Method and device for determining automatic driving map data and readable storage medium

Abstract

The invention provides an automatic driving map data determination method and device and a readable storage medium. The automatic driving map data determination method comprises the steps that global basic data and local operation data of a shutdown period are acquired; acquiring local grid topographic data and first operation boundary feature data according to the local operation data; according to the global basic data, obtaining a point cloud map, an initial grid map and global second operation boundary feature data; determining a target grid map covering the whole domain of the target operation scene according to the plurality of initial grid maps; determining global topographic data according to the local grid topographic data, the preprocessed point cloud map and the target grid map; determining working boundary reference data according to the first working boundary feature data and the second working boundary feature data; and generating automatic driving map data based on the global topographic data and the working boundary reference data. Through double-source acquisition and low-consumption transmission, efficient updating of the mine map is realized.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and more specifically, to a method, apparatus, and readable storage medium for determining autonomous driving map data. Background Technology

[0002] In related technologies, open-pit mines are characterized by harsh environments, lack of fixed roads and signage, frequent updates to work boundaries, and poor network transmission conditions. Therefore, the special environment of mining areas places high demands on the safety performance and operational efficiency of unmanned operations. Because the development of autonomous driving maps within the mining area relies on dedicated map-collecting vehicles, the entire mining environment must be shut down before map data collection can be completed, impacting operational efficiency. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art or related technologies.

[0004] Therefore, the first aspect of this application proposes a method for determining autonomous driving map data.

[0005] The second aspect of this application proposes an apparatus for determining autonomous driving map data.

[0006] A third aspect of this application proposes another apparatus for determining autonomous driving map data.

[0007] The fourth aspect of this application proposes a readable storage medium.

[0008] In view of this, the first aspect of this application provides a method for determining autonomous driving map data, comprising: acquiring full-domain basic data of a target work scenario during downtime and partial work data of the target work scenario during operating hours; acquiring local grid terrain data corresponding to the local work data and first work boundary feature data within the local area based on the partial work data; acquiring a point cloud map, initial grid maps of multiple local areas within the full domain, and second work boundary feature data within the full domain based on the full-domain basic data; determining a target grid map covering the entire target work scenario based on the multiple initial grid maps; determining full-domain terrain data based on the partial grid terrain data, the preprocessed point cloud map, and the target grid map; determining work boundary reference data based on the first work boundary feature data and the second work boundary feature data; and generating autonomous driving map data for the target work scenario based on the full-domain terrain data and the work boundary reference data.

[0009] The target operation scenario provided in this application refers to a specific application scenario that requires the generation of autonomous driving map data. Its core characteristics include harsh environment, frequent terrain changes, and high requirements for the timeliness and accuracy of map updates. A typical scenario is an open-pit mine, which has no fixed roads and signs. The operation mode revolves around "loading-transporting-discharging" and the number of on-site personnel must be strictly limited.

[0010] The shutdown period refers to a specific time period in the target work scenario during which production and operation are suspended. It does not mean that all activities are completely stopped, but rather that core production operations are suspended to carry out non-production-related necessary work, including shift change periods, inspection periods, or maintenance periods. During this period, network communication conditions are relatively stable, which is suitable for large-scale data collection.

[0011] Full-domain basic data refers to the basic data collected during the downtime of the target operation scenario, covering the entire scope of the scenario. It is the core foundation for building autonomous driving maps. Data types include raw point cloud files, LiDAR calibration parameters, etc., which can comprehensively reflect the overall terrain, spatial layout and other basic information of the scenario.

[0012] The operating period refers to the time during which production and operation activities are carried out normally in the target operation scenario. For open-pit mine scenarios, this refers to the time during which operations are carried out around the "loading-transporting-discharging" process. During this period, it is necessary to collect data from key areas without affecting production efficiency.

[0013] Local operation data refers to targeted data collected during the operation period of the target operation scenario, focusing on the current actual operation area. Data types include raw point cloud files after being filtered by preset rules, LiDAR calibration parameters, etc., which are used to supplement and update the real-time terrain and boundary change information of the operation area.

[0014] Local raster terrain data refers to local terrain data presented in raster form after processing local operational data. The raster format can quickly reflect key features of local terrain such as elevation and slope, facilitating efficient transmission and fusion processing.

[0015] The first operational boundary feature data refers to the feature information related to the boundary of the local operational area extracted from the local operational data. Specifically, it includes features such as the terrain edge and operational point boundary of the local operational area, which can accurately reflect the real-time boundary changes of the current operational area.

[0016] Point cloud maps are maps generated based on the original point cloud files in the global basic data. They consist of a large number of discrete three-dimensional point coordinates and can realistically restore the three-dimensional spatial form of the target operation scene. They are an important basis for subsequent terrain data construction and boundary extraction.

[0017] An initial raster map refers to a rasterized map that is generated based on global basic data and covers multiple local areas within the global domain. Each initial raster map corresponds to a partition in the global domain and contains the basic terrain information of that partition, but has not yet undergone global integration and accuracy optimization.

