Aircraft parameter estimation using recurrent neural networks - A critical appraisal

This paper investigates applicability of dynamic neural networks, often called as Recurrent Neural Networks (RNN), to aircraft parameter estimation. It is demonstrated that the RNNs are amenable to state space model representation and hence adoptable for estimation of aerodynamic derivatives retaining the model parametrization based on the physical insight. Several aspects of similarity between the RNN approach and the conventional aircraft parameter estimation methods such as output error, equation error, or extended Kalman filter are brought out. The proposed approach is first validated on simulated data and then extended to estimate from flight data the aerodynamic derivatives pertaining to the lateral-directional motion of an aircraft. Based on the identification results, the practical utility, advantages and limitations of the RNN approach are critically appraised