Low Level Model Identification of a Quadrotor X3D-BL

Abstract

Resumo-This work aims at identifying the low level model of a miniature quadrotor, using a sensor of force and torque in the three axes of the Cartesian system. To do that, the first step is to obtain the low level model of the rotorcraft. Then we obtained the coefficient C f ,τ , which is a meaningful contribution of this work, since there are just a few works in the literature addressing such coefficient, necessary to convert the force and torque on the three axes in the individual strength of each propellant. As a continuation of this work, the parameters associated to the low level model here discussed will be also identified. Index Terms-Low-level modeling; Model identification; UAV model. I. INTRODUCTIONThe action of a controller capable of guiding an aircraft in predefined flight missions is one of the elements necessary for autonomous navigation. The design of such a controller often requires a model that suitably describes (following some criteria previously adopted) the behavior of the vehicle to be controlled. Many design techniques for flight controllers are based on the mathematical model of the aerial vehicle.Regarding the mathematical modeling of a small-scale aircraft, there are two well defined approaches in the reference literature: one based on the physical equations of the system and the other based on system identification techniques [1], [2]. Such approaches are not mutually exclusive: very often the use of one of them is required to simplify the other. In general, the first approach uses the equations of motion for the mechanical representation of a physical system, whereas the second one estimates the dynamic model of the physical system based on the data correspondent to excitation and response. According to [3] and [4], the complete model of a quadrotor (shown in Fig. 1) can be represented by four interconnected subsystems, as shown in Fig. 2 where u θ , u φ , uψ and uż are the excitation correspondent to the angles of pitch (θ ) and roll (φ ), the excitation correspondent to the yaw rate (ψ) and the excitation correspondent to the rate of vertical lift (ż), respectively. In turn, ω i and f i , for i = 1, 2, 3, 4, are the rotational velocities developed by each motor and the forces generated by them, respectively.The first two blocks shown in Fig. 2 are responsible for receiving control signals and generating the forces that will act on the aircraft, which are all normal to the plane b X − b Y that characterizes the frame b X − b Y − b Z attached to the body of the vehicle (see Fig. 1), whose origin is the center of mass

Experimental Parameter Identifications of a Quadrotor by Using an Optimized Trajectory

“…As suggested in [6], [23], [24], [42], [44], [45], [47], [52]- [56], in this work, the nominal parameters provided by the manufacturer and the new estimated parameters are assessed by comparing real-time experiments with simulations results, and by comparing the output prediction error of the regression model, which is also often called identifier error. Our results indicate that the new estimated parameters show more similarity to the experiments and lower output prediction error than the nominal parameter given by the manufacturer.…”