Research On Modeling And Control For Quadrotor UAV

Quadrotor is a kind of simple unmanned aerial vehicle(UAV) drived by four propellers capable of the taking off and landing along the vertical direction that has been widely applied in the military and civilian area. Itis also a typical complex control system that is underactuated, instable, multivariable, uncertain, nonlinear and strongly coupled. Modeling and control of a quadrotor is a hot topic in recent years.Based on a summary of the recent research on quadrotor involving the key technologies and the future applications, this paper concentrates on the modeling and controller design of the quadrotor, Qball, used in the experiments. Some important theoretical analysis and research results are given as follows:Firstly, since the quadrotor is an underactuated, multivariable and coupled plant, its accurate mathematics model is difficult to be obtained. This paper studies the mobile character in five basic motion forms: hover, lifting movement, pitch motion, roll motion and yaw motion. By establishing the Earth-fixed frame and body-fixed frame, we analyze the external moments and forces on the quadrotor. Then, by viewing the quadrotor as a rigid body, this paper analyzes the kinematics and dynamics of the quadrotor, and the nonlinear mathematical model is built up.Secondly, because that the altitude estimation using single inertial sensor connot meet the measurement requirements, this paper studies the fusion algorithm of UAV attitude based on multi-sensors. In this paper, we study the characteristics, noise models and attitude algorithm principle. Then, the classical Kalman filter is adopted to estimate UAV’s attitude. Due to the fact that it highly relies on the accuracy of the noise models, this paper studies the adaptive Kalman filter based on the single innovation. Finally, the effects of attitude estimation of two kinds of filter are compared through the simulation.Thirdly, the linear quadratic regulator(LQR) is designed and implemented on the quadrotor. As the quadrotor system has serious nonlinear characteristics, the traditional LQR cannot deal with nonlinear dynamics leading the degradation of the robustness. In order to solve this problem, this paper proposes a new LQR controller based on Extended State Observes(ESO) named ESO-LQR. The performance of two kinds of controller are compared through the simulation.Fourthly, because that the conventional controllers can’t deal with the physical constraints, the model predictive controller(MPC) is proposed in this paper to control the quadrotor. Another Integral-Action-Embeded MPC controller is designed to eliminate the steady-state error. The performance of two kinds of controller are compared through the simulation.Last, the controllers are tested on the Qball to verify the effectiveness.