Comparison of the methods of parametric identification of linear dynamical systems under mixed noise

In the article we study the possibility of comparison of parametric identification methods by the sensitivity theory via local expansions of the objective functions using as an example three identification methods. The theoretical results are verified by computational identification of the equations of longitudinal motion of the aircraft which parameters are identified by a) the linear least-squares method, b) the method of instrumental variables in frequency domain, c) the variational method (closely related to the STLS and GTLS methods). The simulation used a mixed additive noise: in the observations and in the residuals of the model equations.