A method and device for updating a point cloud map of automatic operation of a shovel, an electronic device, and a storage medium

Abstract

The application provides a method and device for automatic operation point cloud map updating of a shovel, electronic equipment and a storage medium. Based on the spatial characteristics of the "front moving and rear static" of the mining working face, the front perception and the rear positioning are spatially decoupled. The rear point cloud and the IMU are used to determine a stable rear pose reference, and the front pose is calculated based on the hinge angle. When the front pose and the point cloud data are combined to determine the trigger update, the effective point cloud on both sides of the roadway obtained by dimension reduction projection and RANSAC fitting in the front point cloud is extracted, the optimal transformation matrix is obtained by registering the current map, the front point cloud is transformed to the map coordinate system based on the matrix and combined, and the out-of-boundary points are removed through geometric determination to obtain the updated point cloud map and continue the cycle as the next time reference. The application separates the positioning reference from the dynamic boundary, effectively avoids the positioning drift, and ensures the accuracy and spatial consistency of the map updating.

Description

Technical Field

[0001] This application relates to the field of loader technology, and more specifically, to a method, apparatus, electronic device, and storage medium for updating an automatic operation point cloud map of a loader. Background Technology

[0002] Loaders are core equipment in mines, and their intelligent operation relies on high-precision real-time positioning. Due to GPS signal rejection in underground mines, point cloud map-based positioning has become the mainstream technology. Existing solutions typically involve first constructing an initial static map, and then matching the map to achieve vehicle positioning. This method can achieve good positioning accuracy in static tunnel environments.

[0003] However, in longwall mining operations, as mining progresses, the ore-exit boundary of the longwall face recedes, and the working environment continuously changes. The original static map cannot match the actual environmental changes, directly leading to point cloud registration errors and consequently, positioning failure. Most existing dynamic update solutions directly update the entire point cloud map globally, which is not only computationally intensive and difficult to meet the real-time requirements of automated loader operations, but also prone to point cloud registration interference from dynamically changing environments, causing positioning reference drift and ultimately affecting the accuracy of map updates. This makes them unsuitable for the continuous automated operation of loaders during mining longwall mining. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, apparatus, electronic device and storage medium for updating the point cloud map of the automatic operation of a loader, so as to overcome the problems in the prior art.

[0005] In a first aspect, embodiments of this application provide a method for updating a point cloud map of an automated operation of a loader, the method comprising: During the operation of the loader, the point cloud map at the current moment is acquired, and the front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle body articulation angle data are acquired in real time. Among them, the front point cloud data is used to perceive the dynamic retreating mining boundary in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind. Based on the point cloud map at the current moment, the rear point cloud data, and the IMU inertial attitude data, the rear orientation of the loader is determined; and based on the rear orientation and the front and rear vehicle body hinge angle data, the front orientation of the loader is calculated. When the front pose and the front point cloud data meet the map update triggering conditions, the effective point cloud in the front point cloud data is registered with the point cloud map at the current time to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after the front point cloud data is extracted by dimensionality reduction projection and fitting a straight line. Based on the optimal pose transformation matrix, the front point cloud data is transformed to the map coordinate system and merged with the point cloud map at the current moment. The merged point cloud is then clipped and removed from the boundary out-of-bounds points to obtain the updated point cloud map. The updated point cloud map is used as the point cloud map at the next moment.

[0006] In some technical solutions of this application, the above method also includes: When the loader is started for the first time, initial static point cloud data and map coordinate system parameters are acquired; wherein, the initial static point cloud data is the static structure point cloud of the roadway obtained by scanning before the loader is in operation; An initial point cloud map is constructed based on the initial static point cloud data and the map coordinate system parameters; wherein, the initial point cloud map serves as the initial reference benchmark for the point cloud map at the current moment.

[0007] In some technical solutions of this application, determining the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data includes: The rear point cloud data is registered with the stable rear tunnel structure in the point cloud map at the current moment, and the pose is corrected by combining the IMU inertial attitude data to determine the rear pose of the loader.

[0008] In some technical solutions of this application, the calculation of the front orientation of the loader based on the rear orientation and the hinge angle data of the front and rear vehicle bodies includes: The front orientation of the loader is calculated based on the rear orientation, the hinge angle data of the front and rear vehicle bodies, and the preset coordinate transformation relationship between the front and rear coordinate systems.

[0009] In some technical solutions of this application, the above solutions determine whether the front pose and the front point cloud data meet the map update triggering conditions in the following ways: Calculate the distance to the obstacle in front corresponding to the front point cloud data; The map update triggering conditions are determined to be met when the distance to the obstacle in front is less than or equal to a set distance threshold, the obstacle in front does not coincide with the known boundary of the point cloud map at the current moment, and the obstacle in front continues to exist within a preset time; otherwise, the map update triggering conditions are determined not to be met.

[0010] In some technical solutions of this application, the above-mentioned registration of the effective point cloud with the point cloud map at the current time to obtain the optimal pose transformation matrix includes: Using the aforementioned front pose as the initial pose, the maximum number of iterations, convergence threshold, and interior point threshold are set. The effective point cloud is iteratively registered with the point cloud on the corresponding alleyway side in the point cloud map at the current moment; In each iteration, for each valid point after transformation, the matching point with the smallest distance is found in the point cloud map, and the sum of squared errors of all matching point pairs is calculated; the absolute value of the difference between the sum of squared errors of two adjacent iterations is taken as the error change. When the error change is less than the convergence threshold, the iteration stops, and the current pose transformation matrix is ​​taken as the optimal pose transformation matrix.

[0011] In some technical solutions of this application, the above-mentioned out-of-bounds point clipping and removal of the merged point cloud includes: The RANSAC algorithm is used to fit the current boundary line to the newly added part of the merged point cloud; Calculate the intersection points of the current boundary line and the original left and right side line segments of the roadway to obtain the intersection points of the left and right boundaries; Calculate the centroid of the newly added point cloud and determine the forward reference vector pointing from the acquisition position of the front point cloud data to the centroid; For each point in the merged point cloud, calculate its current point vector relative to the acquisition position of the preceding point cloud data, and determine the angle between the current point vector and the forward reference vector. If the included angle exceeds the preset range, the point is retained; if the included angle is within the preset range and the distance from the point to the acquisition position of the front point cloud data is greater than the distance from the intersection of the corresponding side boundary to the acquisition position of the front point cloud data, the point is determined to be located outside the boundary and is cut off.