[0018] The second operational boundary feature data refers to the boundary feature information extracted from the overall basic data, covering the entire range of the target operational scenario. Specifically, it includes features such as terrain boundaries and fixed facility boundaries within the entire range, and can provide the overall boundary framework of the scenario.

[0019] A target raster map is a complete raster map that covers the entire target operation scene, formed by calculating terrain parameters, matching edges, and stitching together multiple initial raster maps. It has the characteristics of complete terrain information and consistent regional connection, and is the core carrier for constructing full-domain terrain data.

[0020] Preprocessed point cloud map refers to the map obtained after thinning and slicing the point cloud map. Thinning can remove redundant point data and reduce the amount of data, while slicing can divide the point cloud map into multiple tiles according to preset rules, which facilitates storage, transmission and subsequent fusion processing.

[0021] Global terrain data refers to complete terrain data that covers the entire range of the target operation scene after integrating local raster terrain data, preprocessed point cloud maps, and target raster maps. It includes the overall terrain framework of the entire region and supplements the real-time terrain details of the operation area, which can comprehensively and accurately reflect the terrain characteristics of the scene.

[0022] Work boundary reference data refers to boundary reference information obtained by fusing and verifying the first work boundary feature data and the second work boundary feature data. It is used to guide autonomous driving operations and has the characteristics of being continuous, complete, accurate and reliable, and can clearly define the work range and safety boundaries.

[0023] Autonomous driving map data refers to the final map data generated based on global terrain data and work boundary reference data, which meets the requirements for use by autonomous driving systems. It includes key content such as 3D terrain information, work boundary markings, road network topology, and work point markings, and can directly support the safe and efficient operation of autonomous driving equipment in the target work scenario.

[0024] In summary, this application addresses the issues of reliance on dedicated data collection vehicles, frequent downtime, high network transmission pressure, and insufficient map accuracy and timeliness in the target operational scenario for autonomous driving map updates. It achieves efficient map data generation without additional equipment costs, without impacting production operations, and adaptable to harsh network environments. Specifically, the process first achieves high efficiency and continuity through "dual-time period data collection," avoiding efficiency losses caused by downtime. This application collects data in different time periods. During downtime, it collects basic data across the entire area. At this time, the target operational scenario suspends core production operations, ensuring no impact on production progress due to data collection. Furthermore, a stable high-speed network can be used to upload large-scale data during this period. During operational periods, it simultaneously collects local operational data, focusing only on the current operational area, without needing to cover the entire area. The collection process runs parallel to the production and operation process, without interfering with core operations such as "loading-transportation-arrangement." This dual-time period collection mode, combining basic data across the entire area with local supplementation, ensures both comprehensive map data coverage and real-time updates of operational area data. It avoids downtime losses at the collection stage and improves the collaborative efficiency between production operations and map updates.

[0025] Secondly, the accuracy and reliability of map data are improved through "multi-source data fusion". This application collects two different types of data: global basic data and local operation data. From the global basic data, point cloud maps, initial raster maps, and second operation boundary feature data are obtained to construct the overall terrain and boundary framework of the scene. From the local operation data, local raster terrain data and first operation boundary feature data are obtained to supplement the real-time terrain details and boundary changes of the operation area. Then, through multiple rounds of data fusion, multiple initial raster maps are stitched together to form a target raster map. The local raster terrain data, the preprocessed point cloud map, and the target raster map are fused to generate global terrain data. The two types of operation boundary feature data are fused to generate working boundary reference data. The multi-source data mutually verify and supplement each other, effectively eliminating errors and abnormal information from a single data source, improving the accuracy and reliability of the map data, and ensuring that the map data can truly reflect the terrain and boundary conditions of the target operation scene.

[0026] Finally, layered data processing and precise transmission reduce network transmission pressure and adapt to harsh network environments. Targeted operational scenarios (such as open-pit mines) are typically located in remote areas, relying on self-built base stations for communication, resulting in poor network transmission conditions. This application implements targeted optimizations in data processing and transmission. Specifically, local operational data is first processed into local raster terrain data using the computing module of the unmanned vehicle, and then only data relevant to the current operation is transmitted using a preset range upload method, eliminating irrelevant and redundant data, significantly reducing the amount of data transmitted. Point cloud maps are thinned and pre-processed by tiling to reduce data storage and transmission pressure. The initial raster map is first partitioned and calculated before stitching, avoiding the inefficiency caused by directly processing the entire raster data. Through this series of layered processing and precise transmission strategies, the dependence on network bandwidth is reduced, enabling efficient data transmission even in environments with poor network conditions and ensuring the timeliness of map data updates.

[0027] In some technical solutions of this application, the overall basic data includes the original point cloud file and the lidar calibration parameters, and the local operation data includes the original point cloud file and the lidar calibration parameters after being filtered by preset rules; the downtime is the shift change time, inspection time or maintenance time of the target operation scenario. The overall basic data is collected at a preset speed and route and then uploaded through a high-speed network, and the local operation data is collected synchronously with the operation process of the target operation scenario.

