Research Of UAV Path Following And Positioning Method Based On Visual Inertial Odometry

In recent years,with the development of UAV(Unmanned Aerial Vehicle)technology,the application of UAVs in urban environment has become more and more extensive,such as search and rescue,public facility monitoring,etc.In most missions,UAVs are required to fly autonomously along a pre-planned path,so efficient UAV path following technology is the basis for the successful completion of the mission.In addition,the realization of path following requires real-time position of the UAV,while the canyon terrain formed by the tall buildings in the city leads to poor or even missing GPS(Global Position System)signals,which cannot guarantee the real-time positioning requirements of the UAV.In order to meet the application requirements of UAVs in urban environments,this thesis studies the UAV path following method in the GPS-denied situation,so that the UAV can accurately and efficiently fly following the pre-planned path in the absence of GPS signals.Under the guidance of this research goal,the main research contents of this thesis are as follows:According to the implementation requirements of UAV path following in urban environment,the UAV path following and positioning system is designed to provide platform support for the subsequent algorithm implementation.For the inertial measurement unit(IMU)and monocular camera used in the UAV positioning method,the error model of IMU,the pinhole camera model,and the corresponding correction methods are analyzed.The calibration method is verified on the data set,and the sensor parameters of the IMU and camera are obtained,which provides more accurate and effective sensor data for the implementation of the positioning method.In order to solve the problem of the lack of optimal selection of guidance law parameters in path following algorithm,a strategy of optimal selection of guidance law parameters based on genetic algorithm was studied,and the parameters selected according to this strategy can minimize the cost of the path following algorithm.Based on the UAV path following and positioning system,the path following algorithm is verified by physical experiment.According to the track error and control benefit,the evaluation index is formulated,and used for the path following results’ quantitative analyzation.The experimental results show that the guidance law parameters selected by genetic algorithm are optimal,which can achieve accurate and efficient path following of UAV.In view of the poor or even missing GPS signals in urban environments,a UAV positioning method based on loosely coupled visual inertial odometer was studied.Firstly,the monocular visual odometer is realized through SLAM(Simultaneous Localization and Mapping)algorithm,and the motion poses with uncertain scale of camera are obtained.Based on the loosely coupled fusion framework,extended Kalman filter is used to achieve the fusion positioning of IMU data and visual odometer result.Tested on the Eu Ro C data set,the experimental results show that the method can provide positioning error within 5 meters when the data collection system moves gently,and realizes the effective positioning of the drone in the GPS-denied situation.Through the research of this thesis,the UAV adaptive path following and the positioning of the UAV based on the loosely coupled visual inertial odometer are realized,which provides technical support for improving the autonomous flying ability of the UAV in the urban environment and is beneficial to expand the application of UAVs in urban observation,logistics and emergency.