[0012] Secondly, embodiments of this application provide an apparatus for updating the point cloud map of an automated loader operation, the apparatus comprising: The data acquisition module is used to acquire the point cloud map at the current moment during the operation of the loader, and to acquire front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle body articulation angle data in real time; wherein, the front point cloud data is used to perceive the dynamic retreating mining boundary in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind. The positioning module is used to determine the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data; and to calculate the front orientation of the loader based on the rear orientation and the front and rear vehicle body hinge angle data. The registration module is used to register the effective point cloud in the front point cloud data with the point cloud map at the current time when the front pose and the front point cloud data meet the map update triggering conditions, so as to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after the front point cloud data is extracted by dimensionality reduction projection and fitting a straight line. The update module is used to transform the front point cloud data to the map coordinate system according to the optimal pose transformation matrix, merge it with the point cloud map at the current time, and perform boundary out-of-bounds point clipping and removal on the merged point cloud to obtain the updated point cloud map; the updated point cloud map is used as the point cloud map at the next time step.

[0013] Thirdly, embodiments of this application provide an electronic device, a processor, a memory, and a bus. The memory stores machine instructions executed by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine instructions are executed by the processor, the steps of the above-described method for updating the automatic operation point cloud map of the shovel loader are performed.

[0014] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when run by a processor, executes the steps of the above-described method for updating the automatic operation point cloud map of a shovel loader.

[0015] The technical solutions provided by the embodiments of this application may include the following beneficial effects: This application discloses a method for updating the point cloud map during the automatic operation of a loader, comprising: acquiring the point cloud map at the current moment during the operation of the loader, and acquiring front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle body hinge angle data in real time; wherein, the front point cloud data is used to perceive the ore extraction boundary of the dynamic retreat in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind; determining the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data; and calculating the front orientation of the loader based on the rear orientation and the front and rear vehicle body hinge angle data; when When the front pose and the front point cloud data meet the map update triggering conditions, the effective point cloud in the front point cloud data is registered with the point cloud map at the current time to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after dimensionality reduction projection and straight line fitting in the front point cloud data; according to the optimal pose transformation matrix, the front point cloud data is transformed to the map coordinate system and merged with the point cloud map at the current time, and the merged point cloud is subjected to boundary out-of-bounds clipping and culling to obtain the updated point cloud map; the updated point cloud map is used as the point cloud map at the next time.

[0016] This application decouples the front point cloud data from the rear point cloud data, using them respectively to perceive the dynamically retreating ore extraction boundary and the stable rear roadway structure. The stable rear roadway structure, combined with IMU inertial attitude data, provides an absolute positioning reference. The front pose is then calculated using the hinge angle, effectively avoiding positioning drift caused by interference from the dynamic environment ahead. Furthermore, this application only updates the map when the map update trigger condition is met, avoiding the large computational load introduced by frequent global updates and improving real-time response capabilities during automated operations. When a map update is triggered, effective point clouds located on the straight segments on the left and right sides of the roadway are selected for registration after dimensionality reduction projection and straight line fitting. This eliminates the interference of dynamic boundary point clouds on the registration process and ensures the accuracy of the pose transformation matrix during map updates. By transforming the previous point cloud data to the map coordinate system and merging it with the point cloud map at the current moment, and clipping and removing points outside the boundary, the spatial consistency and boundary accuracy of the updated map are effectively guaranteed. Moreover, the updated map is automatically used as the current map at the next moment, realizing the continuous dynamic update and accurate matching of the point cloud map during the continuous mining operation of the loader.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a method for updating a point cloud map for automated operation of a loader, as provided in an embodiment of this application, is shown. Figure 2 A schematic diagram of a hardware device for implementing the method of this application is shown in an embodiment of this application; Figure 3 This illustration shows a data interaction diagram provided by an embodiment of this application; Figure 4 A schematic diagram of an apparatus for automatically updating point cloud maps of a loader operation, as provided in an embodiment of this application, is shown. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0021] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0023] Loaders are core equipment in mines, and their intelligent operation relies on high-precision real-time positioning. Due to GPS signal rejection in underground mines, point cloud map-based positioning has become the mainstream technology. Existing solutions typically involve first constructing an initial static map, and then matching the map to achieve vehicle positioning. This method can achieve good positioning accuracy in static tunnel environments.

[0024] However, in longwall mining operations, as mining progresses, the ore-exit boundary of the longwall face recedes, and the working environment continuously changes. The original static map cannot match the actual environmental changes, directly leading to point cloud registration errors and consequently, positioning failure. Most existing dynamic update solutions directly update the entire point cloud map globally, which is not only computationally intensive and difficult to meet the real-time requirements of automated loader operations, but also prone to point cloud registration interference from dynamically changing environments, causing positioning reference drift and ultimately affecting the accuracy of map updates. This makes them unsuitable for the continuous automated operation of loaders during mining longwall mining.

[0025] Based on this, this application provides a method, apparatus, electronic device, and storage medium for updating the point cloud map of an automated loader operation, which will be described below through embodiments.

[0026] Figure 1 The diagram illustrates a flowchart of a method for updating a point cloud map for automated operation of a loader, as provided in an embodiment of this application. The method includes steps S101-S104; specifically: S101. During the operation of the loader, the point cloud map at the current moment is acquired, and the front point cloud data, rear point cloud data, IMU inertial attitude data and front and rear vehicle body hinge angle data are acquired in real time; wherein, the front point cloud data is used to perceive the ore exit boundary of the dynamic retreat in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind. S102. Determine the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data; and calculate the front orientation of the loader based on the rear orientation and the front and rear vehicle body hinge angle data. S103. When the front pose and the front point cloud data meet the map update triggering conditions, the effective point cloud in the front point cloud data is registered with the point cloud map at the current time to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after the front point cloud data is extracted by dimensionality reduction projection and fitting a straight line. S104. Based on the optimal pose transformation matrix, transform the front point cloud data to the map coordinate system and merge it with the point cloud map at the current moment. Perform boundary out-of-bounds point clipping and removal on the merged point cloud to obtain the updated point cloud map; the updated point cloud map is used as the point cloud map at the next moment.

