Unmanned mine car charging parking space positioning method based on laser radar point cloud

Abstract

The invention discloses an unmanned mine car charging parking space positioning method based on laser radar point cloud. The method comprises the following steps: acquiring laser radar point cloud data including excavator and mine car environment; preprocessing the laser radar point cloud data, and establishing a target pose model of a charging parking space based on safety specifications; performing point cloud registration on the current point cloud corresponding to the current position of the mine car and the target pose model, and calculating the relative pose deviation between the current pose of the mine car and the target pose; on the basis of the relative pose deviation, the pose adjustment amount needed by the unmanned mine car to drive into the charging parking space is calculated; and the unmanned mine car is controlled to move to the charging parking space according to the posture adjustment amount. According to the method, the laser radar point cloud is processed, the target parking space model is established and accurately registered with the current point cloud of the mine car, and the mine car is controlled to automatically park in the charging parking space after the pose deviation is calculated. Full-automatic and high-precision positioning of the charging berth of the unmanned mine car is realized, and the operation safety, continuity and overall efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of regulatory information fusion technology for open-pit mines, and in particular to a method for locating unmanned mining truck loading and parking spaces based on lidar point clouds. Background Technology

[0002] Loading and parking mainly refers to the process where, within the drivable area of ​​a mine, an unmanned mining truck, within a specified time, determines the required excavator type and loading position based on information such as the excavator's fixed shovel position and orientation angle, and plans the loading area path. It then efficiently, smoothly, and orderly parks itself into the corresponding shovel position from the tracking endpoint. After loading is complete, it smoothly exits and reaches the designated exit endpoint on the main road, thus successfully completing the loading and parking handover.

[0003] A comprehensive analysis of optimal parking point location methods for unmanned mining trucks, combined with the characteristics of mining operations and technical implementation paths, is divided into three mainstream methods: (1) Collaborative computing method based on excavator positioning The parking location of the mining trucks can be calculated by using the real-time location of the excavator.

[0004] Advantages: Relies on existing equipment, strong real-time performance.

[0005] Limitations: Requires high-precision GPS support; susceptible to signal interference in complex terrain.

[0006] (2) Grid map and trajectory optimization method Candidate parking spaces are generated based on environmental modeling, and then the optimal route is selected: Advantages: Adaptable to complex obstacle environments, with a high success rate in path planning.

[0007] Limitations: Implementation is complex, and grid maps are not suitable for large-scale environments, dynamic scenes, and resource-constrained systems.

[0008] (3) Key control point guidance method Controlling parking accuracy through key path nodes: Advantages: Solves the problem of deviation when reversing over long distances.

[0009] Limitations: It is particularly suitable for mining sites located far from main roads, but its advantages are not obvious in other scenarios.

[0010] To address the problem of low efficiency in finding the optimal loading and parking point for unmanned mining trucks, there is an urgent need for a method based on lidar for finding the optimal loading and parking point for unmanned mining trucks. Summary of the Invention

[0011] The purpose of this invention is to overcome the shortcomings of the existing technology. To achieve the above objective, an unmanned mining truck loading and parking space positioning method based on lidar point cloud is adopted to solve the problems mentioned in the background technology.

[0012] A method for locating loading parking spaces for unmanned mining trucks based on lidar point clouds includes the following steps: Step S1: Obtain lidar point cloud data containing the environment of the excavator and mining truck; Step S2: Preprocess the lidar point cloud data and establish a target pose model of the loading parking space based on safety specifications; Step S3: Perform point cloud registration between the current point cloud corresponding to the current position of the mine car and the target pose model, and calculate the relative pose deviation between the current pose and the target pose of the mine car; Step S4: Based on the relative pose deviation, calculate the attitude adjustment amount required for the unmanned mining truck to enter the loading parking space; Step S5: Control the unmanned mining truck to move to the loading parking position according to the attitude adjustment amount.

[0013] As a further aspect of the present invention: in step S2, establishing the target pose model specifically includes: A global parking coordinate system is established with the center of the excavator bucket as the origin. Based on the processed point cloud data, the geometric contour of the parking space boundary is extracted through clustering and segmentation, thereby generating the target pose model.

