Mobile robot high-precision positioning method, electronic equipment and program product

Abstract

The invention provides a mobile robot high-precision positioning method, electronic equipment and a computer program product. The method comprises the steps of determining a final pose of the mobile robot at a first moment, and estimating an estimated pose of the mobile robot at a second moment; in response to the high-resolution map switching instruction, determining a local high-resolution positioning map corresponding to the target position; based on laser radar scanning data and the local high-resolution positioning map, the estimated pose is optimized; wherein the step of constructing the local high-resolution positioning map comprises the steps of determining an effective calibration laser point set and a weight value of each calibration laser point; determining a weight coefficient of each grid in the second local high-resolution map based on the weight value; calculating a probability density value of each grid based on the weight coefficient; and recording the probability density value to a second local high-resolution map to obtain a local high-resolution positioning map.

Description

Technical Field

[0001] This disclosure relates to the field of mobile robot positioning, specifically to a high-precision positioning method, electronic device, and software product for mobile robots. Background Technology

[0002] Mobile robots (such as automated guided vehicles (AGVs)) serve as key carriers for intelligent logistics and flexible manufacturing systems, and their high-precision, robust positioning capabilities are prerequisites for achieving autonomous navigation and collaborative operations. Current mainstream positioning technologies mainly fall into two categories: passive positioning relying on human landmarks and active perception positioning relying on multi-sensor fusion. Both have their advantages and disadvantages in terms of engineering applicability, environmental adaptability, and system maintainability, and neither can simultaneously meet the requirements of convenient deployment, long-term stability, and reliability under strong interference in industrial environments. Therefore, there is an urgent need for an autonomous positioning method for mobile robots that requires no pre-set markers, does not rely on vision, and can adapt to complex industrial environments. Summary of the Invention

[0003] This disclosure provides a high-precision positioning method, electronic device, and software product for mobile robots.

[0004] According to one aspect of this disclosure, a high-precision positioning method for a mobile robot is provided, comprising: The process involves: constructing local high-resolution localization maps corresponding to various target locations in the mobile robot's operating environment; determining the final pose of the mobile robot at a first moment; estimating the pose of the mobile robot at a second moment based on the final pose at the first moment, where the second moment is a moment after the first moment; responding to a high-resolution map switching command, determining a local high-resolution localization map of the target location corresponding to the mobile robot at the second moment; and optimizing the estimated pose based on LiDAR scanning data and the local high-resolution localization map to obtain a final pose, which is then used as the high-precision localization result of the mobile robot at the second moment. Constructing the local high-resolution localization map corresponding to the target location includes: constructing a basic point cloud map of the mobile robot's operating environment; determining the target location... A corresponding set of valid calibration laser points, comprising multiple calibration laser points; weight values ​​for each calibration laser point are set based on whether it is a point on a straight line and the relationship between its intensity value and an intensity threshold; a local base map is extracted from the base point cloud map centered on the target location; the local base map is segmented to obtain a second local high-resolution map; the weight values ​​of each calibration laser point are mapped to the second local high-resolution map to obtain weight coefficients for each grid cell in the second local high-resolution map; based on the weight coefficients, probability density is calculated for the second local high-resolution map to obtain probability density values ​​for each grid cell in the second local high-resolution map; the probability density values ​​are recorded in the second local high-resolution map to obtain the local high-resolution positioning map.

[0005] According to one technical solution, by constructing a local high-resolution positioning map specific to the target location and integrating the geometric structural features of laser points (such as straight-line orientation) and physical reflection characteristics (intensity values) for differentiated weighting and probabilistic modeling, the accuracy and discriminativeness of the map's environmental representation are significantly improved. During the positioning phase, the local high-resolution map is dynamically switched and matched based on the estimated pose, and pose optimization is performed based on high-confidence areas, effectively overcoming the positioning ambiguity and drift problems caused by insufficient global map resolution, missing environmental textures, or repetitive structures.

[0006] According to at least one embodiment of this disclosure, determining the effective set of calibration laser points corresponding to the target location includes: controlling a lidar installed on a mobile robot located at the target location to collect multiple frames of lidar scanning data; determining an initial set of calibration laser points corresponding to the lidar based on the multiple frames of lidar scanning data; and using the initial set of calibration laser points corresponding to the lidar as the effective set of calibration laser points corresponding to the target location.

[0007] According to the technical solution of this embodiment, it can ensure that the calibration features are highly consistent with the actual working conditions, significantly improve the on-site adaptability and positioning reproduction accuracy of local high-resolution maps, and avoid the matching failure problem caused by the deviation between offline mapping and online working conditions.

[0008] According to at least one embodiment of this disclosure, determining the initial calibration laser point set corresponding to the lidar includes: constructing a first local high-resolution map centered on the mobile robot with a first resolution, wherein the first resolution is higher than the resolution of the base point cloud map; projecting the multi-frame lidar scanning data onto the first local high-resolution map and determining the number of times each grid in the first local high-resolution map is hit; selecting multiple grids with the highest number of hits as candidate grids; obtaining calibration laser points based on the candidate grids; and forming the initial calibration laser point set by assembling multiple calibration laser points.

[0009] According to the technical solution of this embodiment, by constructing a first local high-resolution map and statistically analyzing the grid hit frequency to screen stable feature points, noise and random interference can be effectively filtered out, significantly improving the repeatability, environmental robustness and geometric saliency of the calibration laser points, laying a reliable feature foundation for subsequent high-precision positioning.

[0010] According to at least one embodiment of this disclosure, obtaining a calibration laser point based on the candidate grid includes: using the geometric center of the candidate grid as the coordinate value of the calibration laser point, and using the maximum intensity value hitting the candidate grid as the intensity value of the calibration laser point.

[0011] According to the technical solution of this embodiment, by characterizing the calibration laser point with the geometric center and maximum intensity value of the high-hit-frequency grid, the spatial stability and physical repeatability of the calibration feature point can be improved.

[0012] According to at least one embodiment of this disclosure, determining the effective calibration laser point set corresponding to the target location includes: controlling at least two lidars installed on a mobile robot located at the target location to collect multiple frames of lidar scanning data; determining an initial calibration laser point set corresponding to each lidar based on the multiple frames of lidar scanning data; determining an effective straight line set corresponding to each initial calibration laser point set; determining the minimum distance between each lidar and each straight line in the corresponding effective straight line set; selecting the lidar with the smaller minimum distance from the at least two lidars as the calibration lidar; and using the initial calibration laser point set corresponding to the calibration lidar as the effective calibration laser point set corresponding to the target location.

