Research And Implementation Of A Quad-rotor UAV Positioning System Based On Multi-sensor Fusion

In recent years,with the rapid development of computer technology and electronic information technology,the field of UAV has been blowout development.In the field of UAV,the most basic and important research direction is UAV navigation and positioning.Navigation and positioning algorithm is the basis and key of UAV’s upper application,such as motion control and autonomous navigation.In the actual work and production,people often need quadrotor UAV to complete the operation in the more complex environment such as cave,tunnel,field and so on.When quadrotor UAV works in this complex situation,it can not complete the reliable positioning by relying on a single sensor.In this paper,aiming at the shortcomings of traditional navigation and positioning methods in robustness,accuracy and environmental adaptability,we focus on the positioning algorithm based on GPS / vision / IMU and other multi-sensor fusion,and verify the algorithm on the quadrotor UAV platform.The main work of this paper is as follows:Firstly,for indoor environment positioning without GPS signal,this paper studies the modular positioning framework of vision fusion IMU based on extended Kalman filter.It can fuse the position and attitude of UAV calculated by camera’s visual odometer.Aiming at the problem of inconsistent data acquisition frequency between camera and IMU,a multi-sensor time synchronization algorithm is designed.In order to eliminate the cumulative error and fuse the relative observation information of the visual odometer,a clonal Kalman filter based on the error Kalman filter is implemented.In addition,for the situation that the pose of vision algorithm is not robust enough,the corresponding correction processing is carried out to improve the robustness of the system.Secondly,for the outdoor environment with GPS signal,in order to achieve robust long-term drift free position estimation,this paper studies the GPS fusion vision algorithm based on sliding window optimization,and uses the correlation between GPS and the pose information obtained by the above filtering method.The system state is estimated by minimizing the cost function including vision and GPS global position residuals.In order to reduce the computational cost of the system and deploy on the airborne equipment,this paper uses the method of marginalization to process the old information and keep the optimized data within a certain amount.Finally,the algorithm is tested on public data sets and quadrotor UAVs,which proves the practicability and robustness of the algorithm.