[0014] As a further aspect of the present invention: in step S2, the lidar point cloud data is preprocessed, specifically including: Point clouds acquired through multiple sensors or multiple frames are stitched together to unify them into the same coordinate system, and then denoising and smoothing are performed.

[0015] As a further aspect of the present invention, step S3 specifically includes: Step S31: Divide the current point cloud into a mining truck body point cloud and an excavator point cloud; Step S32: Register the excavator point cloud with the corresponding excavator reference point cloud in the target pose model to obtain the initial pose transformation parameters; Step S33: Transform the point cloud of the mine car body using the initial pose transformation parameters, and perform coarse and fine registration between the transformed point cloud of the mine car body and the target pose model to calculate the relative pose deviation.

[0016] As a further aspect of the present invention: in step S31, the point cloud is divided into the point cloud of the mining truck body and the point cloud of the excavator by a preset segmentation plane equation.

[0017] As a further aspect of the present invention: in step S32, the excavator point cloud and the excavator reference point cloud are registered using an iterative nearest point algorithm to obtain a rotation matrix and a translation vector as the initial pose transformation parameters.

[0018] As a further aspect of the present invention: in step S33, the coarse registration is performed using the sample consistency initial registration algorithm, and the fine registration is performed using the iterative nearest point algorithm.

[0019] As a further aspect of the present invention: during the fine registration process, point cloud data of the excavator cab is extracted from the target pose model as the target point cloud for registration.

[0020] As a further aspect of the present invention: in step S4, the attitude adjustment amount includes the distance deviation between the center of the mine car and the center of the berth in the lateral and longitudinal directions, as well as the angle deviation between the heading angle of the mine car and the fitted straight line of the berth boundary.

[0021] As a further aspect of the present invention: in step S5, control commands are generated based on the attitude adjustment amount and in combination with the mine car kinematic model and steering constraints to dynamically control the mine car to travel to the loading parking space.

[0022] Compared with the prior art, the present invention has the following technical advantages: Using the aforementioned technical solution, raw point cloud data containing the environment of the excavator and mining truck is acquired via LiDAR, and preprocessed and aligned to a coordinate system. Subsequently, a target pose model of the loading parking space is established according to safety regulations (with the bucket center as the origin). The core step is to perform multi-stage point cloud registration (including segmentation, initial registration, coarse registration, and fine registration) between the point cloud corresponding to the current position of the mining truck and the target model, thereby accurately calculating the relative position and angular deviation between the current pose of the mining truck and the ideal parking space. Finally, based on this deviation, a specific attitude adjustment amount is calculated, and control commands are generated accordingly to guide the unmanned mining truck to automatically and accurately drive to the loading parking space.

[0023] This system achieves fully automated, high-precision unmanned mining truck loading and parking positioning. Utilizing lidar point cloud registration technology, it overcomes the challenges of potentially poor GPS signals in complex mining environments and the significant impact of lighting and dust on traditional visual methods. It enables non-contact and precise calibration of the relative spatial relationship between the mining truck and the excavator (loading point). This significantly reduces manual intervention and calibration time, improving the continuity and safety of the workflow. Simultaneously, the calculation method based on the accurate model and registration enhances the repeatability and reliability of parking positioning, laying a precise location foundation for subsequent automated loading, thereby comprehensively improving the intelligence level and operational efficiency of mine transportation. Attached Figure Description

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings: Figure 1 This is a schematic diagram illustrating the steps of a parking space positioning method according to an embodiment of this application. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please refer to Figure 1 In this embodiment of the invention, a method for locating unmanned mining truck loading parking spaces based on lidar point clouds includes the following steps: Step S1: Obtain lidar point cloud data containing the environment of the excavator and mining truck; In this specific implementation, the driverless mining truck enters the loading waiting area. The driverless mining truck receives information from the cloud control platform, including the excavator type, location, and orientation angle.

[0027] Step S2: Preprocess the lidar point cloud data and establish a target pose model of the loading parking space based on safety specifications; In this embodiment, step S2, establishing the target pose model specifically includes: A global parking coordinate system is established with the center of the excavator bucket as the origin. Based on the processed point cloud data, the geometric contour of the parking space boundary is extracted through clustering and segmentation, thereby generating the target pose model.