[0027] This application decouples the front point cloud data from the rear point cloud data, using them respectively to perceive the dynamically retreating ore extraction boundary and the stable rear roadway structure. The stable rear roadway structure, combined with IMU inertial attitude data, provides an absolute positioning reference. The front pose is then calculated using the hinge angle, effectively avoiding positioning drift caused by interference from the dynamic environment ahead. Furthermore, this application only updates the map when the map update trigger condition is met, avoiding the large computational load introduced by frequent global updates and improving real-time response capabilities during automated operations. When a map update is triggered, effective point clouds located on the straight segments on the left and right sides of the roadway are selected for registration after dimensionality reduction projection and straight line fitting. This eliminates the interference of dynamic boundary point clouds on the registration process and ensures the accuracy of the pose transformation matrix during map updates. By transforming the previous point cloud data to the map coordinate system and merging it with the point cloud map at the current moment, and clipping and removing points outside the boundary, the spatial consistency and boundary accuracy of the updated map are effectively guaranteed. Moreover, the updated map is automatically used as the current map at the next moment, realizing the continuous dynamic update and accurate matching of the point cloud map during the continuous mining operation of the loader.

[0028] The following describes some embodiments of this application in detail. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0029] Before introducing the method of this application, let's first introduce the hardware device for implementing the method of this application, such as... Figure 2 As shown, the hardware includes a front lidar 6, a rear lidar 4, an IMU inertial measurement unit 5, and an articulated sensor 7, all mounted on the loader. The loader operates within a spatial environment defined by the unloading point 1 and the roadway boundary 2. The ore discharge boundary 3 in front of the loader dynamically recedes as the mining process progresses, creating a unique "front-moving, rear-static" characteristic of mining operations. The rear lidar 4 is positioned at the rear of the loader to sense the stable roadway structure behind it. The IMU inertial measurement unit 5 works in conjunction with the rear lidar to provide the loader with the data needed for rear orientation calculations. The front lidar 6 is positioned at the front of the loader to detect the dynamically receding ore discharge boundary in real time. The articulated sensor 7 is located at the hinge point between the front and rear bodies of the loader to acquire the angle data between the front and rear bodies.

[0030] like Figure 3As shown, the hardware device also includes a control module, which is communicatively connected to the front lidar, the rear lidar, the IMU inertial measurement unit, and the articulated sensor. This control module receives real-time data collected by the sensors and executes the steps described in this application. Internally, the control module includes a map initialization module, a positioning module, and a map update module, which are used to construct the initial point cloud map, calculate the pose of the front and rear parts of the loader, and dynamically update the point cloud map in real time, respectively.

[0031] Based on the aforementioned hardware, this application provides a method for updating the point cloud map of an automatic operation of a loader. Before executing the cyclic process of positioning and map updating, the method first determines whether the loader is starting up for the first time.

[0032] If it's the first startup, a map initialization step is executed first. This involves acquiring initial static point cloud data and map coordinate system parameters when the loader starts for the first time. The initial static point cloud data is the static structural point cloud of the roadway obtained by scanning before the loader begins operation. In other words, it's the complete structural information of the roadway collected by a static scanning device (e.g., a scanner) before the working environment is disturbed by mining. This data represents the geometric shape of the underground roadway in its initial state and serves as the most original static reference for map construction. The map coordinate system parameters include the basic configuration information of the map coordinate system and the coordinate transformation relationship between the scanning device's coordinate system and the map coordinate system. This ensures that point cloud data from different sources can be expressed and processed under a unified global coordinate system. Based on the acquired initial static point cloud data and map coordinate system parameters, an initial point cloud map can be constructed. This initial point cloud map fully reflects the static spatial structure of the mining face before the loader's operation. Essentially, it's a fixed reference base map that remains unchanged throughout the subsequent operations, serving only as an absolute position reference for later positioning. The point cloud map used in the actual positioning and update process is a dynamic variable. Specifically, it represents the set of map data referenced at the current moment during the positioning and map update cycle. When the loader starts for the first time, the initial point cloud map is assigned to the current point cloud map as its initial reference base. This ensures that subsequent positioning and update cycles can begin on a stable and accurate map foundation, preventing positioning failures or map construction chaos due to a lack of an initial map. Subsequently, as the loader continues to operate, the current point cloud map will iterate and upgrade according to subsequent update steps in this application, while the initial point cloud map continues to provide a stable long-term reference base for the system.

[0033] In practical implementation, the initialization module map construction formula is shown in formula (1), which constructs the initial map using the initial point cloud map scanned by the scanner, the scanner coordinate system, and the map coordinate system parameters. Where, Represents the initial point cloud map; This represents the initial static point cloud scanned by the scanner; This indicates the basic parameters of the map coordinate system and the conversion relationship between the scanner and the map coordinate system.

[0034] (1) If the loader is not being started for the first time during operation, the initialization step is skipped, and the subsequent positioning and update cycle begins. First, the core input data for the subsequent positioning and map update process needs to be acquired, specifically including the current point cloud map, as well as real-time collected front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle hinge angle data.

[0035] Here, the point cloud map at the current moment refers to the set of map data that the system is actually referencing at the current moment. When the loader starts for the first time, this map is the result of the initial point cloud map assignment. In subsequent continuous operations, it is determined by the map after the previous round of updates. It is a dynamic benchmark that supports the operation of the positioning module and the map update module.

[0036] Front point cloud data refers to the spatial point set generated by the LiDAR deployed in front of the loader in real time. Its core function is to perceive and track the ore boundary in front of the loader as it retreats due to the mining process. The dynamic changes of this boundary are obvious and drastic, and it belongs to the dynamic environment part that the system needs to perceive and include in map expansion.

[0037] Rear point cloud data refers to the spatial point set generated in real time by the lidar deployed at the rear of the loader. Its detection object is the stable roadway structure behind the loader. This structure remains basically static throughout the entire mining process and has extremely high spatial stability. Therefore, it can provide a highly reliable spatial feature reference for the system and is the static environment part used to lock the absolute positioning benchmark of the system.

