Research On Key Technologies Of Autonomous Driving System For Load Haul Dump Based On LiDAR

The load-haul-dump(LHD)machine is one of the most important trackless equipment in underground mines,playing a vital role in the process of ore loading and transportation.The automation of equipment,such as LHD,is the foundation of mine intelligence and an inevitable trend in the development of the mining industry.The automatic driving system is the key technology for achieving LHD automation.Although China has made great progress in mine automation research in recent years,and remote operation of LHD has been realized,the industrial application of LHD automatic driving has not yet been achieved due to immature technology.This paper takes LHD as the research object,combines it with the characteristics of the underground mine roadway environment,and carries out in-depth research on the key technologies of automatic driving systems,proposing a complete system for automatic driving of LHD.In the process of key technology research,the validation method that combines simulation and actual data is adopted,which lays a solid foundation for the industrial application of LHD automatic driving.Firstly,an automatic driving simulation platform for LHD was established,laying the foundation for the validation of subsequent key technologies.Based on the geometric structure of LHD,the kinematic model of LHD was derived,and the geometric models of each component were established.They were coupled and connected to form the geometric model of LHD.A tunnel modeling method was proposed to simulate the underground tunnel environment.The performance of the LHD simulation system was analyzed through experiments,and the results showed that the simulation system could be used for algorithm testing.Secondly,the construction and positioning methods of underground environment maps were studied.The performance of commonly used 2D map construction methods was tested through the automatic driving simulation platform,and a method suitable for constructing maps in underground environments was selected.The optimal positioning method under pose constraints in underground restricted spaces was proposed.Submaps were partitioned in the global map to extract features and construct a map feature library.Real-time features were extracted based on real-time Li DAR data,and the rough positioning result was obtained by searching and matching in the map feature library.Precise positioning was achieved based on the rough positioning result.The performance of the proposed positioning method was tested on the simulation platform,and the problem of difficult determination of initial pose was solved.The paper further proposes a local and global path planning method based on a single-line Li DAR.After visualizing the real-time Li DAR data,the skeleton of the image is extracted by a thinning algorithm.The centerline of the current roadway is extracted from the skeleton image,and the local path is obtained by smoothing the centerline using methods such as Bezier curves.The local path planning method can be used for the reactive navigation of the LHD.The skeleton of the global map is obtained by extracting the skeleton of the entire working roadway using a thinning algorithm.The global path is obtained by smoothing the centerline using Bezier curves.The global path planning method can be used for the absolute navigation of the LHD.The performance of the path planning method was verified using datasets collected from two mines.The results show that the local path planning method is accurate and efficient,and can solve the problem that existing methods cannot be used for turning.The three different control methods for the shovel loader path tracking were designed as follows:(1)a path tracking method based on preview incremental PID control was designed,which uses lateral error as the input of the PID controller to control the angular velocity;(2)a path tracking method based on linear model predictive control was designed,which uses the shovel loader state variables as the input of the controller to control the velocity and angular velocity;(3)a path tracking method based on nonlinear model predictive control was designed,which is consistent with the model predictive control and realizes the control of velocity and angular velocity.Multiple control experiments were conducted on the automatic driving simulation platform,and the results showed that the preview incremental PID control method is difficult to tune,the overall performance of the linear model predictive control method is good,and the nonlinear model predictive control method is suitable for low-speed situations.Finally,the organizational relationships between the key technologies were integrated,and the organizational structure of the LHD autonomous driving system was proposed.An industrial experiment of the LHD autonomous driving system was conducted,which showed its feasibility,with stable driving at a linear speed of 1.8 m/s and a turning speed of 0.6m/s.The errors between reactive navigation and absolute navigation were compared,and the reasons for their errors were analyzed.In addition,the problems and shortcomings of the current LHD autonomous driving system were summarized,such as unreasonable speed planning and delayed execution of angular velocity,which require further in-depth research in future work.Figures 166,Tables 25,References 162.