Research On Autonomous Landing Of UAV Based On ROS

In recent years,benefited from the development of sensing technology and related material technology,drone-related sensors have been gradually miniaturized,which has greatly promoted the development of the civil drone industry.Take DJ-Innovations,a Chinese company in the field of consumer drones,as an example.Its annual sales in 2018 reached 17.5 billion,accompanied by the continuous development of international and domestic open source flight control.At the same time,along with the continuous development of international and domestic open source flight control,the threshold for the development of drones has been greatly reduced compared to a few years ago.This has led many companies and institutions at home and abroad to start mass manufacturing or purchasing and invest them in various fields,such as commercial logistics drones.police special drones,geological exploration and measurement drones,etc.have also been widely applied to their related fields.With the continuous development of drone technology,the current development direction of drones is more inclined to the autonomous control of drones,that is,drones no longer rely on human control but rely on onboard computers,while drones land on their own,It is the key stage for the drone to control the flight autonomously.The realization of the multi-rotor UAV's independent precision landing process is of great significance to the logistics drone.This paper will introduce the ROS robot operating system to realize the UAV visual guidance landing process on the quadrotor UAV,and realize the Gazebo 3D simulation software simulation flight.Reduce development test costs.This design is mainly aimed at the problem of autonomous high-precision landing of logistics drones,and aims to use opencv image processing combined with robot operating system and visual guidance algorithm to achieve high-precision flight landing.The work involved in this article is as follows:(1)Through a large number of domestic and foreign UAV visual aid landing data,and comparing various target recognition and localization algorithms,an improved UAV autonomous landing visual guidance scheme was finally determined.The program will combine the high and low target recognition schemes to effectively solve the problem that the UAV cannot effectively identify the target at a certain height and the camera lens cannot completely cover the target.(2)Focusing on the visual guidance strategy of UAV,the pose estimation algorithm and UAV control algorithm in the process of UAV autonomous landing are introduced in detail,and the algorithm used in this design is compared with the classical algorithm in this field.(3)Researched the idea of functional design on ROS system and completed the program software design under this system,node the various functions of the drone and use the relevant ROS function package to simplify the development process and reduce the complexity of the whole system.(4)Using the robot 3D simulation software Gazebo provided by ROS,with its powerful physics engine and high-quality image rendering function,the whole simulation of the entire landing process is realized,and the various costs in the debugging development process are reduced.After completing the simulation experiment of the Gazebo platform,an outdoor flight experiment was carried out on the drone,and finally the experiment measured that the design can correctly match the target position at a height of 20 meters.The system can guide the UAV stably and reliably to achieve accurate landing with a target error of 30 cm.This design can provide a certain reference value for the logistics UAV technology.