[0013] According to the technical solution of this embodiment, by comparing data from multiple lidar systems, the calibration lidar is selected based on its geometric proximity (minimum distance) to an effective straight line. This ensures that the selected lidar has the optimal observation angle and the richest features. It can significantly improve the structural representativeness and matching robustness of the calibration lidar points, providing a more stable and reliable feature basis for high-precision positioning.

[0014] According to at least one embodiment of this disclosure, determining the effective set of straight lines corresponding to each initial calibration laser point set includes: performing a Hough transform based on the initial calibration laser point set to obtain at least one first candidate straight line; determining the angular deviation between each first candidate straight line and the mobile robot; taking the first candidate straight line whose angular deviation is less than an angular deviation threshold as an effective straight line; and obtaining the effective set of straight lines based on the effective straight lines.

[0015] According to the technical solution of this embodiment, falsely detected straight lines caused by environmental stray structures or calibration attitude interference can be effectively eliminated, significantly improving the task relevance and docking scenario adaptability of the effective straight line set.

[0016] According to at least one embodiment of this disclosure, setting the weight value of each calibration laser point includes: performing a Hough transform based on the effective calibration laser point set to obtain a second candidate line; setting the weight value of the calibration laser point located on the second candidate line in the effective calibration laser point set as a first weight value; setting the weight value of the calibration laser point not located on the second candidate line in the effective calibration laser point set, and whose intensity value is greater than or equal to an intensity value threshold, as a second weight value; and setting the weight value of the calibration laser point not located on the second candidate line in the effective calibration laser point set, and whose intensity value is less than an intensity value threshold, as a third weight value, wherein the first weight value is greater than the second weight value, and the second weight value is greater than the third weight value.

[0017] According to the technical solution of this embodiment, by integrating geometric structural saliency (whether it is located on the second candidate line) and physical reflection characteristics (the relationship between the intensity value threshold) to set three levels of differentiated weight values, map matching can be more focused on high reliability feature points, which can significantly improve the positioning accuracy, robustness and anti-interference ability.

[0018] According to at least one embodiment of this disclosure, segmenting the local base map to obtain a second local high-resolution map includes: segmenting each grid cell in the local base map into i i grid cells are used to obtain a second local high-resolution map, where i is the ratio of the resolution of the base point cloud map to the first resolution.

[0019] According to the technical solution of this embodiment, by subdividing the local base map into regular grids according to the resolution ratio, the spatial resolution of the map can be efficiently improved without introducing interpolation distortion, so that the local map can accurately carry the geometric and intensity features of the calibrated laser points.

[0020] According to at least one embodiment of this disclosure, the weight values ​​of each calibrated laser point are mapped to the second local high-resolution map to obtain the weight coefficients of each grid in the second local high-resolution map, including: marking grids belonging to obstacles in the local base map as obstacle grids and grids not belonging to obstacles as idle grids; and dividing the i-th grid obtained by segmenting each grid in the local base map... i grids form a grid set, resulting in multiple grid sets. Based on the coordinate values ​​of each calibrated laser point, the weight value of each calibrated laser point is mapped to the corresponding grid in the second local high-resolution map, serving as the weight coefficient of the corresponding mapped grid in the second local high-resolution map. When the unmapped grid and the mapped grid in the second local high-resolution map belong to the same grid set, the weight coefficient of the unmapped grid in the second local high-resolution map is set to 0. When the unmapped grid and the mapped grid in the second local high-resolution map do not belong to the same grid set, the weight coefficient of the unmapped grid belonging to the obstacle grid is set to the fourth weight value, which is less than the third weight value, and the weight coefficient of the unmapped grid belonging to the idle grid is set to 0.

[0021] According to the technical solution of this embodiment, the representation accuracy and robustness of obstacle boundary regions in the second local high-resolution map are effectively improved by using a refined weight mapping and classification assignment strategy, while taking into account both computational efficiency and environmental semantic consistency.

[0022] According to at least one embodiment of this disclosure, based on the weighting coefficient, a probability density calculation is performed on the second local high-resolution map to obtain the probability density value of each grid in the second local high-resolution map, including: determining the nearest obstacle grid corresponding to each grid in the second local high-resolution map; calculating the Euclidean distance value between each grid and the corresponding nearest obstacle grid; when the Euclidean distance value is less than or equal to 0, setting the probability density value of each grid to the weighting coefficient of the nearest obstacle grid corresponding to each grid; when the Euclidean distance is greater than 0 and less than or equal to the maximum Euclidean distance, setting the probability density value of each grid to the product of the weighting coefficient of the nearest obstacle grid corresponding to each grid and the Gaussian decay function; and when the Euclidean distance is greater than the maximum Euclidean distance, setting the probability density value of each grid to 0.

[0023] According to the technical solution of this embodiment, a physically meaningful continuous likelihood field is constructed by combining the weight coefficients of the nearest obstacle grid with a distance-dependent Gaussian decay model. This preserves the high-confidence core region of the calibration features while achieving a smooth transition and truncation of the influence range.

[0024] According to at least one embodiment of this disclosure, optimizing the estimated pose based on lidar scanning data and the local high-resolution positioning map to obtain a final pose includes: performing a first screening on the lidar scanning data based on the local high-resolution positioning map to obtain an initial set of lidar points within the range of the local high-resolution positioning map; performing a second screening on the initial set of lidar points to filter out lidar points without obstacle grids in the neighborhood to obtain an effective set of lidar points; optimizing the estimated pose using multiple pose optimization algorithms to obtain multiple candidate optimized poses; determining the mapping position of each lidar point in the effective set of lidar points to the local high-resolution positioning map based on each candidate optimized pose; calculating the score of each candidate optimized pose based on the probability density value of the grid corresponding to each mapping position; and selecting the candidate optimized pose with the highest score as the final pose.

[0025] According to the technical solution of this embodiment, multiple pose optimization algorithms work together, and each candidate pose is quantitatively scored and selected based on the probability density in the local high-resolution positioning map. This avoids the possibility of a single algorithm getting stuck in local minima or malconvergence caused by noise interference, significantly improving the reliability, consistency, and environmental adaptability of high-precision positioning results.