[0028] In this embodiment, step S2 involves preprocessing the lidar point cloud data, specifically including: Point clouds acquired through multiple sensors or multiple frames are stitched together to unify them into the same coordinate system, and then denoising and smoothing are performed.

[0029] In a specific implementation, a parking coordinate system is established: The parking coordinate system is established based on the safety regulations for mine car loading coordination. The origin of this coordinate system is located at the center of the bucket, the X and Y axes are parallel to the bucket plane, and the Z axis is perpendicular to the bucket surface.

[0030] In a specific implementation, splicing measurements are performed: Because vehicle-mounted LiDAR has a limited scanning field of view, it is necessary to stitch together point cloud data from multiple frames or multiple sensors to unify them into the same coordinate system through coordinate transformation.

[0031] Multiple vehicle-mounted radars are deployed at fixed intervals and stitched together based on installation parameter constraints. Point clouds acquired from different viewpoints or at different times are then aligned using coordinate transformations (rotation matrices and translation vectors). The key lies in finding the correspondences between the point clouds and calculating the optimal transformation parameters to minimize registration errors.

[0032] Using lidar point cloud data, density peak clustering algorithm is employed to segment the berth boundary and identify the berth's geometric contour.

[0033] Construct a three-dimensional coordinate system for the berth, extract key features (such as boundary corners and centerlines), and generate an ideal pose model of the target berth.

[0034] In a specific implementation, point cloud preprocessing is performed: Remove invalid data: The scanning process may contain environmental interference or equipment noise points, which need to be cleaned up by statistical filtering or radius filtering.

[0035] Smoothing: Use bilateral filtering or Gaussian filtering to optimize the surface point cloud distribution, reduce jagged edges caused by scanning jitter, and improve the accuracy of subsequent registration.

[0036] Step S3: Perform point cloud registration between the current point cloud corresponding to the current position of the mine car and the target pose model, and calculate the relative pose deviation between the current pose and the target pose of the mine car; In this embodiment, step S3 specifically includes: Step S31: Divide the current point cloud into a mining truck body point cloud and an excavator point cloud; In this embodiment, in step S31, the point cloud is divided into the point cloud of the mining truck body and the point cloud of the excavator by a preset segmentation plane equation.

[0037] Step S32: Register the excavator point cloud with the corresponding excavator reference point cloud in the target pose model to obtain the initial pose transformation parameters; In this embodiment, in step S32, the excavator point cloud and the excavator reference point cloud are registered using the iterative nearest point algorithm to obtain the rotation matrix and translation vector as the initial pose transformation parameters.

[0038] Step S33: Transform the point cloud of the mine car body using the initial pose transformation parameters, and perform coarse and fine registration between the transformed point cloud of the mine car body and the target pose model to calculate the relative pose deviation.

[0039] In this embodiment, in step S33, the sample consistency initial registration algorithm is used for coarse registration, and the iterative nearest point algorithm is used for fine registration.

[0040] In this embodiment, during the fine registration process, the point cloud data of the excavator cab is extracted from the target pose model as the target point cloud for registration.

[0041] In a specific implementation, coordinate alignment is performed between the current pose and the target pose: The point cloud data of the current position of the mining truck is converted into a global coordinate system. Through feature matching, the current point cloud is aligned with the target berth model, and the relative offset is calculated.

[0042] The obtained point cloud data includes data from the vehicle-mounted LiDAR and the excavator, both within the same coordinate system. It is primarily used to calibrate the spatial relationship between theoretical and actual parking spaces. By registering the vehicle-mounted LiDAR point clouds in the theoretical and actual parking space point cloud data, both are moved to the same coordinate system.

[0043] Among them, cloud segmentation and registration of loading and parking points are performed: Voxel grid downsampling method is used for point cloud data After simplification, a simplified measurement point cloud P is obtained.

[0044] This point cloud data includes point cloud data from an onboard LiDAR system. Excavator point cloud data .