[0038] IMU inertial attitude data refers to the rear attitude information of the loader (including three-dimensional acceleration and angular velocity) output in real time by the inertial measurement unit. It is used to assist in constraining and correcting the position and angle during the rear attitude calculation process, so as to improve the robustness of positioning under severe vibration or temporary obstruction conditions.

[0039] The front and rear body articulation angle data refers to the included angle value obtained in real time by the articulation sensor set at the front and rear body articulation. This angle directly reflects the instantaneous turning state of the vehicle during operation. Since the front and rear of the loader are based on the same body but belong to different coordinate systems, this angle data becomes the key link connecting the front posture and the rear posture calculation.

[0040] The aforementioned data are collected in real time by corresponding sensors at preset frequencies and uniformly received and timestamped by the control module, serving as the basic input for subsequent positioning calculations and map updates. The core concept supporting the overall technical solution of this application is the spatial decoupling of front and rear point cloud data. The rear point cloud data and IMU inertial attitude data jointly lock an absolutely stable positioning benchmark, while the front point cloud data specifically serves to determine map update trigger conditions and extract valid point clouds. Both perform their respective functions while cooperating to achieve a balance between robust positioning and accurate map updates for the loader during dynamic backfilling operations.

[0041] After obtaining the aforementioned real-time data, the system first enters the positioning calculation stage, the core of which lies in establishing the absolute positioning benchmark of the whole machine in the global map based on the stable tunnel structure behind.

[0042] Specifically, the positioning calculation is divided into two closely linked stages: rear position pose determination and front and rear position pose estimation. First, the system registers the rear point cloud data collected in real time by the rear lidar with the corresponding stable roadway structure in the current point cloud map. During this process, due to the severe vibration and possible brief field-of-view obstruction that may occur when the loader is operating underground, relying solely on the geometric features of the point cloud for registration can easily lead to errors caused by local deviations or frame drops. Therefore, the system simultaneously introduces the inertial attitude data output by the IMU inertial measurement unit and uses sensor fusion to constrain and correct the pose of the registration result to effectively suppress the interference of vibration noise and cumulative errors, thereby obtaining a high-precision rear position pose of the loader in the current global map coordinate system. Here, the rear position pose refers to the six-degree-of-freedom spatial position and attitude angle of the loader's rear body in the global coordinate system. It is the absolute reference benchmark for all subsequent spatial calculations and positioning judgments of the entire system, and it always plays the role of "anchoring" the coordinates of the entire machine throughout the entire mining operation.

[0043] After accurately determining the rear orientation, since the loader is an articulated vehicle, its front and rear bodies are not rigidly fixed, but mechanically connected by an articulation mechanism. Furthermore, the angle between the two parts will continuously change during turning and driving. Therefore, the real-time orientation of the front body cannot be directly equated to the rear orientation. A special coordinate transformation must be performed based on the rear orientation and the articulation angle.

[0044] Specifically, the system reads the angle between the front and rear vehicle bodies obtained by the articulated sensors in real time, and combines it with the rigid transformation relationship between the front coordinate system and the rear coordinate system that has been pre-calibrated and stored in the control module (this transformation relationship is determined by the mechanical structure dimensions and installation position of the front and rear vehicle bodies of the loader, and is a fixed parameter within the system), and transforms the determined rear posture from the rear coordinate system to the front coordinate system step by step, thereby calculating the front posture of the loader at the current moment.

[0045] At this point, the system has obtained both the rear pose and the front pose. The two are independent of each other but also related. The rear pose is responsible for locking an absolutely stable spatial position, while the front pose accurately reflects the spatial state of the front end at the current moment. Together, they constitute the pose input basis for the positioning and subsequent map updates in this application, providing a precise spatial reference for determining the map update trigger conditions.

[0046] In practical implementation, the calculation formula is shown in formula (2). Where, Indicates the rear position; This represents the point cloud map at the current moment; if it is in the initial state, then... = ; This indicates the real-time point cloud of the rear lidar; This represents the inertial attitude data acquired by the IMU; Indicates the front posture; This represents the angle data from the hinged sensor. This indicates the transformation relationship between the front and rear coordinate systems.

[0047]

[0048] (2) After obtaining the front pose and front point cloud data, it is necessary to determine whether they meet the map update trigger conditions to decide whether to initiate the map update process. The map update trigger conditions are determined in the following way: First, the system calculates the spatial distance between obstacles in front of the loader and the front sensors of the loader based on the point cloud data collected by the front LiDAR. Here, the obstacle distance refers to the distance value corresponding to the nearest obstacle point in the point cloud set detected by the front LiDAR to the loader. When the obstacle distance is greater than a preset distance threshold, it indicates that there is still enough working space in front of the loader and the boundary area that needs to be updated has not been reached. The system determines that the map update triggering condition is not met, does not trigger a map update, and continues to execute the current operation process.

[0049] When the distance to the obstacle ahead is less than or equal to the set distance threshold, it indicates that the loader has approached or reached a location where a boundary change may occur. At this point, the system needs to further determine whether the obstacle coincides with the known boundary in the current point cloud map. If the obstacle is exactly located at the geometric position of the known boundary in the current point cloud map, i.e., the two coincide, it means that the obstacle is a re-probing of a boundary already recorded in the map, not a new boundary change, and the system determines that the map update trigger condition is not met. If the obstacle does not coincide with the known boundary in the current point cloud map, it means that the obstacle may be a newly exposed spatial location due to the further retreat of the mining boundary. At this point, the system cannot make a final judgment immediately, but needs to wait for a preset time and continuously monitor whether the obstacle continues to exist within the preset time.

[0050] If the obstacle disappears within a preset time (e.g., due to the movement of dynamic obstacles or the removal of temporary interference factors), it indicates that the obstacle is not a stable new boundary, and the system determines that the map update trigger condition is not met. If the obstacle persists for the preset time, it indicates that the obstacle has become a stable part of the new mining boundary, and the system determines that the map update trigger condition is met, and the subsequent map update process should be initiated. The above three-layer judgment logic is progressive, effectively filtering out dynamic interference and temporary misjudgments, ensuring the accuracy and reliability of the map update trigger condition.

