Research And Implementation Of Indoor Autonomous Flight System For Quadrotor UAV Based On Multi-sensor Fusion

In modern society,the demand for UAV is increasing continuously,because it can help human to complete some difficult or dangerous tasks,such as farmland monitoring,aerial photogrammetry,route inspection and so on.In recent years,the quadrotor UAV system,which can realize autonomous flight in unknown environment,has become a new research hotspot.Although,it has made good progress in the sensor accuracy,simultaneous localization and mapping technology and path planning technology.However,due to the complexity of autonomous navigation system of quadrotor UAV,the autonomous navigation technology needs further research and improvement.Based on the above background,this paper improves the visual-inertial SLAM technology and path planning technology,and designs a complete set of quadrotor UAV system which can realize autonomous navigation indoors.Finally,the software design and hardware construction of the autonomous navigation system are completed,and the reliability and advantages of the system are verified by simulation experiments and actual flight experiments.Firstly,the frame structure and motion principle of the UAV,the imaging model and calibration of the camera are introduced.After selecting the hardware of the system,the paper uses Realsense D435 i depth camera,UP Squared onboard computer,TFmini Plus lidar and Pixhawk flight control modules to build the hardware platform of the quadrotor UAV,and completes the design of the system software.Finally,a hardware-in-the-loop simulation environment is built based on the Gazebo platform to test the algorithm of the system.Secondly,the autonomous navigation algorithm of UAV based on visual-inertial fusion in this system is introduced in detail.In the visual-inertial mileage calculation method,the principle of the algorithm is introduced in detail from the aspects of visual front-end processing,IMU pre-integration,system initialization,nonlinear optimization,loop detection and pose map optimization.Finally,the attitude and position information of the UAV are obtained.In the dense mapping algorithm,the RGB-D camera mapping scheme is used to generate the point cloud map and octomap map of the surroundings;in the path planning algorithm,an improved RRT* algorithm is proposed.By improving the algorithm to generate new nodes of random length and adding dynamic constraints,the optimal path can be generated in real time on the octomap map.Finally,in order to verify the effectiveness of each algorithm in the autonomous navigation system,this paper designs many simulation and actual flight experiments,including indoor positioning experiments,dense mapping experiments,path planning experiments and test experiments in the real environment.The results indicate that the indoor autonomous navigation system of quadrotor UAV proposed in this paper can achieve good positioning,environmental mapping and path planning in indoor unknown environment,and has good real-time performance and robustness.