[0026] According to at least one embodiment of this disclosure, the score of each candidate optimized pose is calculated based on the probability density value of the grid corresponding to the mapped position, including: taking the ratio of the sum of the probability density values ​​of the grid corresponding to each mapped position to the number of effective lasers as the score of the candidate optimized pose, wherein the number of effective lasers is the number of laser points in the set of effective laser points mapped to grids with a probability density value greater than 0.

[0027] According to the technical solution of this embodiment, by introducing the normalization of the number of effective laser points, the scoring deviation caused by the fluctuation of the number of observation points (such as occlusion and changes in viewing angle) can be effectively suppressed, making the pose evaluation more focused on the matching quality rather than the number of points, which significantly improves the fairness of the scoring among multiple candidate poses and the reliability of the optimal solution selection.

[0028] According to at least one embodiment of this disclosure, the high-resolution map switching instruction is generated when the Bezier module starts.

[0029] According to the technical solution of this embodiment, the high-resolution positioning map can be dynamically switched on demand, ensuring that the high-precision map and calibration lidar are only used during the Bezier docking stage, which balances the system's real-time performance and positioning accuracy, and avoids the computational redundancy and resource waste caused by the global high-resolution map.

[0030] According to another aspect of this disclosure, an electronic device is provided, comprising: a memory storing execution instructions; and a processor executing the execution instructions stored in the memory, such that the processor performs a high-precision positioning method for a mobile robot according to any embodiment of this disclosure.

[0031] According to another aspect of this disclosure, a readable storage medium is provided, wherein execution instructions are stored therein, which, when executed by a processor, are used to implement a high-precision positioning method for a mobile robot according to any embodiment of this disclosure.

[0032] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements a high-precision positioning method for a mobile robot according to any embodiment of this disclosure. Attached Figure Description

[0033] The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0034] Figure 1 This is a flowchart illustrating a high-precision positioning method for a mobile robot according to one embodiment of the present disclosure.

[0035] Figure 2 This is a flowchart illustrating a method for constructing a local high-resolution positioning map according to one embodiment of the present disclosure.

[0036] Figure 3 This is a flowchart illustrating a method for determining an effective set of calibrated laser points according to one embodiment of the present disclosure.

[0037] Figure 4 This is a flowchart illustrating the method corresponding to step S320 of one embodiment of the present disclosure.

[0038] Figure 5 This is a flowchart illustrating a method for determining an effective set of calibrated laser points according to another embodiment of this disclosure.

[0039] Figure 6 This is a flowchart illustrating the method corresponding to step S530 of one embodiment of the present disclosure.

[0040] Figure 7 This is a flowchart illustrating the method corresponding to step S230 of one embodiment of the present disclosure.

[0041] Figure 8 This is a flowchart illustrating the method corresponding to step S270 of one embodiment of the present disclosure.

[0042] Figure 9 This is a flowchart illustrating a pose optimization method according to one embodiment of the present disclosure.

[0043] Figure 10 This is a schematic diagram of a method for determining weighting coefficients according to one embodiment of the present disclosure.

[0044] Figure 11 This is a schematic block diagram of a high-precision positioning device for a mobile robot according to one embodiment of the present disclosure.

[0045] Figure 12 This is a schematic structural block diagram of an electronic device employing a processor-based hardware implementation according to one embodiment of the present disclosure. Detailed Implementation

[0046] The present disclosure will now be described in further detail with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0047] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] Existing mobile robot localization solutions suffer from significant application bottlenecks. One bottleneck involves solutions based on manual landmarks such as reflectors and QR codes, which, while offering high positioning accuracy, require high upfront deployment and ongoing maintenance costs. Furthermore, these solutions are prone to failure in environments requiring modification, marker occlusion, or contamination, making them unsuitable for the rapid reconfiguration needs of flexible production lines. Another bottleneck is that while mainstream vision-laser fusion solutions possess environmental adaptability potential, they are severely limited by drastic fluctuations in industrial lighting (such as strong welding light, day-night cycles, and flickering lights), dust and oil mist, and equipment vibration interference. This leads to unstable image feature extraction, increased mismatch rates, and even complete loss of localization capability. Particularly in harsh conditions such as high-temperature, high-humidity, and high-dust automotive welding or foundry workshops, the vision subsystem often becomes a bottleneck in localization reliability. Therefore, there is an urgent need for a localization mechanism that eliminates reliance on manual markers and optical imaging, relying solely on stable physical quantities (such as laser reflection characteristics and geometric structures) for environmental perception and pose calculation, to improve the usability and robustness of mobile robots in real-world industrial scenarios.

[0049] To address this, this disclosure proposes the following technical solution: A local high-resolution positioning map corresponding to each target location in the mobile robot's operating environment is constructed. Upon receiving a map switching command, pose optimization is performed based on LiDAR scanning data and the local high-resolution positioning map, ultimately outputting a high-precision positioning result. Specifically, in constructing the local high-resolution positioning map, a set of effective calibration laser points is extracted based on the base point cloud. The geometric linear features and intensity values ​​of each calibration laser point in the effective calibration laser point set are dynamically weighted and mapped to a segmented high-resolution raster map to generate a probability density representation. This disclosure significantly improves the environmental representation accuracy and discriminability of the map by constructing a local high-resolution positioning map specific to the target location and integrating the geometric structural features (such as linear attribution) and physical reflection characteristics (intensity values) of the laser points for differentiated weighting and probabilistic modeling. During the positioning stage, a second local high-resolution map is dynamically switched and matched based on the estimated pose, and pose optimization is performed based on high-confidence regions, effectively overcoming the positioning ambiguity and drift problems caused by insufficient global map resolution, missing environmental textures, or repetitive structures. It can achieve real-time positioning with sub-centimeter precision and high robustness without relying on vision or manual marking. It is especially suitable for complex lighting and unmarked industrial environments.

[0050] To facilitate description and make the technical solutions of this disclosure easier to understand, the terminology of this disclosure will be explained before describing the technical solutions of this disclosure.

[0051] The mobile robot operating environment refers to the sum of the physical space and related elements in which the mobile robot performs its tasks, including static structures (such as walls, columns, shelves, workbenches, etc.), ground features (such as slope, material, etc.), and environmental information.