[0051] Once the three-layer judgment logic confirms that the map update triggering conditions are met, the system proceeds to the substantive map update operation process. Before executing the map update, the system does not directly use the original front point cloud data collected by the front lidar for full registration. Instead, it first needs to extract effective point clouds that can stably reflect the spatial structure of the alleyway from the original front point cloud data as the data basis for registration.

[0052] Since the original front point cloud contains both static structural point clouds reflecting the stable wall features on both sides of the roadway, and a large number of dynamic point clouds from the dynamic exit boundary of the mine, as well as environmental noise, if the complete front point cloud containing dynamic noise is directly registered with the map, the drastic change of the exit boundary will cause the registration algorithm to produce a large spatial deviation when searching for geometric correspondence, which will seriously interfere with the convergence direction of the registration and may even lead to errors in the calculation of the pose transformation matrix.

[0053] Therefore, this application employs a feature extraction method for effective point clouds: First, the 3D point cloud data collected by the front lidar is vertically projected onto a 2D horizontal plane, ignoring the dimensional information in the height direction and retaining only the planar distribution structure of the point cloud in the horizontal direction; next, on this 2D projection plane, the RANSAC algorithm is used to perform straight-line fitting on the point cloud sets on the left and right sides of the tunnel to extract the stable boundary line segments on the left and right sides; finally, from the original 3D point cloud set, only the spatial points falling on these two fitted line segments are extracted as effective point clouds. This effective point cloud set corresponds only to the stable walls on both sides of the tunnel in spatial location, successfully filtering out the dynamic retreat area of ​​the front ore extraction boundary and random interference points, providing stable and reliable geometric features for subsequent registration, which not only improves the convergence efficiency of registration but also fundamentally eliminates the technical risk of dynamic boundary interference affecting registration accuracy.

[0054] After obtaining the aforementioned effective point cloud, the system performs iterative registration calculations with the stable point clouds on both sides of the roadway at the corresponding location in the current point cloud map to obtain a high-precision spatial transformation relationship. Before the registration process starts, the system uses the forward pose of the loader calculated in the aforementioned positioning steps as the initial pose, and presets the maximum number of iterations, convergence threshold, and interior point threshold as basic control parameters for the registration iteration. In each iteration, the system performs spatial transformation on the current effective point cloud according to the current pose transformation matrix, and for each effective point after transformation, searches for the spatial point with the smallest Euclidean distance in the corresponding area point cloud of the current point cloud map as the matching point. Then, it calculates the sum of squared errors between the effective point and its respective matching point as the overall registration error for that iteration.

[0055] To objectively determine whether the registration has reached its optimal state, the system simultaneously records and compares the overall registration error between two adjacent iterations, calculates the absolute value of the difference between the two errors, and uses this absolute value as the error change for the current iteration to characterize the convergence of registration accuracy as the iteration progresses. When this error change is less than a pre-set convergence threshold, it indicates that the error difference between the two iterations is extremely small, and the registration result has approached the global optimum. The system then terminates the iteration process and outputs the pose transformation matrix corresponding to the current iteration stage as the final optimal pose transformation matrix. This optimal pose transformation matrix fully describes the precise spatial displacement and attitude rotation relationship of the front of the loader relative to the map coordinate system at the current moment, and is the necessary geometric parameter basis for subsequently stitching the front point cloud data to the map coordinate system and performing boundary shearing.

[0056] In specific implementation, the poses of the front and rear parts, the real-time point cloud of the front lidar, and the point cloud map at the current moment are given to the map update module for boundary determination. First, the distance of the obstacle in front is compared with the set threshold. If it is greater than the set threshold, it means that the boundary has not been reached, and the map update value is set to 0. If it is less than or equal to the set threshold, it means that the boundary may have been reached. Then, it is determined whether the obstacle in front coincides with the boundary of the point cloud map at the current moment. If the boundary has been reached, it means that the current position has reached the loading area, and the loading action is performed, and the map update value is set to 0. If the boundary has not been reached, it means that the position in front may be the updated boundary position. Wait for 10 seconds. If the obstacle disappears within 10 seconds, it means that the boundary has not been reached, and the map update value is set to 0. If it always exists, it means that the boundary has been updated, and the map update value is set to 1. The boundary determination formula is shown in formula (3). In the formula, Represents the boundary determination formula; For indicator functions, when the condition within the parentheses is satisfied... =1, when not satisfied =0; Indicates the distance from the front to the obstacle; This indicates that a distance threshold has been set. The result of determining whether the obstacle ahead coincides with the current point cloud map boundary. = 1 indicates that the boundaries coincide. =0 indicates that the boundaries do not coincide; This indicates the determination that the obstacle persists for 10 seconds. = 1 indicates that it lasts for 10 seconds. =0 means it disappears within 10 seconds.

[0057] (3) when = 1, map update is performed first. Point cloud preprocessing is then performed, the 3D point cloud is projected onto a 2D plane, and the RANSAC method is used to fit straight lines on both sides of the alley. The point clouds on the left and right straight line segments are extracted and registered with the map at the current time. The point cloud set after dimensionality reduction after point cloud preprocessing is shown in formula (4). The formula for the minimum number of RANSAC iterations is shown in formula (5), the expression for fitting straight lines using the RANSAC method is shown in formula (6), and the method for extracting point clouds on the left and right straight line segments is shown in formula (7). Where, The point cloud set representing the projection of the front lidar 3D point cloud data onto the xy plane is used to extract the left and right side point clouds by fitting straight line segments on the 2D plane, and finally these points are extracted from the 3D point cloud. represents the minimum number of iterations for the RANSAC algorithm, ensuring the reliability of the line fitting; p represents the fitting success rate, its core function being to guarantee the reliability of the line fitting and adapt to the accuracy requirements of the tunnel scenario; e represents the proportion of outliers, used to represent... The proportion of interference points is considered to avoid the impact of interference points on the calculation of the number of iterations. This represents the minimum number of sampling points required to fit a two-dimensional straight line. and These are the straight segments of the left and right roadways fitted using the RANSAC method; This represents the interior point distance threshold, used to determine whether a point cloud contains interior points; This represents the set of point clouds on the left and right sides selected from the front sensor.