[0045] Next, after obtaining the reference parking spot cloud Q (including those in the same coordinate system) , It is necessary to unify the measured parking spot cloud and the reference parking spot cloud into a single coordinate system.

[0046] This is based on dividing the measurement point cloud into two parts: the vehicle point cloud and the excavator point cloud.

[0047] Points closer to the plane are assigned to one side of the vehicle, thus obtaining the vehicle's point cloud. and excavator measuring point cloud .

[0048]

[0049]

[0050] In the formula, a, b, and c are the coefficients of the equation for the dividing plane; , , These are the 3D coordinates of a point in the point cloud, derived from data in point cloud P. Setting a small offset ensures that all point clouds acquired by the vehicle-mounted LiDAR are assigned to the excavator side.

[0051] The initial pose is obtained through point cloud registration using an onboard LiDAR: Subsequently, the ICP algorithm was used to analyze the point cloud of the loading parking space. Reference parking point cloud Registration is performed to unify the position of the vehicle-mounted LiDAR in the vehicle coordinate system to the parking coordinate system, serving as the basis for subsequent optimal parking point cloud registration.

[0052] The rotation matrix from the measured point cloud of the vehicle-mounted lidar to the reference point cloud is solved. Translation vector Afterwards, through and Update measurement point cloud The point cloud of the vehicle-mounted lidar measurement in the reference coordinate system is obtained. ; +

[0053] Point cloud measurement via vehicle-mounted lidar registration Both the reference point cloud Q and the reference point cloud Q are roughly located at the center of the vehicle-mounted LiDAR, possessing good initial pose. Based on this, the measurement point cloud is accurately registered. And reference point cloud Q.

[0054] This includes coarse registration of the berth scan point cloud with the reference point cloud: Analyzing the excavator's point cloud information, if the complete measured point cloud is used for registration, the point cloud will have a lot of noise and the registration accuracy will be low due to the differences in excavator reflections under different conditions. Therefore, only the measured point cloud and the point cloud data Qum of the excavator's cab in the reference point cloud Q are registered. Extracting point cloud data of the excavator cab from the reference point cloud. The initial excavator cab is obtained. The point cloud changes of the excavator cab are assessed; if they exceed a certain threshold, indicating a boundary has been reached, the selected excavator cab area is then converted into point cloud data. .

[0055] by Target point cloud, Instead of coarse registration, the Sample Consensus Initial Alignment (SAC-IA) algorithm is used for... Perform coarse registration.

[0056] Step S4: Based on the relative pose deviation, calculate the attitude adjustment amount required for the unmanned mining truck to enter the loading parking space; In this embodiment, in step S4, the attitude adjustment amount includes the distance deviation between the center of the mine car and the center of the berth in the lateral and longitudinal directions, as well as the angle deviation between the heading angle of the mine car and the fitted straight line of the berth boundary.

[0057] In a specific implementation, the steps for calculating the attitude adjustment of the unmanned mining truck are as follows: Calculate the lateral (X-axis) and longitudinal (Y-axis) distances between the center of the mine car and the center of the berth, and determine the required movement amount by combining the kinematic model of the mine car.

[0058] The adjustment angle is determined by fitting a straight line at the berth boundary and the heading angle of the mine car.

[0059] Step S5: Control the unmanned mining truck to move to the loading parking position according to the attitude adjustment amount.

[0060] In this embodiment, in step S5, control commands are generated based on the attitude adjustment amount and combined with the mine car kinematic model and steering constraints to dynamically control the mine car to drive to the loading parking space.

[0061] Specifically, the excavator is loaded with materials. After the driverless vehicle completes parking, the excavator begins loading.

[0062] The beneficial effects of this invention are: (1) The accuracy of perception and positioning has been significantly improved. By deeply integrating LiDAR with point cloud classification technology, centimeter-level positioning accuracy is achieved. Lateral / heading errors are controlled within the centimeter level, maintaining 360-degree coverage even in dusty mining environments, thus solving the problem of traditional cameras being susceptible to interference.

[0063] The point cloud model, combined with dynamic environmental feature extraction, improves the accuracy of loop closure detection and supports precise parking in complex scenarios.