[0052] The target location refers to the critical work point where the mobile robot needs to accurately reach and complete a specific operation in the working environment. For example, the location where it needs to dock with equipment such as machine tools and workbenches (such as loading and unloading ports, assembly stations, charging interfaces, etc.).

[0053] This disclosure applies to mobile robot working environments with high-precision positioning requirements, and is especially suitable for harsh industrial scenarios where manual marking such as QR codes and reflectors are not allowed, and where lighting is complex and there is interference from dust / oil mist / strong light, such as automotive welding workshops, precision machining production lines, casting workshops, and cleanroom docking stations.

[0054] Figure 1 A schematic flowchart illustrating a high-precision positioning method for a mobile robot according to one embodiment of this disclosure is shown. Figure 1 The method shown includes steps S110 to S150.

[0055] In step S110, a local high-resolution localization map is constructed corresponding to each target location in the mobile robot's operating environment. The local high-resolution localization map refers to a refined raster map with increased resolution specifically constructed for each target location, which is used in subsequent high-resolution localization.

[0056] In step S120, the final pose of the mobile robot at the first moment is determined. The final pose is the pose after high-precision positioning at the first moment, and can also be called the prior pose.

[0057] In step S130, based on the final pose of the first moment, the pose of the mobile robot at the second moment is estimated to obtain the predicted pose of the mobile robot at the second moment, where the second moment is the moment after the first moment, for example, the first moment is the previous moment and the second moment is the current moment.

[0058] The pose at the second moment is estimated by superimposing odometry (i.e., the relative motion determined by the number of revolutions of the robot's left and right wheels and the relative attitude change obtained by the inertial measurement unit (IMU)) on the final pose at the first moment.

[0059] In step S140, in response to the high-resolution map switching command, a local high-resolution positioning map of the target position corresponding to the mobile robot at the second time moment is determined.

[0060] As one possible implementation, the high-resolution map switching command is generated when the Bézier module starts. The Bézier module is used to generate a Bézier curve and trigger the high-precision positioning process. A Bézier curve, also known as a Bézier curve, is a mathematical parametric curve that generates a smooth curve through control points. In this disclosure, the Bézier curve is used in the final stage of the mobile robot approaching the target position to replace traditional polylines or spline paths, generating a continuous, differentiable, acceleration-controllable, and shock-free motion trajectory, ensuring the stability and safety of the high-precision docking process. Exemplarily, the Bézier module acts as a client, sending a Bézier start message to the positioning module. The Bézier module indirectly indicates that the mobile robot needs high-precision positioning as it moves to the target position. Upon receiving this message, the positioning module immediately generates a high-resolution map switching command. Based on this command, the current global base point cloud map (e.g., with a resolution of 0.05m) is switched to a local high-resolution navigation map (with a resolution of 0.01m) corresponding to the target position. In one example, when the mobile robot moves to a distance of 2m from the platform, it calls the Bézier module to follow the Bézier curve to approach the platform. The machine number that needs to be located can be included in the Bezier activation message sent from the Bezier module to the positioning module. The positioning module can then find the corresponding local high-resolution positioning map based on the machine number and complete the high-resolution map switch.

[0061] In step S150, the estimated pose is optimized based on the lidar scanning data and the local high-resolution positioning map to obtain the final pose, which is then used as the high-precision positioning result of the mobile robot.

[0062] In optimizing the estimated pose, existing pose optimization algorithms (such as ScanMatch, ICP, Ceres, etc.) can be used, and there are no restrictions here.

[0063] Through steps S110 to S150, the mobile robot can be positioned with high precision at the target location and with conventional positioning at non-target locations.

[0064] Figure 2 A flowchart illustrating a local high-resolution positioning map construction method according to one embodiment of this disclosure is shown. Figure 2 The method shown includes steps S210 to S280.

[0065] In step S210, a basic point cloud map of the mobile robot's operating environment is constructed.

[0066] The base point cloud map is typically a global point cloud map of the mobile robot's operating environment. As one possible implementation, the pose information of the mobile robot and environmental information are collected by odometry, inertial measurement unit (IMU) and lidar sensors configured on the mobile robot, and then a two-dimensional grid-like base point cloud map is constructed.

[0067] In step S220, the set of valid calibration laser points corresponding to the target position is determined. The set of valid calibration laser points consists of multiple calibration laser points.

[0068] Figure 3 A flowchart illustrating an embodiment of the present disclosure of a method for determining an effective calibration laser point set is shown. Figure 3 The method shown includes steps S310 to S330.

[0069] In step S310, the lidar installed on the mobile robot located at the target position is controlled to collect multiple frames of lidar scanning data.

[0070] As one possible implementation, when the mobile robot moves to the target position, the mobile robot is controlled to remain stationary, and the lidar installed on the mobile robot is controlled to continuously collect N frames (e.g., N≥50) of lidar scanning data.

[0071] In step S320, the initial calibration laser point set corresponding to the lidar is determined based on multi-frame lidar scanning data.

[0072] Regarding step S320, in some embodiments of this disclosure, it may include, for example... Figure 4 Steps S3201 to S3205 are shown.

[0073] In step S3201, a first local high-resolution map is constructed with the mobile robot as the center and a resolution of the first resolution, which is higher than the resolution of the base point cloud map.

[0074] As one possible implementation, in constructing the first local high-resolution map, a map is constructed centered on the mobile robot with a range of 5... A 5m local high-resolution map with a first resolution of 0.01m.

[0075] In step S3202, multiple frames of LiDAR scanning data are projected onto a first local high-resolution map, and the number of times each grid in the first local high-resolution map is hit is determined.

[0076] As one possible implementation, the lidar scanning data is converted to a relative position in a two-dimensional coordinate system with the mobile robot as the origin, and projected onto a first local high-resolution map. The number of times each grid in the first local high-resolution map is hit is counted (i.e., the amount of data projected onto each corresponding grid).

[0077] In step S3203, the grid cells with the highest frequency are selected as candidate grid cells.

[0078] As one possible implementation, firstly, the grids with a hit count less than the first hit count threshold T1 (e.g., T1∈[0.5N,1N]) are filtered out. Then, among all the remaining grids after filtering, the M grids with the largest hit count (e.g., M≥100) are selected as candidate grids.

[0079] In step S3204, calibration laser points are obtained based on the candidate grid.

