the airframe, especially for small UAVs, can be achieved. For this purpose, virtual software-based sensors, based on neural networks prediction techniques have been developed and proven to be suitable for the aerodynamic angles estimation of small UAVs. The aim of this work is to evaluate the accuracy in the estimation of neural network solvers for the angle of attack and the angle of sideslip of a model free scheme named ASSE \cite{ASSE}. In this work, all the tests have been carried out considering a null wind velocity but, for sake of clarity, conclusions of this work are also applicable when the wind speed is not zero. The suitability of the model-free ASSE scheme will be tested for both traditional solution methods (i.e. non-linear solver) and neaural network based methods.