Research On UAV System Identification And Control Algorithms

Unmanned Aerial Vehicle(UAV)have been widely used in military and civilian field and the development of control theory has drastically improved the performance of UAV control systems.The core part in design of robust flight control systems is the precise dynamics model of UAVs.With the innovation and improvement of control theory,many researches of dynamic modelling of UAV s have been proposed.Among of these methods,the system identification method needs a smaller amount of calculation than traditional mechanism modelling and lower cost than wind tunnel tests.The rapid development of related theories of flight control systems demands for higher accuracy of system identification and faster convergence speed of parameter estimation.At the same time,the design of controllers for UAVs is also a hot and difficult topic in UAV control.Therefore,this thesis will study the identification of the dynamic model of a new kind of variable structure vertical take-off and landing(VTOL)UAV,and design the LQR controller based on the identified model.Finally,a nonlinear dynamic inverse controller is designed directly for the UAV.The main work is as follows:First,according to the mechanism modeling method,the dynamics and kinematics models of a new kind of vertical take-off and landing UAV and the mathematical model of the wind are established.The nonlinear model is decoupled to obtain the longitu-dinal model of the UAV in fixed-wing mode.By selecting the state of cruising flight as the reference state,longitudinal linear mathematical model of the UAV is gained through linearization method of small disturbances.The modal analysis of the longitu-dinal model is performed,and the dynamic performance and stability of the system are improved through the pole configuration.Then the model of the UAV is built in MATLAB/Simulink.The input-output data of the UAV is obtained by appropriate excitation signals.Based on these input-output data,the estimation of the parameters in UAV model is gained from the joint estimation of state and parameter of nonlinear system,by unscented Kalman filter(UKF)and square root-unscented Kalman filter(SRUKF).Prior to the parameters estimation,a series of pre-processing such as de-noising and filtering are performed on these data.In order to verify the accuracy of the identification results,the system responses before and after the model identification are compared under the same excitation.Finally,the time-scale separation method is used to separate the attitude control of the UAV into fast and slow loops,and the nonlinear dynamic inverse controller is de-signed for the attitude control loop.One of the difficulties in the control of the variable structure VTOL UAV is the control distribution of the vector engine and the aerody-namic rudder surface during the mode conversion.This thesis introduces the minimum energy allocation and chain allocation scheme,and adopts the chain allocation scheme to design the nonlinear dynamic inverse controller for the first stage of the mode con-version process,and to design the controller of the height channel of the UAV through dynamic inverse.Simulations verify the effectiveness of the designed controller in the VTOL mode,the first stage of mode conversion process and the fixed wing mode.