[0080] As one possible implementation, obtaining the calibration laser point based on the candidate grid includes: using the geometric center of the candidate grid as the coordinate value of the calibration laser point, and using the maximum intensity value hitting the candidate grid as the intensity value of the calibration laser point. Since each laser point hitting the grid corresponds to an intensity value, the maximum intensity value is the maximum value among all the intensity values ​​corresponding to the laser points hitting the grid.

[0081] In step S3205, multiple calibration laser points are combined to form an initial calibration laser point set.

[0082] Steps S3201 to S3205 construct a first local high-resolution map and statistically analyze the grid hit frequency to screen stable feature points. This effectively filters out noise and random interference, significantly improving the repeatability, environmental robustness, and geometric saliency of the calibrated laser points, thus laying a reliable feature foundation for subsequent high-precision positioning.

[0083] In step S330, the initial set of calibration laser points corresponding to the lidar is used as the set of valid calibration laser points corresponding to the target position.

[0084] Steps S310 to S330 can ensure that the calibration features are highly consistent with the actual working conditions, significantly improve the on-site adaptability and positioning reproduction accuracy of the local high-resolution positioning map, and avoid the matching failure problem caused by the deviation between offline mapping and online working conditions.

[0085] Figure 5 A flowchart illustrating another embodiment of the present disclosure of a method for determining an effective calibration laser point set is shown. Figure 5 The method shown includes steps S510 to S560.

[0086] In step S510, at least two lidars installed on the mobile robot located at the target position are controlled to collect multiple frames of lidar scanning data.

[0087] As one possible implementation, the mobile robot is equipped with two lidar sensors, located at the left front and right rear of the robot, respectively. Each lidar sensor acquires N consecutive frames of lidar scan data.

[0088] In step S520, based on multi-frame lidar scanning data, the initial calibration laser point set corresponding to each lidar is determined. Each lidar has a corresponding initial calibration laser point set. The method for determining the initial calibration laser point set is the same as in step S320, and will not be repeated here.

[0089] In step S530, the set of valid lines corresponding to each initial calibration laser point set is determined. The set of valid lines consists of at least one valid line, which is a line that can be fitted from the initial calibration laser point set.

[0090] Regarding step S530, in some embodiments of this disclosure, it may include, for example... Figure 6 Steps S5301 to S5304 are shown.

[0091] In step S5301, a Hough transform is performed based on the initial set of calibrated laser points to obtain at least one first candidate straight line. The Hough transform is a classic parameter space voting method used to detect predefined shapes (such as straight lines) from images or point clouds. As one possible implementation, the straight line detected from the initial set of calibrated laser points is used as the first candidate straight line.

[0092] In step S5302, the angular deviation between each first candidate line and the mobile robot is determined. As one possible implementation, in determining the angular deviation between each first candidate line and the mobile robot, the minimum angle between the direction of the first candidate line and the current orientation of the mobile robot (i.e., the direction of the central axis, typically the positive x-axis of the vehicle body) is taken as the angular deviation between the first candidate line and the mobile robot.

[0093] In step S5303, the first candidate straight line with an angle deviation less than the angle deviation threshold is selected as the valid straight line. As one possible implementation, the angle deviation threshold is set to ±T2, where T2∈[0°,5°].

[0094] In step S5304, a set of valid lines is obtained based on the valid lines.

[0095] Steps S5301 to S5304 effectively eliminate false straight lines caused by environmental clutter or calibration attitude interference, significantly improving the task relevance and docking scenario adaptability of the effective straight line set.

[0096] In step S540, the minimum distance between each lidar and each line in its corresponding set of valid lines is determined. For each lidar's set of valid lines, the distance from each lidar to each valid line in its set is calculated, and the minimum distance is extracted as the minimum distance for that lidar. Finally, the minimum distance for each lidar is obtained.

[0097] In step S550, the lidar with the smaller minimum distance is selected from at least two lidars and used as the calibration lidar.

[0098] As one possible implementation, the mobile robot is equipped with two lidars, and the minimum distances corresponding to the two lidars are a and b, respectively, where a > b. Therefore, the lidar corresponding to the minimum distance b is selected as the calibration lidar.

[0099] In step S560, the initial set of calibration laser points corresponding to the calibration lidar is taken as the set of valid calibration laser points corresponding to the target position.

[0100] Steps S510 to S560 describe the process of determining the effective set of calibration laser points based on at least two lidars. By comparing data from multiple lidars, the calibration lidar is selected based on its geometric proximity (minimum distance) to the effective straight line. This ensures that the selected lidar has the optimal observation angle and the richest features. It significantly improves the structural representativeness and matching robustness of the calibration laser points, providing a more stable and reliable feature basis for high-precision positioning.

[0101] In step S230, the weight values ​​of each calibration laser point are set according to whether the calibration laser point belongs to a point on a straight line and the relationship between the intensity value of the calibration laser point and the intensity value threshold.

[0102] Regarding step S230, in some embodiments of this disclosure, it may include, for example... Figure 7 Steps S2301 to S2304 are shown.

[0103] In step S2301, a Hough transform is performed based on the effective set of calibrated laser points to obtain a second candidate line. As one possible implementation, the line detected from the effective set of calibrated laser points is used as the second candidate line.

[0104] In step S2302, the weight value of the calibration laser point located on the second candidate straight line in the set of valid calibration laser points is set as the first weight value.

[0105] In step S2303, the weight value of the calibration laser point in the effective calibration laser point set that is not located on the second candidate straight line and whose intensity value is greater than or equal to the intensity value threshold is set as the second weight value. In one possible implementation, the intensity value threshold is the intensity value of the laser hitting a conventional rigid material.

[0106] In step S2304, the weight value of the calibration laser point in the effective calibration laser point set that is not located on the second candidate straight line and whose intensity value is less than the intensity value threshold is set as the third weight value. The first weight value is greater than the second weight value, and the second weight value is greater than the third weight value. In one possible implementation, the first weight value is less than or equal to 1, and the third weight value is greater than 0.9.

[0107] By combining geometric saliency (whether it lies on the second candidate line) and physical reflection characteristics (its relationship with the intensity threshold) in steps S2301 to S2304 to set three levels of differentiated weight values, map matching can be more focused on high-reliability feature points. This implementation can significantly improve the accuracy, robustness, and anti-interference ability of positioning.

[0108] In step S240, a local base map is extracted from the base point cloud map with the target location as the center.