[0028] In this technical solution, the raw point cloud file refers to the original data file composed of a large number of discrete 3D point coordinates acquired by acquisition devices such as LiDAR. Each 3D point contains spatial location information (X, Y, Z coordinates) and can directly reflect the spatial form of the acquisition area, such as terrain and features. It is the core raw data for generating point cloud maps and raster maps. LiDAR calibration parameters are key parameters used to calibrate the acquisition accuracy of LiDAR, including LiDAR installation position offset parameters, angle deviation parameters, and ranging accuracy correction parameters. These parameters can correct system errors during the LiDAR acquisition process, ensuring that the acquired raw point cloud file accurately corresponds to the actual spatial location. Preset rule filtering refers to the pre-set rules for filtering local operation data based on the operational needs and data processing requirements of the target operation scenario. These filtering rules include data validity rules (removing invalid data acquired abnormally by sensors), correlation rules (retaining data directly related to the current operation area), and accuracy threshold rules (retaining data that meets preset accuracy requirements). The aim is to improve the quality and relevance of local operation data. Shift change periods refer to the time when workers in the target work scenario hand over responsibilities. During this period, production operations are temporarily suspended, there is no large-scale movement of equipment, and the environment is relatively stable, making it suitable for large-scale data collection. Inspection periods refer to the time periods for regular inspections and maintenance of production equipment and infrastructure in the target work scenario. During this period, production operations are suspended, facilitating the movement of data collection equipment within the scenario without affecting inspection work. Maintenance periods refer to the time periods for repairing faulty equipment and damaged facilities in the target work scenario. During this period, production operations in the relevant areas are suspended, allowing for the use of maintenance gaps to conduct comprehensive basic data collection, achieving coordination between data collection and equipment maintenance. Preset speed and route refer to the pre-set movement speed and route of the data collection equipment for comprehensive basic data collection. The preset speed must balance collection accuracy and efficiency, avoiding incomplete or inaccurate data collection due to excessive speed. The preset route must cover the entire area of ​​the target work scenario to ensure comprehensive data coverage. High-speed networks refer to networks with high transmission rates and good stability that are available during downtime in the target work scenario. These are typically dedicated communication networks built by the scenario itself, capable of meeting the needs for rapid and complete transmission of basic data across the entire domain (such as large-capacity raw point cloud files). Operational processes refer to the standardized procedures for carrying out core production activities in the target work scenario. For open-pit mines, this is the "loading-transporting-discharging" process (the complete process of loading, transporting, and discharging ore). The collection of localized operational data needs to be synchronized with this process, collecting data from the corresponding areas simultaneously when the equipment performs loading, transporting, and discharging operations.

[0029] This application further refines the technical solution by clearly defining the specific types of overall basic data and localized operational data, the specific scope of downtime periods, and the methods for collecting and uploading both types of data. It clarifies data types and filtering rules, ensuring the effectiveness and relevance of the collected data. Raw point cloud files and LiDAR calibration parameters are core data for generating high-precision maps; preset filtering rules can eliminate invalid and redundant information in localized operational data, improving data processing efficiency. Secondly, it clarifies the specific scenarios of downtime periods, providing an operable time window for overall basic data collection and avoiding conflicts between collection time and production operations. Simultaneously, by standardizing collection and uploading methods, preset speeds and routes ensure the overall coverage and accuracy of overall basic data collection, high-speed network uploading ensures the transmission efficiency of large-capacity overall basic data, and synchronous collection with operational processes ensures the real-time nature of localized operational data.

[0030] In some technical solutions of this application, local grid terrain data corresponding to local operation data is obtained based on local operation data, including: calculating the local operation data through the computing power module of the unmanned operation vehicle in the target operation scene to determine the local grid terrain data; the local grid terrain data is transmitted to the computing end using a preset range upload method; wherein, the preset range upload method is to upload local grid terrain data related to the current operation of the target operation scene.

[0031] In this technical solution, unmanned operating vehicles refer to vehicles that have been modified for unmanned operation and have the ability to drive and operate autonomously. In the open-pit mine scenario, they are unmanned mining trucks. Their core components include domain controllers, gateway controllers, vehicle controllers, sensors that can support map acquisition (such as LiDAR), gateways, etc., which can complete production operations and data acquisition tasks without human driving.

[0032] The computing module refers to a hardware module integrated on an unmanned operating vehicle that has data processing and computing capabilities. It can quickly process and analyze the local operating data collected by the vehicle, and can generate local grid terrain data without transmitting the raw data to a remote computing device, thereby improving data processing efficiency.

[0033] Solving refers to the process of analyzing, transforming, and calculating local operation data through a preset algorithm model and calculation logic, converting discrete raw point cloud files and lidar calibration parameters into structured, rasterized local raster terrain data. The solving process includes key steps such as coordinate transformation, elevation calculation, and raster division.

[0034] The preset range upload method refers to the pre-defined data upload range rules, which only upload local raster terrain data that is directly related to the current operation activities of the target work scenario, and remove redundant terrain data that is not related to the current operation. The core purpose is to reduce the amount of data transmitted and reduce network transmission pressure.

