| Tunnels are a common way to increase the utilization rate of land resources.There is lack of proper maintenance during operation may lead to catastrophic consequences.The traditional inspection method has the disadvantages of heavy workload and high danger coefficient,which makes the inspection robot popular.Among them,the micro multi-rotor UAVs have the characteristics of light weight,small size,strong environmental resilience and so on,which has gradually attracted attention in indoor application fields such as tunnel inspection.However,there are many technical problems in tunnel inspection,such as autonomous navigation,which limit its rapid development.In response,Euclidean Signed Distance Fields(ESDF)is selected as the high-precision map for the autonomous navigation of the micro multi-rotor UAVs in tunnel,this paper focuses on the construction method of ESDF and the navigation planning based on ESDF.The specific work of the paper is as follows:Firstly,according to the inspection requirements of tunnel environment with limited global positioning system,the system scheme is designed.This system adopts the D435 as the environmental perception unit and the microcomputer Larkbox Pro as the airborne computing unit.In addition,the tunnel inspection UAV system is built based on the quadrotor UAV platform.Secondly,this paper studies UAV positioning and high-precision map construction based on depth camera in tunnel environment.Aiming at the problem that the uneven brightness of tunnel environment makes it difficult to extract image feature points,the adaptive threshold FAST(features from accelerated segment test)algorithm and Q-tree are used to improve the ORB(oriented fast and rotated brief)feature algorithm.The pose of the UAV is estimated in tunnel environment by feature matching,pose estimation and optimization.Based on this,the ESDF is constructed incrementally by fusing the real-time pose data of the camera and color point cloud data.In addition,a total of 30 mapping experiments were carried out to evaluate the accuracy of mapping.The experimental results show that the average error was within 0.62 m,and the accuracy of mapping reach decimeter level.Thirdly,the navigation planning method based on ESDF constructed is designed.Aiming at the extension of BRRT*(Bidirectional Rapidly Exploring Random Trees star)navigation algorithm is lack of teleonomy.APF-BRRT* method is designed by combining the gravitational field and repulsion field in the Artificial Potential Field.Then,the path smoothing based on polynomial and Local Continuous Optimization is adopted to smoothing,and the generated path is constrained and screened by attitude control of quadrotor UAV.In addition,the local online re-planning strategy is adopted to aiming at the possible changes in the environment and improving the ability of the UAV to reach the target point.Finally,experiments are designed to verify the feasibility and practicability of the research method.The experimental results of map construction to simulated tunnel environment under different lighting conditions show that the system has strong adaptability to the weak light environment.The map construction of simulated tunnel environment and tunnel environment and 40 navigation experiments show that: the success rate of APF-BRRT*algorithm is increased by about 5% compared with BRRT* algorithm,and the time consumption is reduced by 7.35%.So APF-BRRT* has better timeliness.At the same time,the feasibility of the method studied in this paper is verified.The inspection experiment of the real tunnel environment shows that the system has the ability of autonomous inspection and has a certain practicability. |
PDF Full Text Request