[0109] As one possible implementation, a range of 3 is captured in the base point cloud map. A 3m map is used as a local base point cloud map (resolution 0.05m).

[0110] In step S250, the local base map is segmented to obtain a second local high-resolution map.

[0111] As one possible implementation, segmenting a local base map to obtain a second local high-resolution map includes: segmenting each grid cell in the local base map into i A second local high-resolution map is obtained by dividing the base point cloud map into i grids, where i is the ratio of the base point cloud map's resolution to the first resolution. For example, if the base point cloud map's resolution is 0.05m and the first resolution is 0.01m, then i = 0.05 / 0.01 = 5, meaning each grid in the local base map is equally divided into 5 grids. 5 = 25 grid cells.

[0112] In step S260, the weight values ​​of each calibrated laser point are mapped to the second local high-resolution map to obtain the weight coefficients of each grid in the second local high-resolution map.

[0113] As one possible implementation, the weight values ​​of each calibrated laser point are mapped to a second local high-resolution map to obtain the weight coefficients of each grid in the second local high-resolution map. This includes marking grids belonging to obstacles in the local base map as obstacle grids and grids not belonging to obstacles as idle grids. The i values ​​obtained by segmenting each grid in the local base map are then... i grids form a grid set, resulting in multiple grid sets. Based on the coordinates of each calibrated laser point, the weight value of each calibrated laser point is mapped to the corresponding grid in the second local high-resolution map, serving as the weight coefficient of the corresponding mapped grid in the second local high-resolution map. If the unmapped grid in the second local high-resolution map belongs to the same grid set as the mapped grid, the weight coefficient of the unmapped grid in the second local high-resolution map is set to 0. If the unmapped grid in the second local high-resolution map does not belong to the same grid set as the mapped grid, the weight coefficient of the unmapped grid belonging to obstacle grids is set to a fourth weight value, which is less than the third weight value. The weight coefficient of the unmapped grid belonging to idle grids is set to 0. For example, the fourth weight value can be set to 0.9. Figure 10 As shown, in the rasterized local base map, obstacle grids are represented by black, and free grids by white. In the calibration of laser point markers, each grid in the local base map is first divided into 5... Five grids are used, and the corresponding grid labels remain unchanged (black indicates obstacle grids, and white indicates free grids). In the second local high-resolution map, the weight coefficient of the mapped grid is the weight value of the calibrated laser point (represented by black), and it belongs to the same grid set as the mapped grid (belonging to the same 0.05m). The weight coefficients of unmapped graticules (0.05m large grid) are reset to 0 (represented by white) to indicate idle graticules. This implementation effectively improves the representation accuracy and robustness of obstacle boundary regions in the second local high-resolution map through a refined weight mapping and classification assignment strategy, while taking into account both computational efficiency and environmental semantic consistency.

[0114] In step S270, based on the weighting coefficients, the probability density of the second local high-resolution map is calculated to obtain the probability density value of each grid cell in the second local high-resolution map. Probability density is a function describing the distribution characteristics of a continuous random variable. In this disclosure, probability density refers to a non-negative real value assigned to each grid cell in the local high-resolution positioning map, serving as the basis for subsequent pose optimization.

[0115] Regarding step S270, in some embodiments of this disclosure, it may include, for example... Figure 8 Steps S2701 to S2705 are shown.

[0116] In step S2701, the nearest obstacle grid cell corresponding to each grid cell in the second local high-resolution map is determined. An obstacle grid cell is a grid cell in the second local high-resolution map whose weight coefficient is not 0. The nearest obstacle grid cell can be understood as the obstacle grid cell closest to each other.

[0117] In step S2702, the Euclidean distance between each grid cell and the corresponding nearest obstacle grid cell is calculated.

[0118] In step S2703, when the Euclidean distance value is less than or equal to 0, the probability density value of each grid is set to the weight coefficient of the nearest obstacle grid corresponding to each grid.

[0119] In step S2704, when the Euclidean distance is greater than 0 and less than or equal to the maximum Euclidean distance, the probability density value of each grid is set to be the product of the weight coefficient of the nearest obstacle grid corresponding to each grid and the Gaussian decay function.

[0120] In step S2705, when the Euclidean distance is greater than the maximum Euclidean distance, the probability density value of each grid cell is set to 0.

[0121] The probability density calculation method corresponding to this embodiment can be expressed by the following formula: in, h This represents the probability density value of the current grid cell. This represents the weight coefficient of the nearest obstacle grid cell to the current grid cell. This represents the Euclidean distance between the current grid cell and the nearest obstacle grid cell, expressed in grid cells. This indicates the maximum Euclidean distance set. This indicates the raster resolution, which is 0.01m / cell. This represents the standard deviation of the Gaussian model. This represents the Gaussian decay function.

[0122] In step S280, the probability density value is recorded in the second local high-resolution map to obtain a local high-resolution positioning map.

[0123] Steps S210 to S280 differentiate and weight the laser points by fusing geometric structural saliency (line assignment) and physical reflection characteristics (intensity threshold), and map the weights to the subdivided high-resolution grid. Simultaneously, a local high-resolution positioning map with continuous probabilistic semantics is generated by combining a distance attenuation model. This significantly enhances the map's ability to represent and distinguish key positioning features (such as machine edges). This implementation can effectively improve the accuracy of subsequent positioning without relying on manual labeling and visual information.

[0124] Figure 9 A flowchart illustrating a pose optimization method according to one embodiment of this disclosure is shown. Figure 9 The method shown includes steps S910 to S960.

[0125] In step S910, the lidar scanning data is first filtered based on the local high-resolution positioning map to obtain an initial set of lidar points within the range of the local high-resolution positioning map.

[0126] As one possible implementation, the local high-resolution positioning map is used as the bounding box. The LiDAR scanning data (after adaptive voxel filtering) is converted to the map coordinate system corresponding to the local high-resolution positioning map. Only the LiDAR points whose coordinates fall within the bounding box are retained to form an initial set of LiDAR points.

[0127] In step S920, the initial laser point set is filtered a second time to remove laser points that do not have an obstacle grid in the neighborhood, thus obtaining an effective laser point set.

