Research On Aircraft Localization And Navigation Based On Vision And Inertia

With the development of science and technology,multi-rotor aircraft are widely used in surveying,rescue,aerial photography and agriculture.The application challenge of aircraft lies in how to efficiently and accurately perform user tasks and achieve autonomous navigation and the key technologies of autonomous navigation are positioning and motion planning..This thesis mainly focus on the visual inertial navigation positioning and the track generation algorithm,and applies the algorithm to the aircraft to complete the indoor autonomous navigation task.The main research contents of the thesis are:（1）For the positioning algorithm,a tightly coupled VIO positioning algorithm is designed.By fusing camera image data and IMU data,the state of the system is accurately estimated,and the camera pose is obtained in real time.In the front-end,the relative pose between frames is calculated using the epipolar geometry method for the image data;the relative pose within the frame interval is calculated using the pre-integration algorithm for the IMU data.In the back-end,the objective function is to minimize the re-projection error,and multiple key frames adopt a sliding window strategy to complete the pose optimization,and the pose is re-optimized in the loop-closure.The algorithm proposed in this thesis is evaluated in the data set environment.The results show that the proposed algorithm can perform the best in most datasets compared with ROVIO and VINSMONO,which verifies the accuracy of the algorithm.（2）Aiming at the motion planning,this thesis designs a trajectory generation algorithm for quadrotor aircraft.Firstly,the quadrotor dynamics model is introduced,and the differential flatness of the quadrotor is deduced to provide theoretical support for the generation of subsequent trajectories.Then,each trajectory is fitted with a high-order polynomial,and the objective function is to minimize the fourth-order derivative of the position.Equation constraints are established for the starting and ending points,and then an optimization model is built.In order to verify the proposed trajectory generation algorithm,a simulation environment was built under the robot operating system,and the trajectory of the quadrotor was drawn in real time,and several experiments were carried out to verify the trajectory generation algorithm.The experimental results show that the root mean square error of the simulation system is centimeter-level.The feasibility of the autonomous navigation system of the aircraft is proved.On the basis of verifying the feasibility of visual inertial odometry,trajectory generation algorithm and simulation system,in order to improve the frequency of positioning information of the odometer,the multi-sensor fusion framework is used to fuse the state estimation of the body IMU and odometer,and the positioning frequency is increased to 35 HZ,which is beneficial to motion control.In order to obtain good flight control performance,dynamic system identification is used to obtain roll,pitch and vertical velocity transfer functions,and model predictive control is used to obtain sixdegree-of-freedom precise control.Using the visual positioning system as the ground truth,the camera motion is first tracked under a small airborne computing platform,and the obtained trajectory error is at the centimeter level,which verifies the real-time and robustness of the algorithm proposed in this thesis under low indoor lighting conditions.The multi-waypoint flight test of the aircraft and the error evaluation of the system have been completed.Finally,compared with the performance of the same type of aircraft platform,the results show that the designed system error index is better than the control group.After many experiments,it has been proved that the system can run smoothly and accurately to complete the track task specified by the user.