Research On Autonomous Navigation Method Of Quadrotor Based On RGB-D Camera

Due to its excellent flexibility and maneuverability,quadrotors can replace human beings to carry out some special tasks in complex and dangerous environment,improving the work efficiency and reducing the risk of casualties.Recently,small drones with autonomous navigation capabilities have gradually become a new research hotspot in military and civil areas.However,the onboard processor resources of small drones are usually limited.At the same time,the commonly used localization and navigation algorithm requires high performance of the processor,so it is hard to realize real-time autonomous navigation on drones.This paper attaches importance to these key technologies,such as real-time pose estimation,three-dimensional map construction and autonomous trajectory planning for quadrotors.In view of the limited processor resources of quadrotor,the cost of largescale dense map construction,and the lack of real-time and security of trajectory planning,an autonomous navigation method based on RGB-D camera is proposed and applied to the self-built quadrotor platform.The main work and achievements of this article are as follows:(1)Research on pose estimation and map reconstruction method: In order to solve the problem of scale uncertainty in the monocular visual SLAM(Simultaneous Localization and Mapping)and the resource-consuming problem of large-scale dense map reconstruction,this paper uses RGB-D camera to directly obtain the real scale information,and fuses the IMU data to improve the accuracy of pose estimation,and introduces the memory management mechanism of RTAB-Map(Real-Time Appearance-Based Mapping)to achieve real-time accurate pose estimation and largescale dense map reconstruction of quadrotors;(2)Research on quadrotor trajectory planning method: A real-time quadrotor trajectory planning method based on key waypoints is proposed,which improves the heuristic function of the A* algorithm.It uses gradient information to screen key waypoints,and optimizes collision cost function and explicitly introduces the feasible constraints of the quadrotor to construct a quadratic programming model,which can optimize and make the trajectory more suitable for quadrotor’s flight,meanwhile ensure the safety and real-time performance of the autonomous navigation of the quadrotor;(3)Experimental platform construction and method verification: Firstly,under the Linux system,the ROS(Robot Operating System)is deployed,and a simulation environment is constructed to preliminarily test the feasibility and stability of the SLAM algorithm,the trajectory planning algorithm and the whole system.Then,the hardware platform of quadrotor with pixhawk2 flight controller,Intel Real Sense RGBD camera and Intel NUC Mini computer as the important components is built independently to further verify the feasibility and effectiveness of the proposed method.