[0128] As one possible implementation, for each initial laser point in the initial laser point set, firstly, query the neighborhood range (e.g., a 3×3 or 5×5 grid window) corresponding to the initial laser point in the second local high-resolution map; secondly, determine whether the probability density values ​​of all grids within the neighborhood range are all 0 (a probability density value of 0 indicates that it does not belong to an obstacle grid). If so, the initial laser point is determined to be an invalid laser point; otherwise, the initial laser point is determined to be a valid laser point. Finally, a valid laser point set is constructed based on all valid laser points.

[0129] In step S930, the estimated pose is optimized using a variety of pose optimization algorithms to obtain multiple candidate optimized poses.

[0130] As one possible implementation, the predicted pose is optimized using pose optimization algorithms such as ScanMatch, ICP, and Ceres to obtain candidate optimized poses output by each pose optimization algorithm.

[0131] In step S940, based on each candidate optimized pose, the mapping position of each laser point in the effective laser point set to the local high-resolution positioning map is determined.

[0132] Under different candidate optimized poses, the mapping positions of each laser point in the effective laser point set to the local high-resolution localization map are different. Therefore, each candidate optimized pose has its corresponding mapping position.

[0133] In step S950, the score of each candidate optimized pose is calculated based on the probability density value of the grid corresponding to each mapping position.

[0134] As one possible implementation, the score of each candidate optimized pose is calculated based on the probability density value of the grid corresponding to the mapped position. This includes using the ratio of the sum of the probability density values ​​of the grids corresponding to each mapped position to the number of effective lasers as the score of the candidate optimized pose. The number of effective lasers is the number of laser points in the set of effective laser points that are mapped to grids with a probability density value greater than 0.

[0135] In step S960, the candidate optimized pose with the highest score is selected as the final pose.

[0136] Steps S910 to S960 employ multiple pose optimization algorithms working collaboratively, and quantify and select the best pose based on the probability density in the local high-resolution positioning map. This avoids the possibility of a single algorithm getting stuck in local minima or malconvergence due to noise interference, significantly improving the reliability, consistency, and environmental adaptability of the high-precision positioning results.

[0137] According to any of the above embodiments, this disclosure also provides a high-precision positioning device 100 for a mobile robot. Figure 11 This is a schematic block diagram of a high-precision positioning device 100 for a mobile robot according to one embodiment of this disclosure. Figure 11As shown, the high-precision positioning device 100 for mobile robots includes a map building module 110, a first-moment final pose determination module 120, a second-moment predicted position estimation module 130, a map selection module 140, and a second-moment final pose determination module 150. The map building module 110 is used to construct local high-resolution positioning maps corresponding to each target location in the mobile robot's operating environment. The construction process includes: constructing a basic point cloud map of the mobile robot's operating environment; determining the set of valid calibration laser points corresponding to the target location, which consists of multiple calibration laser points; setting the weight value of each calibration laser point based on whether it belongs to a straight line and the relationship between the intensity value and the intensity threshold; extracting a local basic map from the basic point cloud map centered on the target location; segmenting the local basic map to obtain a second local high-resolution map; and mapping the weight values ​​of each calibration laser point to the second local high-resolution map to obtain the weight coefficients of each grid in the second local high-resolution map. Based on weighting coefficients, probability density is calculated on the second local high-resolution map to obtain the probability density value of each grid in the second local high-resolution map. The probability density values ​​are recorded in the second local high-resolution map to obtain a local high-resolution positioning map. The first-time final pose determination module 120 is used to determine the final pose of the mobile robot at the first time. The second-time predicted position estimation module 130 is used to estimate the pose of the mobile robot at the second time based on the final pose at the first time, obtaining the predicted pose of the mobile robot at the second time, where the second time is a time after the first time. The map selection module 140 is used to determine the local high-resolution positioning map of the target position corresponding to the mobile robot at the second time in response to a high-resolution map switching command. The second-time final pose determination module 150 is used to optimize the predicted pose based on LiDAR scanning data and the local high-resolution positioning map to obtain the final pose, and uses the final pose as the high-precision positioning result of the mobile robot at the second time.

[0138] According to further embodiments of this disclosure, an electronic device is also provided. Figure 12This diagram illustrates a schematic block diagram of an electronic device employing a processor-based hardware implementation according to an embodiment of the present disclosure. The hardware structure of the electronic device of the present disclosure can be implemented using a bus architecture. The bus architecture can include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the hardware. Bus 1100 connects various circuits including one or more processors 1200, memory 1300, and / or hardware modules. Bus 1100 can also connect various other circuits 1400 such as peripheral devices, voltage regulators, power management circuits, external antennas, etc. Bus 1100 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, only one connecting line is used in this figure, but this does not imply that there is only one bus or one type of bus. Memory 1300 stores a computer program, and when processor 1200 executes the computer program, processor 1200 is able to perform the following processes. A local high-resolution localization map corresponding to each target location in the mobile robot's operating environment is constructed. The final pose of the mobile robot at a first moment is determined. Based on the final pose at the first moment, the pose of the mobile robot at a second moment is estimated to obtain the predicted pose of the mobile robot at the second moment, where the second moment is a moment after the first moment. In response to a high-resolution map switching command, a local high-resolution localization map of the target location corresponding to the mobile robot at the second moment is determined. Based on the LiDAR scanning data and the local high-resolution localization map, the predicted pose is optimized to obtain the final pose, which is used as the high-precision localization result of the mobile robot at the second moment. Constructing the local high-resolution localization map corresponding to the target location includes: constructing a basic point cloud map of the mobile robot's operating environment; determining a set of valid calibration LiDAR points corresponding to the target location, the set of valid calibration LiDAR points consisting of multiple calibration LiDAR points; setting the weight value of each calibration LiDAR point based on whether it belongs to a straight line and the relationship between the intensity value and the intensity value threshold of the calibration LiDAR point; and extracting a local basic map from the basic point cloud map with the target location as the center. The local base map is segmented to obtain a second local high-resolution map. The weight values ​​of each calibrated laser point are mapped to the second local high-resolution map to obtain the weight coefficients of each grid in the second local high-resolution map.Based on the weighting coefficients, probability density is calculated on the second local high-resolution map to obtain the probability density value of each grid cell in the second local high-resolution map. The probability density values ​​are recorded in the second local high-resolution map to obtain the local high-resolution positioning map.

[0139] This disclosure also provides a readable storage medium storing a computer program that, when executed by a processor, is used to implement the methods described above. A "readable storage medium" can be any means capable of containing, storing, communicating, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples of a readable storage medium include: an electrical connection with one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM), etc.