[0064] (2) Optimization of computational efficiency and real-time performance By employing point cloud rasterization processing and acceleration technology, the algorithm complexity is reduced, enabling it to run in real time on low-computing-power industrial control computers, thus meeting the rapid response requirements of mining scenarios.

[0065] Path planning, combined with curvature optimization and coordinate system transformation, shortens computation time while improving trajectory following performance (such as improved success rate after parking path optimization).

[0066] (3) Enhanced adaptability to dynamic environments and resistance to interference AI point cloud model and collaborative control dynamically adjust the excavator's actions in real time based on vehicle deviation (such as fine adjustments caused by loading vibration), reducing material spillage rate.

[0067] The lidar's anti-interference capabilities ensure stable operation under complex lighting and dust conditions in mining areas.

[0068] (4) Multi-system collaborative and integrated control Parking path optimization technology effectively reduces the number of times you need to adjust your parking position while reversing, improving operational safety in confined spaces.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention. The scope of the invention is defined by the appended claims and their equivalents, all of which should be included within the scope of protection of the invention.

Claims

1. A method for locating loading parking spaces for unmanned mining trucks based on lidar point clouds, characterized in that, Includes the following steps: Step S1: Obtain lidar point cloud data containing the environment of the excavator and mining truck; Step S2: Preprocess the lidar point cloud data and establish a target pose model of the loading parking space based on safety specifications; Step S3: Perform point cloud registration between the current point cloud corresponding to the current position of the mine car and the target pose model, and calculate the relative pose deviation between the current pose and the target pose of the mine car; Step S4: Based on the relative pose deviation, calculate the attitude adjustment amount required for the unmanned mining truck to enter the loading parking space; Step S5: Control the unmanned mining truck to move to the loading parking position according to the attitude adjustment amount.

2. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 1, characterized in that, In step S2, establishing the target pose model specifically includes: A global parking coordinate system is established with the center of the excavator bucket as the origin. Based on the processed point cloud data, the geometric contour of the parking space boundary is extracted through clustering and segmentation, thereby generating the target pose model.

3. The method for locating unmanned mining truck loading and parking spaces based on lidar point clouds according to claim 2, characterized in that, In step S2, the lidar point cloud data is preprocessed, specifically including: Point clouds acquired through multiple sensors or multiple frames are stitched together to unify them into the same coordinate system, and then denoising and smoothing are performed.

4. The method for locating unmanned mining truck loading and parking spaces based on lidar point clouds according to claim 1, characterized in that, Step S3 specifically includes: Step S31: Divide the current point cloud into a mining truck body point cloud and an excavator point cloud; Step S32: Register the excavator point cloud with the corresponding excavator reference point cloud in the target pose model to obtain the initial pose transformation parameters; Step S33: Transform the point cloud of the mine car body using the initial pose transformation parameters, and perform coarse and fine registration between the transformed point cloud of the mine car body and the target pose model to calculate the relative pose deviation.

5. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 4, characterized in that, In step S31, the point cloud is divided into the mine car body point cloud and the excavator point cloud by a preset segmentation plane equation.

6. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 4, characterized in that, In step S32, the excavator point cloud is registered with the excavator reference point cloud using the iterative nearest point algorithm to obtain the rotation matrix and translation vector as the initial pose transformation parameters.

7. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 4, characterized in that, In step S33, the coarse registration is performed using the sample consistency initial registration algorithm, and the fine registration is performed using the iterative nearest point algorithm.

8. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 7, characterized in that, In the fine registration process, point cloud data of the excavator cab is extracted from the target pose model as the target point cloud for registration.

9. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 1, characterized in that, In step S4, the attitude adjustment amount includes the distance deviation between the center of the mine car and the center of the berth in the lateral and longitudinal directions, as well as the angle deviation between the heading angle of the mine car and the fitted straight line of the berth boundary.

10. The method for locating unmanned mining truck loading parking spaces based on lidar point clouds according to claim 1, characterized in that, In step S5, based on the attitude adjustment amount and combined with the mine car kinematic model and steering constraints, control commands are generated to dynamically control the mine car to travel to the loading parking space.

Images