SYSTEM IDENTIFICATION OF PARAMETERIZED STATE-SPACE MODEL OF A SMALL SCALE UAV HELICOPTER * ISMAILA B. TIJANI, RINI AKMELIAWATI † AND ARI LEGOWO Abstract. The success of model-based flight control law design for autonomous helicopter is largely dependent on the availability of reliable model of the system. Considering the complexity of the helicopter dynamics, and inherent difficulty involves with physical measurement of the system parameters, the grey modeling approach which involves the development of parameterized model from first principles and estimation of these parameters using system identification (sysID) technique has been proposed in the literatures. Prediction Error Modeling (PEM) algorithm has been identified as an effective system identification technique. However, application of this method to complex system like helicopter is not a trivial exercise due to inherent coupling in the system states and the challenges associated with parameter initialization in PEM algorithm. In this work, an effective procedure in application of PEM algorithm available in MATLAB toolbox is presented for small scale helicopter using real-time flight data. The approach was able to yield satisfactory model suitable for model-based flight control design. Key words. system identification, modeling, prediction-error modeling AMS subject classifications. 93B30, 97M10 1. Introduction. The need for simple and effective mathematical representa- tion of a helicopter dynamics has been one of the major challenges in model-based flight control design for the system. However, modeling of helicopter dynamics is not a trivial task due to the system’s complex dynamics, nonlinearities, instability and high degree of coupling among its state variables. Several approaches have been adopted in helicopter modelling since the inception of research activities on small-scale heli- copters in academic institutions in the early 1990’s [1]. The modeling approaches can be generally categorized into the following: first principle approach (white model), parametric approach based on system identification (grey model) and non-parametric approach (black box). For model-based control application, only the first two ap- proaches yield the needed mathematical representation for controller design. The first principle approach to modeling involves derivation of mathematical model describing the system of interest by using the fundamental laws of mechan- ics and aerodynamics [2]. Full modeling includes flexibility of the rotors and fuselage, stabilizer effects, dynamics of the actuators and combustion engine[3], [4]. A rigid- body assumption imposed on the system is used as the starting point with the inputs are forces and torques applied at center of gravity (cg), and outputs are the linear position and velocity of the cg as well asthe rotation angles and angular velocities. This is then followed by augmentation with the system sub-dynamics such as rotor- fuselage dynamics, flybar dynamics, flapping dynamics, stabilizer bar and etc. [2], [4]. This approach usually leads to a high-order nonlinear coupled model useful mostly for simulation purposes [3]. In addition, this approach requires a detailed knowledge of and experience with all the phenomena involves in rotorcraft system. These have been the major demerits associated with the use of purely first-principle model approach. Also, it has been observed, that however detailed the resulting model could be, a rig- orous validation against flight data collected in real-time from the intending platform * This research was supported by RMGS (Research Matching Grant Scheme), RMGS-09-02, Re- search Management Center, IIUM, Malaysia. † rakmelia@iium.edu.my, Intelligent Mechatronics Research units (IMSRU), Department of Mecha- tronics Engineering, International Islamic University Malaysia (IIUM). 1