[0035] The computing end refers to the core computing device that receives and processes uploaded data. In this application, it is the server computing unit, which has powerful data analysis, fusion and storage capabilities. It can integrate and process the global basic data and the uploaded local raster terrain data to generate key data such as target raster maps and global terrain data.

[0036] This application further improves the technical solution by clarifying the calculation subject, calculation method, and upload method of local raster terrain data. It enables local calculation using the computing module built into the unmanned operating vehicle, generating local raster terrain data without transmitting the original local operation data to a remote computing terminal. This reduces bandwidth consumption and transmission latency of the original data transmission, improving the real-time performance of data processing. Furthermore, the use of a preset range upload method transmits only data relevant to the current operation, significantly reducing the amount of uploaded data, effectively reducing network transmission pressure, adapting to the harsh network environment of the target operation scenario, and avoiding data loss or update delays due to network congestion or poor transmission quality. Simultaneously, the calculation and upload processes are performed synchronously with the operation flow during operating hours, without affecting production efficiency.

[0037] In some technical solutions of this application, the first operation boundary feature data is the terrain edge and operation point boundary feature information of the local operation area, and the second operation boundary feature data is the terrain boundary and fixed facility boundary feature information of the entire area; determining the work boundary reference data based on the first operation boundary feature data and the second operation boundary feature data includes: performing fusion verification on the first operation boundary feature data and the second operation boundary feature data, removing abnormal feature information, and obtaining the work boundary reference data.

[0038] In this technical solution, terrain edges refer to the boundary lines formed by changes in terrain undulations, such as the boundary between hillsides and flat land, and the boundary between depressions and the ground. These reflect changes in the spatial morphology of the terrain and are an important component of the operational boundary. Operational point boundaries refer to the boundary range of specific operational locations within the target operational scenario, such as the boundaries of ore loading points, unloading points, and transportation routes. These directly relate to the scope definition and safety management of autonomous driving operations. Fixed facility boundaries refer to the boundary range of fixed facilities and equipment within the target operational scenario, such as the boundaries of mine walls, equipment parking areas, office areas, and roads. These facility boundaries are relatively stable and serve as an important basis for constructing the overall boundary framework of the scenario. Fusion verification refers to the process of integrating, comparing, and verifying the feature data of the first and second operational boundaries. Through preset verification rules (such as spatial location consistency verification, boundary continuity verification, and data credibility verification), the consistency and accuracy of the two types of data are determined. Abnormal feature information refers to feature information in the two types of operational boundary feature data that does not conform to the actual scenario, contains errors or contradictions, such as false terrain edges caused by sensor failure, boundary overlap or breakage caused by data acquisition deviation, and invalid fixed facility boundaries caused by facility dismantling and failure to update.

[0039] This application clarifies the specific content of two types of operational boundary feature data and the generation method of operational boundary reference data, thereby defining the specific types of boundary feature data. This ensures that the boundary information extracted from the two data sources is targeted and complementary. Terrain edges and operational point boundaries reflect dynamically changing operational boundaries, while fixed facility boundaries provide a stable overall boundary framework. The combination of these two types of data achieves comprehensive coverage of boundary information. Simultaneously, through a fusion verification process, abnormal feature information can be effectively eliminated, addressing potential errors and unreliability issues associated with single data sources and ensuring the accuracy and continuity of the operational boundary reference data. Furthermore, the generated operational boundary reference data includes both a globally stable boundary framework and supplements the dynamic boundary changes of local operational areas, providing precise range guidance for autonomous driving operations and improving the safety and accuracy of autonomous driving operations.

[0040] In some technical solutions of this application, determining the target grid map covering the entire target operation scene based on multiple initial grid maps includes: calculating the terrain elevation and slope of each initial grid map to obtain multiple initial grid maps after calculation; and performing edge matching and stitching of the multiple initial grid maps after calculation according to the spatial coordinate association relationship of the target operation scene to obtain the target grid map covering the entire target operation scene.

[0041] In this technical solution, terrain elevation refers to the vertical height of a point on the terrain relative to a preset reference surface (such as sea level). It is a core parameter reflecting terrain undulation and can directly affect the driving safety and operational efficiency of autonomous vehicles.

[0042] Slope refers to the degree of inclination of the terrain surface, usually expressed as a percentage or angle. It is a key parameter reflecting the steepness of the terrain, and the slope directly determines the passability and operational feasibility of autonomous vehicles.

[0043] Spatial coordinate correlation refers to the spatial positional correspondence of various local areas in the target operation scene, that is, the relative position, overlap range, and other relationships of the partitions corresponding to different initial raster maps in the global coordinate system. It is the core basis for realizing raster map stitching.

[0044] Edge matching refers to the process of comparing and aligning the features of the edge parts of the initial raster maps of adjacent partitions. By identifying the consistency of features such as terrain elevation and slope in the edge areas, it ensures that adjacent raster maps can be accurately connected, avoiding splicing gaps or overlapping conflicts.

