Research On The Key Technology Of Integrated Navigation For Multi-rotor UAV Based On Low-cost Inertial/GNSS/Vision

Multi-rotor unmanned aerial vehicles(UAVs)are widely used in military,public security,agriculture and forestry and other fields in recent years due to their advantages of strong environmental adaptability,easy control,strong flight mobility,stability and reliability.In the integrated navigation algorithm for multi-rotor UAV,many researchers have proposed sensor errors calibration and attitude estimation algorithms based on high-precision sensors,and have achieved many remarkable results.However,with the application and popularization of rotary-wing UAV,especially expensive special-purpose UAV,integrated navigation algorithms based on low-cost sensors are urgently needed,and higher requirements are put forward for the accuracy,reliability and real-time performance of the solution.In this paper,CQ series of heavy-duty UAV independently developed by Changchun Institute of Optics,Fine Mechanics and Physics are taken as the research platform.On the basis of in-depth research on sensor error calibration of existing high-precision sensors and integrated navigation EKF algorithm,the sensor error calibration algorithm and integrated navigation EKF algorithm based on low-cost sensors are proposed.The navigation key information calculation with high precision,high stability and high real-time performance of low-cost sensors is realized.Aiming at the indoor environment without GNSS signals,an EKF attitude estimation method based on vision-inertial navigation is proposed,which realizes high precision attitude estimation assisted by vision in the environment without GNSS signals.The main work of this paper includes the following aspects:(1)The existing sensor data calibration method,integrated navigation algorithm model and data fusion process based on high-precision sensors for multi-rotor UAV are deeply studied,and various problems existing when the algorithms are applied to low-cost sensors are analyzed.The improved method of integrated navigation algorithm is proposed,which lays a theoretical foundation for the construction of original sensor data calibration and integrated navigation algorithm for low-cost sensors.(2)The error sources of low-cost magnetometers and accelerometers are deeply studied,and their error models are established respectively.In view of the complex magnetic field when magnetometers are used in multi-rotor UAV,a fast calibration method for magnetometers is proposed through the combination of normal sphere fitting and ellipsoid fitting with constraints based on L-M algorithm.It has good calibration effect on “Hard iron” error and “Soft iron” error which affect the magnetometer on UAV.For the calibration of low-cost accelerometers,a six-position calibration method based on L-M algorithm is proposed,which effectively calibrates the zero bias error,scale factor error and non-orthogonal error of accelerometers.The proposed algorithm gives a detailed mathematical model and solution process,and verifies the effectiveness and reliability of the algorithm through measured data.The calibration method for magnetometers and accelerometers proposed in this paper can effectively calibrate various errors existing in low-cost sensors without external high-precision auxiliary equipment.(3)An integrated navigation system platform based on low-cost inertial measurement unit,magnetometer,barometer,GNSS module and binocular camera is built.Aiming at the requirements of low cost and high precision,a high-dimensional EKF data fusion algorithm based on 22-dimensional state variables is proposed.A complete mathematical model and solution steps are constructed.MATLAB simulation of measured data proves the stability and effectiveness of the algorithm.In the engineering implementation of integrated navigation algorithm,aiming at the problem that abnormal interference data may cause EKF divergence,a complementary filter monitoring module is introduced and a CPF-EKF monitoring algorithm is constructed to reset and align immediately when EKF diverges,thus ensuring the high solution accuracy and stable operation of the navigation system.The effectiveness of the proposed CPF-EKF algorithm is verified by actual flight tests.(4)The vision-inertial navigation integrated navigation algorithm of unmanned aerial vehicle in GNSS signal-free environment is studied.A high-dimensional EKF algorithm based on 22-dimensional state variables has been established.Within thisframework,binocular vision sensors are introduced and a vision-inertial navigation EKF integrated navigation attitude estimation algorithm combining binocular vision and inertial navigation system is proposed.According to the relative motion and attitude transformation between the adjacent key frames of the binocular camera,the bias error of the gyroscope and the accelerometer in the IMU is obtained as an observation measure of the bias of the gyroscope and the accelerometer in the state quantity,and is introduced into the EKF data fusion process to further update the optimal estimation value of navigation information.The validity of the vision-inertial navigation combination algorithm is verified by EuRoC data set,and the precision of attitude estimation of the algorithm is fully verified in actual flight tests.