[0058] (4) (5)

[0059] (6) (7) Perform point cloud registration between the current map point cloud and the real-time point cloud of the front lidar on both sides of the alleyway. The initial pose is based on the current front pose:

[0060] The maximum number of iterations is preset to The convergence threshold is Interior point threshold is These parameters are used as the core input parameters for iteration. The iterative formula for point cloud registration pose transformation is shown in formula (8). In the formula, This represents the set of point clouds selected from the left and right sides in the i-th iteration. point The point after pose transformation; and Let these represent the rotation matrix and translation vector for the i-th iteration, respectively. They are obtained through the pose transformation matrix, and the initial pose is... Perform the calculation.

[0061] (8) After each pose transformation iteration, each point needs to calculate its distance with each point in the left and right lanes of the static point cloud map. The point with the smallest distance is taken as the matching point in this iteration. After a successful match, the matched point no longer needs to calculate distances with other points. After matching, an error calculation is required. The error is calculated by directly calculating the sum of the squares of the distances between the transformed point cloud and the current map point cloud. After calculating the error, the absolute values ​​of the error at the current moment and the error at the previous iteration need to be calculated for convergence verification. The objective function for error calculation is shown in formula (9), and the convergence verification function is shown in formula (10). Where, This represents the points selected from the left and right sides after pose transformation. The distance between each point and the points matched in the point cloud set of the left and right lanes of the static map is calculated, and the square of the distance directly represents the registration error of a single point.

[0062] This represents the difference in objective function values ​​between two consecutive iterations. This represents the error calculated in the i-th iteration. When i=1, it represents the error calculated in the first iteration. = 0. When the calculated difference Less than the convergence threshold This indicates that the pose is already optimal. If it exceeds this threshold, the current pose needs to be corrected using gradient descent. Calculate the partial derivatives and solve for the pose transformation matrix that minimizes the error. Then iterate again to calculate the difference in the objective function to ensure it is less than the convergence threshold.

[0063] (9) (10) After obtaining the optimal pose transformation matrix, the system enters the final and closed-loop stage of map updating. First, the system applies this optimal pose transformation matrix to the front point cloud data collected by the front LiDAR, transforming the spatial 3D point cloud, originally located in the local coordinate system of the front of the loader, point by point to the global coordinate system of the map, ensuring that the front point cloud data and the current point cloud map have the same spatial reference. This "transformation" is not a simple spatial translation, but rather, based on the rotation and translation components contained in the optimal pose transformation matrix, accurately mapping each 3D spatial point in the front point cloud to its corresponding geometric position in the map coordinate system.

[0064] After this transformation, the system merges the transformed front point cloud with the current point cloud map, that is, it integrates the newly perceived point cloud information of the mining boundary into the original map frame, thus initially forming a new map set covering the newly added boundary area.

[0065] However, the merged point cloud set may contain redundant point clouds located outside the new mining boundary and no longer physically part of the tunnel space. If these are not removed, it will cause point cloud tailing or error accumulation outside the boundary on the map, seriously damaging the spatial consistency of the map and the reliability of subsequent positioning. Therefore, the system needs to further perform boundary clipping and removal on the merged point cloud to restore the accuracy and clarity of the map boundary.

[0066] The specific execution process of boundary point shearing and removal is as follows: First, the system uses the RANSAC algorithm to refit the current mining boundary line of the part newly added from the front point cloud in the merged point cloud. This line represents the newly formed end boundary of the roadway at the current moment. Then, the system calculates the intersection points between the current boundary line and the stable line segments on the left and right sides of the original roadway, and obtains two boundary intersection points on the left and right sides. These two intersection points represent the spatial positions of the new boundary intersecting with the roadway wall on the left and right sides, and are the benchmark reference points for subsequent distance comparison.

[0067] After obtaining the boundary intersection points, the system needs to establish a geometric reference for spatial location determination. Specifically, the system calculates the geometric centroid of the newly added front point cloud set and determines the direction vector from the acquisition position of the front point cloud data (i.e., the spatial position of the front lidar when acquiring this frame of point cloud) to this centroid, using it as the forward reference vector to distinguish whether the spatial point is located directly in front of the loader. Subsequently, for each spatial point in the merged point cloud, the system calculates the line vector connecting it to the acquisition position of the front point cloud data, denoted as the current point vector, and determines the angle between the current point vector and the forward reference vector.

[0068] It should be noted that this included angle reflects the degree of deviation of the current point relative to the forward direction of the loader: when the included angle exceeds the preset angle range (e.g., greater than 90 degrees or less than -90 degrees), it indicates that the spatial point is located behind the loader or deviates too far from the front, belonging to the confirmed stable structural area behind the roadway, and is directly retained; when the included angle is within the preset angle range (e.g., between -90 degrees and 90 degrees), it indicates that the spatial point is located in front of the loader, and the system needs to further enter the distance comparison stage.

[0069] In the distance comparison process, the system compares the distance from the spatial point to the front point cloud data acquisition location with the distance from the intersection of the corresponding side boundary to the same acquisition location. If the distance from the spatial point to the acquisition location is greater than the distance from the intersection of the corresponding side boundary to the acquisition location, it means that the spatial point has exceeded the current mining boundary and is located outside the roadway boundary. In other words, the point is determined to be an outside point and is cut off. If the distance of the spatial point is not greater than the distance of the intersection of the corresponding side boundary, it means that the point is still located inside the roadway and is retained.

[0070] After the above-mentioned spatial angle determination and distance comparison filtering, all redundant point clouds exceeding the mining boundary were successfully removed. The merged point cloud was precisely trimmed to within the current mining boundary, forming a new map with clear boundaries and consistent space. This new map serves as the updated point cloud map, and the system automatically assigns it to the current point cloud map variable, which will then be used as the current point cloud map for the next time step.

[0071] Subsequently, the entire process returns to the initial data acquisition and positioning steps, and continues to execute the next round of front point cloud data acquisition, rear pose calculation, front pose estimation, map update trigger condition determination and map update based on the new point cloud map at the current moment. This cycle repeats continuously, achieving continuous dynamic map updates that are highly consistent with the "front moving, rear static" process characteristics of mining operations, ensuring that the loader always has an accurate, consistent and real-time point cloud map that matches the current mining progress throughout the entire operation cycle of the mining face.