[0045] Stitching refers to the process of integrating multiple initial raster maps that have undergone edge matching into a complete map. By merging the overlapping area data of adjacent raster maps, a continuous and complete target raster map covering the entire target operation scene is formed.

[0046] This application, by clearly defining the calculation content of the initial raster map and the stitching method of the target raster map, enables the calculation of terrain elevation and slope for each initial raster map, supplementing the core terrain parameters of the raster map. This allows the raster map to more accurately reflect terrain features, providing high-quality foundational data for the subsequent construction of global terrain data. Simultaneously, edge matching and stitching through spatial coordinate relationships ensure precise connection between multiple initial raster maps, avoiding gaps, overlaps, or abrupt terrain changes in the stitched map, thus guaranteeing the integrity and consistency of the target raster map. Furthermore, the generated target raster map covers the entire target operational scenario, providing a complete terrain framework for the construction of global terrain data, while also improving the practicality and reliability of the map data.

[0047] In the open-pit mine scenario, the global basic data is processed to generate eight initial raster maps covering the entire mine (each initial raster map corresponds to a section of the mine, with a 10% overlap between sections). First, terrain elevation and slope are calculated for each initial raster map: based on the original point cloud data in the initial raster map, an interpolation algorithm is used to calculate the terrain elevation of each raster unit (accuracy ±0.1m), and the slope of each raster unit is calculated (accuracy ±1°) based on the elevation difference between adjacent raster units, resulting in eight initial raster maps containing terrain elevation and slope parameters. Subsequently, the spatial coordinate relationships of each initial raster map are determined according to the mine's global coordinate system, and overlapping areas between adjacent initial raster maps (each overlapping area contains 50 raster units) are identified. Edge matching is performed on the terrain elevation and slope features of the overlapping areas, and the spatial positions of adjacent initial raster maps are adjusted to ensure that the terrain parameter deviation in the overlapping areas is less than 0.2m (elevation deviation) and 2° (slope deviation). Finally, the eight initial raster maps after matching are stitched together, and the terrain data of the overlapping areas are merged (the elevation and slope values ​​of the overlapping areas are calculated using a weighted average method). The connection traces of the partition boundaries are eliminated, and a target raster map covering the entire mine is generated. This map can completely and continuously reflect the overall terrain features of the mine, providing a reliable foundation for the construction of subsequent overall terrain data.

[0048] In some technical solutions of this application, after obtaining the point cloud map based on the global basic data, the method further includes: thinning and slicing the point cloud map to obtain the preprocessed point cloud map.

[0049] In this technical solution, thinning refers to the process of removing redundant point data that have little impact on terrain features from the point cloud map by using a preset thinning algorithm (such as uniform thinning method, feature point retention thinning method, etc.), thereby reducing the total amount of point cloud data without losing key terrain information.

[0050] Slicing refers to dividing a point cloud map into multiple independent point cloud tiles of the same size according to a preset spatial range (such as square tiles). Each point cloud tile contains point cloud data of the corresponding spatial range, which is convenient for storage, transmission and parallel processing.

[0051] Preprocessed point cloud maps refer to point cloud maps obtained after thinning and slicing. They consist of multiple point cloud tiles, each containing thinned key terrain point data. This process preserves the core features of the terrain while significantly reducing the amount of data, making it more efficient for subsequent construction of global terrain data.

[0052] This application clarifies the preprocessing method for point cloud maps, achieving the elimination of redundant data through thinning, reducing the total amount of point cloud data, lowering data storage pressure and subsequent computational load, and improving data processing efficiency. Simultaneously, tiling divides the point cloud map into multiple independent tiles, facilitating parallel processing and precise data retrieval. When constructing full-domain terrain data, the corresponding area's point cloud tiles can be called as needed, eliminating the need to load the complete point cloud map, further improving processing efficiency. Furthermore, the preprocessed point cloud map achieves data lightweighting while preserving key terrain features, ensuring the accuracy of full-domain terrain data construction while improving data processing and transmission efficiency.

[0053] In the open-pit mine scenario, the point cloud map generated from the global baseline data contains 100 million 3D points, amounting to 50GB. Directly using this data for subsequent processing would result in low computational efficiency and excessive storage pressure. Therefore, the point cloud map is preprocessed: First, a feature point thinning method is used to thin the data, setting a thinning threshold (one feature point is retained when the distance between adjacent points is less than 0.2m). Redundant point data is removed, reducing the point cloud data size to 1GB, only 2% of the original data size, while fully preserving key features of the mine terrain such as peaks, valleys, and pits. Then, 20m×20m square tiles are used to slice the thinned point cloud map, dividing it into 1000 independent point cloud tiles, each with a data size of approximately 1MB. When constructing global terrain data, the preprocessed point cloud map can accurately call the point cloud tiles of the corresponding area according to the partition range of the target raster map, and merge them with the local raster terrain data and the target raster map. This avoids the inefficiency caused by loading the full point cloud data, ensures the accuracy of terrain data construction, and significantly reduces the pressure of data storage and transmission.

