Comparison of Numerical Differentiation Techniques for Aircraft Identification

External disturbance and measurement noise during flight tests inevitably degrade the identification of the aircraft aerodynamic models. Traditional approaches, however, need to differentiate the measured signals to build the identification models, which results in a dedicated preprocessing to avoid noise amplification. The aim of this paper is to assess the influence of four derivative estimation techniques on the parameter estimation of an aircraft aerodynamic model. Among the four studied techniques, two come from the field of robot identification. The other two techniques are the standard one in aircraft identification based on the Savitzky-Golay algorithm and a suggested one based on wavelet denoising coupled with finite differences. The two techniques coming from robot identification are the usual one relying on a low-pass filter applied in both forward and backward directions and a recently suggested method based on a Kalman filter with a first-order random walk model. The comparison simulation results illustrate that the first robot differentiation strategy not only performs well in providing accurate stability and control derivatives even in the presence of colored disturbance but also is competitive with respect to the standard method in aircraft identification.