GPS Free Localization And Implementation In Robot Operating System For Unmanned Aerial Vehicles

Nowadays,unmanned aerial vehicle(UAV)system shows great efficiency in urban anti-terrorism,disaster detection,logistics,traffic dispersion and other military and civilian fields.These applications give the UAV system a promising future.During these tasks,it is necessary to locate the UAV accurately and robustly,and improving this self-localization ability shows a great significance to UAV system development.Aiming at the need of the UAV application in urban buildings/forest/indoor weak-GPS and pseudo-GPS environments,it is meaningful to develop the GPS free UAV localization technollogy significantly and search for the sensor fusion localization with onboard or auxiliary sensors.This thesis concentrates on the ground-based and onboard localization systems respectively,using cameras,laser scanner,IMU etc.to capture inner and internal information,and employ image segmentation,filtering,optimization etc.,to estimate the pose and other state of UAV during its flying.The major focuses and innovations are as follows:(1)Chan-Vese model-based UAV image detection algorithmThe Chan-Vese model is employed to segment UAV in the images captured from the two cameras of the ground-based localization system during UAV's flying.Then locating the true UAV region according to the shape and motion characteristics of UAV,and computing the coordinate of UAV center finally.In the image segmentation part,according to the specific application scenario,ROI set,initial condition and iteration stop condition configuration improve the accuracy and real-time capability significantly(2)Kalman filter-based UAV spatial localization algorithmIn the light of the detected UAV coordinates in left and right images and the feedback attitudes of the PTUs(Pan Tilt Units),an EKF(Extended Kalman Filter)is designed to compute the UAV spatial position,which improves localization accuracy and robustness efficiently.This EKF uses the continuity of the UAV movement and PTUs rotation,which significantly decreases localization error caused by UAV detection error and PTUs attitude measurement error.(3)Apriltag-based multi-cameras/laser sensors joint extrinsic calibrationMulti-sensors extrinsic calibration is to estimate the transformations among the sensor coordinate systems,which directly affects the sensors data fusion results.This algorithm designs an Apriltag array,distributing multiple apriltags on a plane,to be our aided calibration parttern.This algorithm calibrates camera-camera and camera-laser extrinsic parameters respectively firstly and optimizes these parameters then.Finally,integrate all the extrinsic parameters and a globally nonlinear optimization process is designed to further improve the calibration accuracy.(4)Onboard camera/laser fusion UAV localization algorithmCamera and laser scanner capture the environment information from different aspects,and their availabilities need different environment conditions such as light,structure etc..Fusing camera and laser scanner and combining their advantages could make up the defects of single sensor and promote adaptability to complex environment;on the other hand,it is efficient to improve the localization accuracy.This algorithm employs optic to track the laser point projections to obtain their depths.Then PnP computing,Kalman Filting and nonlinear optimization are employed to estimate UAV pose accurately.(5)Algorithms implementation in ROSWith the purpose that enhancing the code reusability in Robot research,Robot Operating System(ROS)was designed,which catches much attention from researchers and develops fast.With the basis with ROS,a Gazebo-based simulator was built to validate the localization algorithms and real experiments have been done showing great localization accuracy,robustness and real-time capability.The four algorithms designed by this paper have been packaged in ROS and published,including Chan-Vese model-based UAV detection package,Kalman Filter-based UAV localization package,onboard multi-cameras/laser extrinsic calibration package and onboard multiple sensors fusion-based UAV localization package.