[0054] In some technical solutions of this application, generating autonomous driving map data for the target operation scenario based on global terrain data and work boundary reference data includes: aligning spatial coordinates based on global terrain data and work boundary reference data to obtain coordinate-matched global terrain data and work boundary reference data; correcting terrain elevation deviations and handling areas with abrupt changes in terrain slope based on the coordinate-matched global terrain data to obtain accuracy-optimized global terrain data; performing fusion encoding based on the accuracy-optimized global terrain data and work boundary reference data to obtain a map data prototype containing 3D terrain information and operation boundary markers; and supplementing road network topology and operation point labeling information according to map data specifications recognizable by the autonomous driving system based on the map data prototype to obtain autonomous driving map data for the target operation scenario.

[0055] In this technical solution, spatial coordinate alignment refers to aligning the coordinate systems of the global terrain data and the working boundary reference data to the same preset coordinate system (such as the local coordinate system or global coordinate system of the target operation scene), and adjusting the spatial positions of the two types of data to ensure that the working boundary reference data can accurately match the corresponding spatial position of the global terrain data. The core is to eliminate the coordinate deviation between the two types of data.

[0056] The coordinate-matched global terrain data and working boundary reference data refer to the global terrain data and working boundary reference data that have been processed by spatial coordinate alignment and whose coordinate system and spatial location are accurately matched. The two types of data can correspond to each other within the same spatial framework, laying the foundation for subsequent fusion coding.

[0057] Topographic elevation deviation refers to the difference between the topographic elevation value in the global topographic data and the actual topographic elevation value. The causes of deviation include sensor acquisition error, data processing error, etc., and it needs to be corrected to ensure the accuracy of topographic elevation.

[0058] Abrupt changes in terrain slope refer to areas in the overall terrain data where the slope value changes abruptly and significantly. These areas are usually abnormal regions that occur during data processing and do not conform to the natural variation patterns of actual terrain, thus requiring smoothing processing.

[0059] Accuracy-optimized full-domain terrain data refers to full-domain terrain data that has undergone terrain elevation deviation correction and slope change abrupt area processing, resulting in more accurate terrain parameters and terrain changes that better reflect the actual situation, thus possessing higher accuracy and reliability.

[0060] Fusion coding refers to the process of integrating the optimized global terrain data and the working boundary reference data into a whole, and formatting the data according to the preset coding rules to transform the three-dimensional terrain information and the working boundary markers into structured data that can be recognized and stored by computers.

[0061] Map data prototypes refer to the initial map data generated after fusion encoding, containing 3D terrain information and work boundary markers. This is the basic version of autonomous driving map data, but it lacks supplementary information such as road network topology and work point annotations. Map data specifications recognizable by autonomous driving systems refer to the technical specifications such as map data format, data structure, and parameter indicators preset by the autonomous driving system. Different types of autonomous driving systems may have different specification requirements; map data must conform to these specifications to be recognized and used by the autonomous driving system. Road network topology refers to the connection relationships and traffic rules of roads in the target work scenario, including information such as the start and end points of roads, intersections, traffic directions, and number of lanes. This is the core information guiding the autonomous vehicle's driving route. Work point annotations refer to the identification information of key work locations in the target work scenario, including the name, type (e.g., loading point, unloading point, maintenance point), coordinates, and work area of ​​the work point, facilitating accurate identification of work locations by autonomous vehicles.

[0062] This application achieves the following technical effects by clearly defining the steps for generating autonomous driving map data: First, spatial coordinate alignment ensures the spatial consistency between the overall terrain data and the working boundary reference data, laying the foundation for the fusion of the two types of data and avoiding map data errors caused by coordinate deviations. Second, terrain accuracy optimization corrects elevation deviations and abrupt slope changes, improving the accuracy and rationality of the overall terrain data and ensuring that the map data truly reflects the actual terrain. Third, fusion coding integrates terrain data and boundary data into a structured map data prototype. After supplementing road network topology and work point annotations, it generates final map data that meets the requirements of autonomous driving systems, realizing the transformation from basic data to practical map data and ensuring that the generated map data can directly support autonomous driving operations.

[0063] A second aspect of this application provides an apparatus for determining autonomous driving map data, comprising: a first acquisition module, a second acquisition module, a third acquisition module, a first determination module, a second determination module, a third determination module, and a generation module. The first acquisition module is used to acquire full-domain basic data of a target work scenario during downtime and partial work data of the target work scenario during operating hours. The second acquisition module is used to acquire local raster terrain data corresponding to the local work data and first work boundary feature data within the local area based on the partial work data. The third acquisition module is used to acquire a point cloud map, initial raster maps of multiple local areas within the full domain, and second work boundary feature data within the full domain based on the full-domain basic data. The first determination module is used to determine a target raster map covering the entire target work scenario based on the multiple initial raster maps. The second determination module is used to determine full-domain terrain data based on the partial raster terrain data, the preprocessed point cloud map, and the target raster map. The third determination module is used to determine work boundary reference data based on the first work boundary feature data and the second work boundary feature data. The generation module is used to generate autonomous driving map data for the target work scenario based on the full-domain terrain data and the work boundary reference data.

[0064] The autonomous driving map data determination device of this application acquires full-domain basic data and local operation data in time periods through a first acquisition module, which can supplement the data source without the need for additional equipment, avoiding frequent downtime caused by reliance on dedicated data collection vehicles. The second acquisition module uses the computing power module of the unmanned operation vehicle to locally solve local raster terrain data and uploads it with a preset range. The third acquisition module performs thinning and slicing preprocessing on the point cloud map, which greatly reduces the network transmission pressure and is suitable for the harsh network environment of remote scenarios such as open-pit mines. The first determination module generates a full-domain target raster map through partition calculation and stitching. The second determination module integrates multi-source terrain data to construct accurate full-domain terrain data. The third determination module fuses and verifies the two types of operation boundary feature data to ensure boundary accuracy. Finally, the generation module generates map data that meets the requirements of the autonomous driving system. The multi-module collaboration realizes efficient acquisition, transmission and generation of map data, which not only improves the operational efficiency of target operation scenarios such as open-pit mines, but also enhances the accuracy and reliability of map data, effectively solving the pain point of high dependence on data collection vehicles and network environment in existing technologies.

[0065] A third aspect of this application provides an apparatus for determining autonomous driving map data, comprising: a processor and a memory, wherein the memory stores a program or instructions, and the processor, when executing the program or instructions in the memory, implements the steps of the autonomous driving map data determination method as described in any of the above-described technical solutions. Therefore, the apparatus for determining autonomous driving map data possesses all the beneficial effects of the autonomous driving map data determination method as described in any of the above-described technical solutions.

[0066] A fourth aspect of this application provides a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps of the method for determining autonomous driving map data as described in any of the above-described technical solutions. Therefore, the readable storage medium possesses all the beneficial effects of the method for determining autonomous driving map data as described in any of the above-described technical solutions.

[0067] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0068] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0069] Figure 1 This is a flowchart illustrating a method for determining autonomous driving map data according to an embodiment of this application;

[0070] Figure 2 This is a schematic diagram of an overall system for crowdsourced data collection, processing, and map generation for autonomous driving maps in open-pit mines, according to an embodiment of this application.

[0071] Figure 3 This is one of the schematic block diagrams of an autonomous driving map data determination device according to an embodiment of this application;

[0072] Figure 4 This is a second schematic block diagram of an autonomous driving map data determination device according to an embodiment of this application. Detailed Implementation

[0073] To better understand the above-mentioned objectives, features, and advantages of this application, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0074] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below.

[0075] The following reference Figures 1 to 4 This application describes a method, apparatus, and readable storage medium for determining autonomous driving map data according to some embodiments.

[0076] like Figure 1 As shown, an embodiment of this application provides a method for determining autonomous driving map data, the steps of which include:

[0077] Step 102: Obtain the overall basic data of the target work scenario during the downtime period, and the partial work data of the target work scenario during the operation period;

[0078] Step 104: Based on the local operation data, obtain the local raster terrain data corresponding to the local operation data, and the first operation boundary feature data within the local area;

[0079] Step 106: Based on the global basic data, obtain the point cloud map, the initial raster map of multiple local areas within the global area, and the second operation boundary feature data within the global area;

[0080] Step 108: Based on multiple initial grid maps, determine the target grid map that covers the entire target operation scene.

[0081] Step 110: Determine the global terrain data based on the local raster terrain data, the preprocessed point cloud map, and the target raster map;

[0082] Step 112: Determine the work boundary reference data based on the first work boundary feature data and the second work boundary feature data;

[0083] Step 114: Generate autonomous driving map data for the target operation scenario based on global terrain data and work boundary reference data.

[0084] The target operation scenario provided in this application refers to a specific application scenario that requires the generation of autonomous driving map data. Its core characteristics include harsh environment, frequent terrain changes, and high requirements for the timeliness and accuracy of map updates. A typical scenario is an open-pit mine, which has no fixed roads and signs. The operation mode revolves around "loading-transporting-discharging" and the number of on-site personnel must be strictly limited.

[0085] The shutdown period refers to a specific time period in the target work scenario during which production and operation are suspended. It does not mean that all activities are completely stopped, but rather that core production operations are suspended to carry out non-production-related necessary work, including shift change periods, inspection periods, or maintenance periods. During this period, network communication conditions are relatively stable, which is suitable for large-scale data collection.

[0086] Full-domain basic data refers to the basic data collected during the downtime of the target operation scenario, covering the entire scope of the scenario. It is the core foundation for building autonomous driving maps. Data types include raw point cloud files, LiDAR calibration parameters, etc., which can comprehensively reflect the overall terrain, spatial layout and other basic information of the scenario.

[0087] The operating period refers to the time during which production and operation activities are carried out normally in the target operation scenario. For open-pit mine scenarios, this refers to the time during which operations are carried out around the "loading-transporting-discharging" process. During this period, it is necessary to collect data from key areas without affecting production efficiency.

[0088] Local operation data refers to targeted data collected during the operation period of the target operation scenario, focusing on the current actual operation area. Data types include raw point cloud files after being filtered by preset rules, LiDAR calibration parameters, etc., which are used to supplement and update the real-time terrain and boundary change information of the operation area.

[0089] Local raster terrain data refers to local terrain data presented in raster form after processing local operational data. The raster format can quickly reflect key features of local terrain such as elevation and slope, facilitating efficient transmission and fusion processing.

[0090] The first operational boundary feature data refers to the feature information related to the boundary of the local operational area extracted from the local operational data. Specifically, it includes features such as the terrain edge and operational point boundary of the local operational area, which can accurately reflect the real-time boundary changes of the current operational area.

[0091] Point cloud maps are maps generated based on the original point cloud files in the global basic data. They consist of a large number of discrete three-dimensional point coordinates and can realistically restore the three-dimensional spatial form of the target operation scene. They are an important basis for subsequent terrain data construction and boundary extraction.

[0092] An initial raster map refers to a rasterized map that is generated based on global basic data and covers multiple local areas within the global domain. Each initial raster map corresponds to a partition in the global domain and contains the basic terrain information of that partition, but has not yet undergone global integration and accuracy optimization.

[0093] The second operational boundary feature data refers to the boundary feature information extracted from the overall basic data, covering the entire range of the target operational scenario. Specifically, it includes features such as terrain boundaries and fixed facility boundaries within the entire range, and can provide the overall boundary framework of the scenario.

[0094] A target raster map is a complete raster map that covers the entire target operation scene, formed by calculating terrain parameters, matching edges, and stitching together multiple initial raster maps. It has the characteristics of complete terrain information and consistent regional connection, and is the core carrier for constructing full-domain terrain data.

[0095] Preprocessed point cloud map refers to the map obtained after thinning and slicing the point cloud map. Thinning can remove redundant point data and reduce the amount of data, while slicing can divide the point cloud map into multiple tiles according to preset rules, which facilitates storage, transmission and subsequent fusion processing.

[0096] Global terrain data refers to complete terrain data that covers the entire range of the target operation scene after integrating local raster terrain data, preprocessed point cloud maps, and target raster maps. It includes the overall terrain framework of the entire region and supplements the real-time terrain details of the operation area, which can comprehensively and accurately reflect the terrain characteristics of the scene.

[0097] Work boundary reference data refers to boundary reference information obtained by fusing and verifying the first work boundary feature data and the second work boundary feature data. It is used to guide autonomous driving operations and has the characteristics of being continuous, complete, accurate and reliable, and can clearly define the work range and safety boundaries.

[0098] Autonomous driving map data refers to the final map data generated based on global terrain data and work boundary reference data, which meets the requirements for use by autonomous driving systems. It includes key content such as 3D terrain information, work boundary markings, road network topology, and work point markings, and can directly support the safe and efficient operation of autonomous driving equipment in the target work scenario.

[0099] In summary, this application addresses the issues of reliance on dedicated data collection vehicles, frequent downtime, high network transmission pressure, and insufficient map accuracy and timeliness in the target operational scenario for autonomous driving map updates. It achieves efficient map data generation without additional equipment costs, without impacting production operations, and adaptable to harsh network environments. Specifically, the process first achieves high efficiency and continuity through "dual-time period data collection," avoiding efficiency losses caused by downtime. This application collects data in different time periods. During downtime, it collects basic data across the entire area. At this time, the target operational scenario suspends core production operations, ensuring no impact on production progress due to data collection. Furthermore, a stable high-speed network can be used to upload large-scale data during this period. During operational periods, it simultaneously collects local operational data, focusing only on the current operational area, without needing to cover the entire area. The collection process runs parallel to the production and operation process, without interfering with core operations such as "loading-transportation-arrangement." This dual-time period collection mode, combining basic data across the entire area with local supplementation, ensures both comprehensive map data coverage and real-time updates of operational area data. It avoids downtime losses at the collection stage and improves the collaborative efficiency between production operations and map updates.

[0100] Secondly, the accuracy and reliability of map data are improved through "multi-source data fusion". This application collects two different types of data: global basic data and local operation data. From the global basic data, point cloud maps, initial raster maps, and second operation boundary feature data are obtained to construct the overall terrain and boundary framework of the scene. From the local operation data, local raster terrain data and first operation boundary feature data are obtained to supplement the real-time terrain details and boundary changes of the operation area. Then, through multiple rounds of data fusion, multiple initial raster maps are stitched together to form a target raster map. The local raster terrain data, the preprocessed point cloud map, and the target raster map are fused to generate global terrain data. The two types of operation boundary feature data are fused to generate working boundary reference data. The multi-source data mutually verify and supplement each other, effectively eliminating errors and abnormal information from a single data source, improving the accuracy and reliability of the map data, and ensuring that the map data can truly reflect the terrain and boundary conditions of the target operation scene.

[0101] Finally, layered data processing and precise transmission reduce network transmission pressure and adapt to harsh network environments. