| The mine fill pipeline is a key facility used in the mining process to transport the fill material to the extraction area and does not allow the pipeline to fail or leak,requiring regular inspection of the mine fill pipeline.The traditional manual inspection method is not only time-consuming and labour-intensive,but also poses certain safety risks.The use of drones with cameras and other equipment for inspection of filling pipes allows for fast and accurate detection of leaks and fault information in filling pipes,while improving the safety of the inspection.As the filling pipe is installed on one of the side walls of the tunnel,the installation position has a fixed distance from the ground.Therefore,when conducting pipeline leak inspections,UAVs must fly in a certain fixed attitude and form a certain fixed angle and distance from the pipeline in order to complete the identification and inspection of pipeline leaks.It requires a precise navigation line and stable flight control to accomplish this task.Accurate acquisition of the navigation line for the UAV inspection flight and real-time flight control of the UAV along this navigation line are therefore key technologies for the inspection of filled pipelines.According to this need,a research has been carried out on the construction of maps and navigation in the underground tunnel of the UAV.The main contents are as follows:(1)For the problem of weak GPS signals and light in underground tunnels,where accurate maps cannot be obtained using a single sensor,3D Li DAR and IMU are used as sensors,and data fusion is carried out by iterative extended Kalman filtering.At the same time,in order to improve graph building efficiency and reduce memory consumption,the incremental KD-tree data structure was used in the algorithm.(2)The loop closure detection algorithm is improved.For the first time,a loop closure detection algorithm based on NDT-KD-ICP point cloud descriptors,which is introduced in loop closure detection,as well as in the back-end part,a factor map structure is built to optimize the UAV poses,resulting in a highly accurate point cloud map.(3)Improved navigation line extraction method.Precise navigation lines were obtained from the point cloud map.According to the obtained point cloud map,the noise and ground points are eliminated by using SOR filtering algorithm and fabric simulation filtering algorithm;then the map is projected in both directions,and the boundary of the point cloud map in two planes is extracted by the latitude and longitude scanning algorithm,as well as the boundary line is fitted;in order to obtain the median line,sampling equally spaced on the fitted boundary and calculating the midpoint of its normal intersection line;finally,the center axis is rounded to obtain a precise navigation line suitable for smooth flight control.(4)The UAV dynamics model is established,and a string-level PID control system based on the outer-loop position controller with the inner-loop attitude controller is built.Also,in order to improve the UAV’s ability to resist external disturbances during flight,the angular velocity controller is introduced in the inner-loop attitude controller,while a series-level fuzzy PID controller is constructed on this basis,and the stability of this control system is verified through simulation.Finally,the simulation experiment of flight control of patrol line is conducted on the extracted navigation line,and the test results show that the control system built achieves the goal of smooth flight control along the navigation line,as well as meeting the needs of UAV inspection in the underground roadways. |
