Visual Servo Control Of Fixed-wing Unmanned Aerial Vehicle For Ground Target Tracking

Persistent robust tracking of ground target based on airborne sensors is one of the basic forms for the unmanned aerial vehicle(UAV)to perform earth observation mission.Besides,it is also the basis for subsequent target behavior(or intension)analysis and situational generation.Most of the traditional methods adopt the "Identification-PositioningControl" strategy,that is,after identifying the target,the position of it in the space can be determined by internal and external parameters of the airborne sensors,and then the tracking control law is designed based on the position.However,there usually exist big errors in the process of positioning,and the fixed-wing UAV is in high-speed motion all over the flight,which results in poor tracking performance of the UAV.Therefore,it is difficult to implement the long-term robust tracking of the target.For this problem,this dissertation studies the visual servo control law for target tracking without position calculation,based on the fixed-wing UAV and airborne vision sensor.After that,the "Identification-Control”integration is realized,the hardware-in-the-loop(HIL)simulation and real flight tests are conducted to verify the algorithms.The main work of the dissertation includes:(1)Proposing a visual servo control law for the fixed-wing UAV with a fixed camera,which can track stationary or slow moving target(speed is no more than 3m/s)without position calculation.When the camera is fixed on the UAV,the change of UAV attitude will affect the pose of camera,and then causes the target feature point to shift.Therefore,the dissertation proposes an "Ideal Camera" model to compensate the influence,which makes the pixel coordinates of the feature point unchanged in the "Ideal Camera" model even if the yaw and pitch angles of the UAV have changed.Next,the mapping relationship between the velocity of the feature point and that of the UAV is constructed based on the image Jacobian matrix.Furthermore,the least squares method is used to design the control law,and the stability of it is proved by Lyapunov method.The constructed HIL simulation environment is based on Gazebo and autopilot.The experiments verify the effectiveness of the control law for the tracking of stationary or slow moving target in the environment,under the circumstance that the flight speed is 16m/s.(2)Proposing a visual servo control law for the fixed-wing UAV with a pan-tilt camera,which can track medium-speed moving target(speed is between 3m/s and 6m/s)without position calculation.The “Ideal Camera”model proposed in the previous work has limited ability to compensate for the effect caused by target motion.When the target moves fast,it will be easily out of view.To solve the problem,the dissertation introduces a pan-tilt to enable the camera to yaw and pitch to some extent.Based on the "Ideal Camera" model,the "UAV + pan-tilt" motion model is considered,and the improved visual servo control law is designed to make the UAV robustly track not only the stationary or slow moving target,but also the medium-speed moving target.Besides,the stability of the servo control system is proved by Lyapunov method.The HIL simulation experiments verify the effectiveness of the control law for the tracking of moving target with a speed less than 6m/s,under the circumstance that the flight speed is 16m/s.Furthermore,the flight tests verify the effectiveness of the control law when the speed of UAV and target are 16m/s and 5.6m/s,respectively.(3)Proposing a strategy for the fixed-wing UAV with a pan-tilt camera to achieve long-term tracking of fast moving target(speed is more than 6m/s).Due to the limitation of mechanical characteristics for the pan-tilt,the adjustment of its attitude will soon reach saturation when the target speed is fast enough,which results in the loss of target in the image.To address the problem,the dissertation comprehensively considers the motion characteristics of the UAV and pan-tilt.After combining the estimation for the moving target,a state machine with three states and switching strategy among them are designed to further improve the tracking ability of fast moving target for the UAV.The simulation experiments verify that when the speed of target reaches 9m/s,the UAV can still achieve effective tracking of it with a flight speed of 16m/s.