[0072] In practical implementation, the optimal pose transformation matrix is ​​obtained through point cloud registration. Then, the current laser point cloud set is transformed according to the pose change matrix, and the point clouds of the left and right lanes are removed. This transformed set is then merged with the static point cloud map. After merging, the point clouds outside the updated boundaries are clipped and removed. The clipping strategy is as follows: Fit a straight line to the newly added point cloud set using the RANSAC method. .

[0073] calculate and and intersection and intersection .

[0074] Calculate the centroid of the newly added point cloud. Calculate the direction vector m between the front point O and the centroid M, and the vectors a and b between the front point and the intersection points A and B.

[0075] Calculate the angles between the vectors o and m of all points in the merged point cloud and the preceding points. ,when <-90° or When the angle is greater than 90°, it indicates that the point is behind the loader, and the current point is retained; when -90° ≤ When the angle is less than 0°, if the magnitude of the current point and the original vector o is greater than the magnitude of the left vector a, it means that the point is outside the boundary and needs to be cut; when 0° < When the angle is ≤ 90°, if the magnitude of the current point relative to the original vector o is greater than the magnitude of the vector b on the right, it means that the point is outside the boundary and needs to be clipped. Filtering out points outside the boundary yields an updated point cloud map. The updated map is then fed into the current point cloud map to complete the expansion of the current loop's point cloud map. The expression for the fitted straight line is shown in formula (11).

[0076] (11) Figure 4 This illustration shows a structural diagram of a device for updating an automated point cloud map of a loader, as provided in an embodiment of this application. The device includes: The data acquisition module is used to acquire the point cloud map at the current moment during the operation of the loader, and to acquire front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle body articulation angle data in real time; wherein, the front point cloud data is used to perceive the dynamic retreating mining boundary in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind. The positioning module is used to determine the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data; and to calculate the front orientation of the loader based on the rear orientation and the front and rear vehicle body hinge angle data. The registration module is used to register the effective point cloud in the front point cloud data with the point cloud map at the current time when the front pose and the front point cloud data meet the map update triggering conditions, so as to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after the front point cloud data is extracted by dimensionality reduction projection and fitting a straight line. The update module is used to transform the front point cloud data to the map coordinate system according to the optimal pose transformation matrix, merge it with the point cloud map at the current time, and perform boundary out-of-bounds point clipping and removal on the merged point cloud to obtain the updated point cloud map; the updated point cloud map is used as the point cloud map at the next time step.

[0077] The data acquisition module is used to acquire initial static point cloud data and map coordinate system parameters when the loader is started for the first time; wherein, the initial static point cloud data is the static structure point cloud of the roadway obtained by scanning before the loader is in operation; A construction module is used to construct an initial point cloud map based on the initial static point cloud data and the map coordinate system parameters; wherein the initial point cloud map serves as the initial reference benchmark for the point cloud map at the current moment.

[0078] The step of determining the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data includes: The rear point cloud data is registered with the stable rear tunnel structure in the point cloud map at the current moment, and the pose is corrected by combining the IMU inertial attitude data to determine the rear pose of the loader.

[0079] The step of calculating the front posture of the loader based on the rear posture and the hinge angle data of the front and rear vehicle bodies includes: The front orientation of the loader is calculated based on the rear orientation, the hinge angle data of the front and rear vehicle bodies, and the preset coordinate transformation relationship between the front and rear coordinate systems.

[0080] Whether the front pose and the front point cloud data meet the map update triggering conditions is determined by the following method: Calculate the distance to the obstacle in front corresponding to the front point cloud data; The map update triggering conditions are determined to be met when the distance to the obstacle in front is less than or equal to a set distance threshold, the obstacle in front does not coincide with the known boundary of the point cloud map at the current moment, and the obstacle in front continues to exist within a preset time; otherwise, the map update triggering conditions are determined not to be met.

[0081] The step of registering the effective point cloud with the point cloud map at the current moment to obtain the optimal pose transformation matrix includes: Using the aforementioned front pose as the initial pose, the maximum number of iterations, convergence threshold, and interior point threshold are set. The effective point cloud is iteratively registered with the point cloud on the corresponding alleyway side in the point cloud map at the current moment; In each iteration, for each valid point after transformation, the matching point with the smallest distance is found in the point cloud map, and the sum of squared errors of all matching point pairs is calculated; the absolute value of the difference between the sum of squared errors of two adjacent iterations is taken as the error change. When the error change is less than the convergence threshold, the iteration stops, and the current pose transformation matrix is ​​taken as the optimal pose transformation matrix.

[0082] The process of removing out-of-bounds points from the merged point cloud includes: The RANSAC algorithm is used to fit the current boundary line to the newly added part of the merged point cloud; Calculate the intersection points of the current boundary line and the original left and right side line segments of the roadway to obtain the intersection points of the left and right boundaries; Calculate the centroid of the newly added point cloud and determine the forward reference vector pointing from the acquisition position of the front point cloud data to the centroid; For each point in the merged point cloud, calculate its current point vector relative to the acquisition position of the preceding point cloud data, and determine the angle between the current point vector and the forward reference vector. If the included angle exceeds the preset range, the point is retained; if the included angle is within the preset range and the distance from the point to the acquisition position of the front point cloud data is greater than the distance from the intersection of the corresponding side boundary to the acquisition position of the front point cloud data, the point is determined to be located outside the boundary and is cut off.

[0083] like Figure 5 As shown, this application provides an electronic device for executing the method for updating the automatic operation point cloud map of a loader in this application. The device includes a memory, a processor, a bus, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method for updating the automatic operation point cloud map of a loader.

[0084] Specifically, the aforementioned memory and processor can be general-purpose memory and processor, without any specific limitations. When the processor runs the computer program stored in the memory, it can execute the aforementioned method for updating the automatic operation point cloud map of the loader.

[0085] Corresponding to the method for updating the point cloud map of the automatic operation of the loader in this application, this application embodiment also provides a computer storage medium storing a computer program, which is executed by a processor to perform the steps of the above-described method for updating the point cloud map of the automatic operation of the loader.

