The field of Unmanned Aerial Vehicles (UAV) has gained great significance in the R&D activities of several institutions, and numerous realizations have been constructed. Controlling the movement of a UAV is one of the most significant tasks in the hierarchical control system that should be realized in order to successfully manage UAVs in several missions. Precise control of the movement requires accurate mathematical models of the aircraft that correspond to the several control schemes that are intended to apply, including linearised, LPV, and nonlinear ones. The source of mathematical models includes the physical model constructed on the basis of the Newtonian equations and the rules of aerodynamics, as well as the use of empirical observations, measurements. The methodology of system identification offers the tools that can be used to manage this field. The Systems and Control Laboratory of the Computer and Automation Research Institute in cooperation with University of Minnesota and Budapest University of Technology and Economics has built - as a part of a more general framework established to solve complex UAV control problems - small UAVs, equipped with on-board sensors and embedded computer with the purpose of obtaining an adequate test platform for the aircraft experiments. Based on on-board measurements acquired during the flight tests performed with the UAVs, the opportunity has been given to test several system identification approaches corresponding to the control tasks to be solved. This paper gives an outline of the system identification and modelling methods that have successfully been applied, and also analyzes the problems encountered in the data-acquisition - signal processing - computing chain.