Application And Research Of Multi-sensor Fusion In Indoor UAV Path Planning

Multi-sensor fusion refers to the alignment and coupling of data between binocular cameras and inertial navigation equipment.Due to the strong complementarity,high precision and robustness of camera vision and inertial data,it has gradually become a leading role in drones,robots and other systems.Common means of combining positioning and navigation on other platforms.However,the current mainstream methods have problems such as large amount of data calculation,difficult time stamp alignment,and difficulty in achieving real-time performance,which make this navigation method face many problems.Therefore,this paper conducts in-depth research on the existing problems,and designs a quadrotor UAV binocular vision-inertial navigation combined positioning and navigation system.This thesis firstly introduces the positioning principle of binocular vision-inertial navigation and the state estimation method of binocular vision-inertial navigation fusion,and focuses on the visual-inertial navigation pose estimation method based on nonlinear optimization method to provide accurate and accurate information for subsequent path planning.Pose and environment map information.Then,the key technologies are studied,including the fourth-order Runge-Kutta integration of IMU and the back-end nonlinear tightly coupled least squares optimization.While considering the accuracy and speed,a fourth-order Runge-Kutta integral is proposed to complete the discrete integration of IMU data.In the nonlinear optimization part,a tightly coupled data fusion method of least squares is proposed to complete the processing of visual inertial navigation data,and a variety of constraints are added to make the pose estimation more robust.Secondly,the ESDF navigation map with Euclidean distance information is constructed by using the binocular dense point cloud information.On this basis,the improved A*algorithm combining local path planning and global path planning is used to plan a feasible path.The smooth constraint performs B-spline optimization on the entire trajectory,so as to achieve the purpose of optimizing the trajectory.Finally,in order to verify the design proposed in this thesis,this thesis designs a set of four-rotor UAV autonomous positioning and navigation system based on binocular visual inertial navigation from the hardware composition and software architecture,which can realize real-time accurate positioning and path planning.In the analysis and verification part of the experimental results,the combined positioning and navigation performance of the system is evaluated from the data set simulation and the actual indoor environment test,which verifies the validity and reliability of the UAV system in this thesis.