Research On Technologies Of Accurate Localization And Mapping For Underground Drilling Robot Application

With the rapid development of intelligent robot,self-driving,artificial intelligence and other related disciplines,the intelligence and robotization of coal mine equipment is the new development trend after the transformation of mechanization,electrification and automation,which will change the existing production mode of coal mine,bring great changes in productivity and safety for the development of coal mining industry.As one research of development and application for coal mine robots,the underground gas prevention and drilling robot is a fully autonomous robot that can realize the automatic gas pumping in underground working face and laneway without manual intervention,and has the functions of autonomous walking and automatic drilling.Accurate positioning and high-precision map construction in complex underground scenes are key technologies that support drilling robots to achieve autonomous walking.However,there is no effective solution yet.Coal mine environment is complex,the working conditions are bad,and there is no GPS in this area.The localization and mapping methods used on the ground can not be applied directly in this environment.The existing underground positioning technologies,such as dead reckoning,inertial navigation and positioning,RF tag positioning,video surveillance and other means,can not adapt to the drilling robot in various complex environments in the long-term large-scale positioning,can not provide accurate position estimation for autonomous navigation and automatic drilling operations.The existing surveying and mapping methods in underground coal mine based on survey equipment such as laser scanners and total stations are inefficient,and are not suitable for 3D map construction and rapid model update in a changing environment,which can not provide accurate and efficient prior maps for drilling robots.Therefore,in this paper,in view of the actual requirements of drilling robot,the two key technical issues of accurate localization and mapping are studied,including:In view of the actual requirment of walking autonomously to drilling robot,the working environment of drilling robot is analyzed,the functional requirements of drilling robot to achieve autonomous navigation are discussed,and the nature of the problems of localization and mapping of mobile robot are studied.The three-dimensional motion description of euclidean space and manifold space is analyzed.The state estimation solution based on filtering and optimization,and the mathematical description of sensor observation model are discussed.All these reserarch lay a theoretical foundation for proposing the localization and mapping methods for drilling robot application.Aiming at the lack of positioning system and pose estimation method that can provide reliable positioning services for mobile robots in underground coal mine,a method for constructing an underground pseudo-GPS positioning system based on EKF-UWB is proposed,and an optimal anchor node deployment strategy is proposed.Further,based on this positioning system,a 6-DOF pose estimation method based on ESKF to fuse positioning system and IMU observation is developed.The test results in simulation and real scenes show that the proposed positioning system based on EKF-UWB and pose estimation method based on ESKF-Fusion can achieve robust and accurate positioning for long-term running of underground robots,and at the same time estimate the extrinsic parameters between UWB and IMU,and the biases of the IMU online.In order to solve the problem of efficient point cloud map construction and localization in convertional unstructured underground environments,a laser-based 3D NDT-graph-SLAM method is proposed.According to the characteristics of the actual working conditions of the underground laneway,a pose graph optimization method based on the lidar odometry constraint factor,the plane feature constraint factor,and the loop constraint factor are designed to realize the online map construction and localization.Tests were conducted on public datasets and simulated underground laneways.The results show that the proposed method can achieve real-time high-precision map construction of ground and unstructured underground scenes,and can meet the application requirements of robots for efficient localization and mapping in underground unstructured environments.In order to achieve reliable positioning and high-precision map construction in uneven terrain and complex scenes in underground coal mine,a multi-sensor fusion method based on the tightly-coupled of LiDAR and IMU—LI-SLAM is proposed.In order to cope with working conditions such as rapid rotation and violent movement of the robot,IMU data is used to assist LiDAR for the correction of point cloud distortion.The LiDAR relative pose factor,IMU pre-integration factor,marginal prior factor and loop detection factor are designed.Based on the factor graph optimization framework,the function of tightly-coupled sensor information fusion is realized.A large number of tests in complex terrain scenes in the field show that the proposed method has strong adaptability to working conditions such as complex terrain and violent movement,and can meet the requirements of high-precision priori map construction and positioning in underground uneven terrain and complex scenes.Aiming at the fact that there is no positioning and map construction method that contains absolute geographic information and overcomes the problem of scene degradation in underground mines,LIU-SLAM,which is based on the fusion of LiDAR,IMU and UWB,is proposed.Using the tightly-coupled local lidar-inertial odometer provided by the LI-SLAM,the absolute position and distance observations provided by the UWB positioning system are further fused by the the global factor graph optimization.The field tests in the underground garage show that the proposed method achieves the precise positioning of local areas and long-distance large-scale scenes within the UWB signal coverage of the robot,and can realize the alignment of the point cloud map and the geographic coordinate system online,which improves the accuracy and robustness of laser-based localization and mapping.In order to verify the practicability and feasibility of the algorithms proposed in this paper,the positioning & navigation software and hardware system of drilling robot were developed,and the simulated coal mine laneway tests were carried out based on the coal mine rescue robot platform.The results show that the UWB positioning system based on the EKF-UWB method can achieve an average static absolute positioning accuracy of less than 10 cm in the coverage area of positioning system,which can meet the precise positioning requirements of the drilling robot in local areas,and can expand the application range by moving or deploying more UWB nodes.The proposed ESKF-Fusion and LIU-SLAM methods can both achieve accurate pose estimation of the robot in the local area within the UWB positioning system during continuous motion.LIU-SLAM is more robust and accurate,and achieves an average absolute positioning accuracy of less than 25 cm in these area.The LIU-SLAM method can also achieve high-precision map construction and precise positioning in a large-scale scene covered by UWB signals,obtain good map consistency and local modeling effects,achieving average absolute positioning accuracy within 25 cm,aligning the point cloud map and geographic coordinate system reliably,and has the best robustness for complex and degraded underground scenarios.Through the construction of the UWB positioning system,the ESKF-Fusion pose estimation method based on the fusion of UWB and IMU,the efficient map construction and positioning method based on 3D NDT-graph-SLAM,and the high-precision map construction and positioning method—LI-SLAM and LIU-SLAM based on the fusion of LiDAR,IMU and further UWB,the feasible solutions for the positioning and map construction of drilling robots are provided,laying the research foundation and application experience for the practical application of drilling robots in the next step.The paper has 116 pictures,34 tables,and 197 references.