[0140] This disclosure also provides a computer program product, the methods of which can be implemented wholly or partially through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented wholly or partially as a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed, all or part of the processes or functions of this disclosure are performed.

[0141] Computer programs or instructions can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any available medium capable of access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video optical disc; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or it can include both volatile and non-volatile types of storage media.

[0142] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0143] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0144] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0145] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0146] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment / mode or example, which are included in at least one embodiment / mode or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0147] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A high-precision positioning method for a mobile robot, characterized in that, include: Construct local high-resolution localization maps corresponding to the locations of each target in the mobile robot's operating environment; Determine the final pose of the mobile robot at the first moment; Based on the final pose at the first moment, the pose of the mobile robot at the second moment is estimated to obtain the predicted pose of the mobile robot at the second moment, where the second moment is the moment after the first moment. In response to the high-resolution map switching command, a local high-resolution positioning map of the target location of the mobile robot at the second moment is determined; as well as Based on the lidar scanning data and the local high-resolution positioning map, the estimated pose is optimized to obtain the final pose, which is then used as the high-precision positioning result of the mobile robot at the second time step. The construction of a local high-resolution positioning map corresponding to the target location includes: A basic point cloud map for constructing the operating environment of mobile robots; Determine the set of valid calibration laser points corresponding to the target location, wherein the set of valid calibration laser points consists of multiple calibration laser points; The weight values ​​of each calibration laser point are set based on whether the calibration laser point belongs to a straight line and the relationship between the intensity value of the calibration laser point and the intensity value threshold. Using the target location as the center, a local base map is extracted from the base point cloud map; The local base map is segmented to obtain a second local high-resolution map; The weight values ​​of each calibrated laser point are mapped to the second local high-resolution map to obtain the weight coefficients of each grid in the second local high-resolution map. Based on the weighting coefficients, the probability density of the second local high-resolution map is calculated to obtain the probability density value of each grid cell in the second local high-resolution map. The probability density value is recorded in the second local high-resolution map to obtain the local high-resolution positioning map.

2. The method as described in claim 1, characterized in that, Determining the set of valid calibration laser points corresponding to the target location includes: Control the lidar installed on the mobile robot located at the target position to collect multiple frames of lidar scanning data; Based on the multi-frame lidar scanning data, the initial calibration laser point set corresponding to the lidar is determined; and The initial set of calibration laser points corresponding to the lidar is taken as the effective set of calibration laser points corresponding to the target position.

3. The method as described in claim 2, characterized in that, Determining the initial calibration laser point set corresponding to the lidar includes: Construct a first local high-resolution map centered on the mobile robot, with a resolution of a first resolution, wherein the first resolution is higher than the resolution of the base point cloud map; The multi-frame LiDAR scanning data is projected onto the first local high-resolution map, and the number of times each grid in the first local high-resolution map is hit is determined. Select the grid cells with the highest frequency as candidate grid cells; Based on the candidate grid, calibration laser points are obtained; and The initial calibration laser point set is composed of multiple calibration laser points.

4. The method as described in claim 3, characterized in that, Based on the candidate grid, the calibration laser points are obtained, including: The geometric center of the candidate grid is used as the coordinate value of the calibration laser point, and the maximum intensity value hitting the candidate grid is used as the intensity value of the calibration laser point.

5. The method as described in claim 1, characterized in that, Determining the set of valid calibration laser points corresponding to the target location includes: Control at least two lidars mounted on a mobile robot located at the target position to collect multiple frames of lidar scanning data; Based on the multi-frame lidar scanning data, the initial calibration laser point set corresponding to each lidar is determined. Determine the set of valid straight lines corresponding to each initial set of calibration laser points; Determine the minimum distance between each lidar and each line in the corresponding set of valid lines; Select the lidar with the smaller minimum distance from the at least two lidars as the calibration lidar; and The initial set of calibration laser points corresponding to the calibration lidar is taken as the effective set of calibration laser points corresponding to the target position.

6. The method as described in claim 4, characterized in that, The weight values ​​for each calibration laser point are set, including: Based on the set of effective calibrated laser points, a Hough transform is performed to obtain a second candidate straight line; The weight value of the calibration laser point located on the second candidate straight line in the set of effective calibration laser points is set as the first weight value; The weight value of the calibration laser point in the effective calibration laser point set that is not located on the second candidate line and whose intensity value is greater than or equal to the intensity value threshold is set as the second weight value; and The weight value of the calibration laser point that is not located on the second candidate line and whose intensity value is less than the intensity value threshold in the set of effective calibration laser points is set as the third weight value, wherein the first weight value is greater than the second weight value, and the second weight value is greater than the third weight value.

7. The method as described in claim 3, characterized in that, The local base map is segmented to obtain a second local high-resolution map, including: Divide each grid in the local base map into i i grid cells are used to obtain a second local high-resolution map, where i is the ratio of the resolution of the base point cloud map to the first resolution.

8. The method as described in claim 7, characterized in that, The weight values ​​of each calibrated laser point are mapped to the second local high-resolution map to obtain the weight coefficients of each grid cell in the second local high-resolution map, including: In the local base map, grids that belong to obstacles are marked as obstacle grids, and grids that do not belong to obstacles are marked as empty grids; i is obtained by segmenting each grid in the local base map i grid cells form a grid set, resulting in multiple grid sets; Based on the coordinate values ​​of each calibrated laser point, the weight value of each calibrated laser point is mapped to the corresponding grid in the second local high-resolution map, which serves as the weight coefficient of the corresponding mapped grid in the second local high-resolution map. If the unmapped rasters and mapped rasters in the second local high-resolution map belong to the same raster set, the weight coefficient of the unmapped rasters in the second local high-resolution map is set to 0; and In the case where the unmapped graticule and the mapped graticule in the second local high-resolution map do not belong to the same graticule set, the weight coefficient of the unmapped graticule belonging to the obstacle graticule is set to the fourth weight value, the fourth weight value is less than the third weight value, and the weight coefficient of the unmapped graticule belonging to the idle graticule is set to 0.

9. An electronic device, characterized in that, include: The memory stores execution instructions; as well as A processor that executes the execution instructions stored in the memory, causing the processor to perform the high-precision positioning method for a mobile robot as described in any one of claims 1 to 8.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the high-precision positioning method for mobile robots as described in any one of claims 1 to 8.

Images