[0086] Specifically, the storage medium can be a general-purpose storage medium, such as a portable disk or hard disk. When the computer program on the storage medium is run, it can execute the above-mentioned method for updating the automatic operation point cloud map of the loader.

[0087] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0088] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0089] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0090] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0091] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0092] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for updating the point cloud map of an automated loader operation, characterized in that, The method includes: During the operation of the loader, the point cloud map at the current moment is acquired, and the front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle body articulation angle data are acquired in real time. Among them, the front point cloud data is used to perceive the dynamic retreating mining boundary in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind. Based on the point cloud map at the current moment, the rear point cloud data, and the IMU inertial attitude data, the rear orientation of the loader is determined; and based on the rear orientation and the front and rear vehicle body hinge angle data, the front orientation of the loader is calculated. When the front pose and the front point cloud data meet the map update triggering conditions, the effective point cloud in the front point cloud data is registered with the point cloud map at the current time to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after the front point cloud data is extracted by dimensionality reduction projection and fitting a straight line. Based on the optimal pose transformation matrix, the front point cloud data is transformed to the map coordinate system and merged with the point cloud map at the current moment. The merged point cloud is then clipped and removed from the boundary out-of-bounds points to obtain the updated point cloud map. The updated point cloud map is used as the point cloud map at the next moment.

2. The method according to claim 1, characterized in that, The method further includes: When the loader is started for the first time, initial static point cloud data and map coordinate system parameters are acquired; wherein, the initial static point cloud data is the static structure point cloud of the roadway obtained by scanning before the loader is in operation; An initial point cloud map is constructed based on the initial static point cloud data and the map coordinate system parameters; wherein, the initial point cloud map serves as the initial reference benchmark for the point cloud map at the current moment.

3. The method according to claim 1, characterized in that, The step of determining the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data includes: The rear point cloud data is registered with the stable rear tunnel structure in the point cloud map at the current moment, and the pose is corrected by combining the IMU inertial attitude data to determine the rear pose of the loader.

4. The method according to claim 1, characterized in that, The step of calculating the front posture of the loader based on the rear posture and the hinge angle data of the front and rear vehicle bodies includes: The front orientation of the loader is calculated based on the rear orientation, the hinge angle data of the front and rear vehicle bodies, and the preset coordinate transformation relationship between the front and rear coordinate systems.

5. The method according to claim 1, characterized in that, Whether the front pose and the front point cloud data meet the map update triggering conditions is determined by the following method: Calculate the distance to the obstacle in front corresponding to the front point cloud data; The map update triggering conditions are determined to be met when the distance to the obstacle in front is less than or equal to a set distance threshold, the obstacle in front does not coincide with the known boundary of the point cloud map at the current moment, and the obstacle in front continues to exist within a preset time; otherwise, the map update triggering conditions are determined not to be met.

6. The method according to claim 1, characterized in that, The step of registering the effective point cloud with the point cloud map at the current moment to obtain the optimal pose transformation matrix includes: Using the aforementioned front pose as the initial pose, the maximum number of iterations, convergence threshold, and interior point threshold are set. The effective point cloud is iteratively registered with the point cloud on the corresponding alleyway side in the point cloud map at the current moment; In each iteration, for each valid point after transformation, the matching point with the smallest distance is found in the point cloud map, and the sum of squared errors of all matching point pairs is calculated; the absolute value of the difference between the sum of squared errors of two adjacent iterations is taken as the error change. When the error change is less than the convergence threshold, the iteration stops, and the current pose transformation matrix is ​​taken as the optimal pose transformation matrix.

7. The method according to claim 1, characterized in that, The step of clipping and removing out-of-bounds points from the merged point cloud includes: The RANSAC algorithm is used to fit the current boundary line to the newly added part of the merged point cloud; Calculate the intersection points of the current boundary line and the original left and right side line segments of the roadway to obtain the intersection points of the left and right boundaries; Calculate the centroid of the newly added point cloud and determine the forward reference vector pointing from the acquisition position of the front point cloud data to the centroid; For each point in the merged point cloud, calculate its current point vector relative to the acquisition position of the preceding point cloud data, and determine the angle between the current point vector and the forward reference vector. If the included angle exceeds the preset range, the point is retained; if the included angle is within the preset range and the distance from the point to the acquisition position of the front point cloud data is greater than the distance from the intersection of the corresponding side boundary to the acquisition position of the front point cloud data, the point is determined to be located outside the boundary and is cut off.

8. A device for updating the point cloud map of an automatic operation of a loader, characterized in that, The device includes: The data acquisition module is used to acquire the point cloud map at the current moment during the operation of the loader, and to acquire front point cloud data, rear point cloud data, IMU inertial attitude data, and front and rear vehicle body articulation angle data in real time; wherein, the front point cloud data is used to perceive the dynamic retreating mining boundary in front of the operation, and the rear point cloud data is used to perceive the stable roadway structure behind. The positioning module is used to determine the rear orientation of the loader based on the current point cloud map, the rear point cloud data, and the IMU inertial attitude data; and to calculate the front orientation of the loader based on the rear orientation and the front and rear vehicle body hinge angle data. The registration module is used to register the effective point cloud in the front point cloud data with the point cloud map at the current time when the front pose and the front point cloud data meet the map update triggering conditions, so as to obtain the optimal pose transformation matrix; wherein, the effective point cloud represents the point cloud located on the straight line segments on the left and right sides of the alley after the front point cloud data is extracted by dimensionality reduction projection and fitting a straight line. The update module is used to transform the front point cloud data to the map coordinate system according to the optimal pose transformation matrix, merge it with the point cloud map at the current time, and perform boundary out-of-bounds point clipping and removal on the merged point cloud to obtain the updated point cloud map; the updated point cloud map is used as the point cloud map at the next time step.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine instructions that the processor executes. When the electronic device is running, the processor communicates with the memory via the bus. When the machine instructions are executed by the processor, they perform the steps of the method for updating the automatic operation point cloud map of a loader as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The computer storage medium stores a computer program that, when executed by a processor, performs the steps of the method for updating the automatic operation point cloud map of a loader as described in any one of claims 1 to 7.

Images