Research On Position And Attitude Estimation Of Multi-Rotor UAV Based On Multi-Source Information Fusion

Miniature multi-rotor unmanned aerial vehicle(UAV)becomes more and more important in the field of aviation.The measurement of position and attitude,as well as the attitude control,are critical for the flight control system of multi-rotor UAV.In this thesis,two major problems are discussed,and corresponding solutions are proposed to achieve precise measurement of multi-rotor UAV's position and attitude.The first problem is to measure the UAV's position and attitude with high accuracy and reliability,while eliminating or compensating the sensors' errors.The second problem is to further enhance the precision of UAV's position and attitude through multi-source information fusion.The theoretical and methodological research of the above two problems has great academic and practical significance.The contributions and innovations of this thesis can be summarized as follows.Firstly,the problem of sensors' errors in UAV are discussed.A unified error model is introduced for three-axis sensors,including magnetometer,accelerometer,and gyroscope.The errors of attitude and heading angles caused by misalignment error are analyzed according to this model.Through theoretical analysis,it is pointed out that the scalar calibration method based on vector modulus invariance has unavoidable residual heading error.For this reason,the double-vector inner product method and two-step method are used for joint calibration of magnetometer and accelerometer,so as to better compensate the misalignment error.However,the vector inner product method needs to constrain the reference vector,while the two-step method requires that the error compensation matrix being an orthogonal matrix when compensating the misalignment error.Therefore,a double inner product method is proposed for the calibration of three-axis vector sensor.The double inner product method combines the advantages of scalar calibration method and vector inner product method.It constructs the error objective function using double inner products,and can solve the compensation matrix by optimizing the objective function.This method is able to eliminate the misalignment error from magnetometer to either another sensor or the body frame,and it can maintain better effect when affected by measurement noise.Secondly,for the gyroscope,since its errors can only be fully observed under dynamic conditions,the above calibration methods that designed for magnetometer and accelerometer can only partially calibrate the errors of gyroscope,such as the bias.Therefore,a vector outer product algorithm is proposed to calibrate the three-axis MEMS gyroscope in the attitude and heading reference system(AHRS)of UAV.This algorithm can essentially unify all the existing calibration methods based on gravity vector,and it has the advantages of more concise expressions and more convenient calculations.Meanwhile,this method does not need turntable or other precision equipment.Thus,it is suitable for in-field calibration of three-axis MEMS gyroscope.The numerical simulation,in-field calibration and flight experiment of six-rotor UAV show that the proposed method can accurately determine the error coefficients of gyroscope,and its accuracy is comparable to that of turntable-based calibration method.The combination of vector outer product calibration method and recursive filtering algorithm can obtain stable attitude angles,which is conducive to flight control and load task execution of UAV.Thirdly,the attitude fusion algorithm for multi-rotor UAV is also discussed.The architecture of attitude information filter based on MARG sensor is designed,and the attitude filtering algorithm is analyzed and improved by the proposed structure.It is verified that the proposed attitude filter can improve the accuracy of attitude information output by UAV navigation system.In order to improve the accuracy of horizontal attitude,an attitude fusion algorithm based on vector observation and motion acceleration adaptive estimation is proposed.Based on this adaptive algorithm,a vector-based parallel attitude filter structure is designed to suppress the interference of both acceleration and magnetic field.This structure can be utilized with various filtering and data fusion algorithms that commonly used.It can better suppress strong and persistent interference,and also enhance the reliability of attitude estimation and navigation information accuracy of multi-rotor UAV.Finally,for the airborne multi-sensor system,a multi-level decentralized fusion structure is designed,and a federal Kalman filter algorithm is used for information fusion.Through the study of information allocation strategy of federated filtering algorithm,the defect of the current information allocation methods based on covariance and failure probability is revealed.That is,there is a conflict between the system's accuracy and fault tolerance.Then,an adaptive information allocation strategy based on trade-off factors is proposed.The fusion errors of three different methods are compared through simulation,which proves the feasibility and reliability of the proposed method.On the basis of federal filtering algorithm,taking the elevation information measurement of GPS/electronic compass/SINS/Barometer integrated navigation system as an example,the mathematical models are given,including the state equations and measurement equations of sub-filters and main filter,so as to implement information fusion.UAV flight experiment shows that the proposed method can ensure the accuracy and reliability of navigation state estimation of multi-rotor UAV in complex near ground environment.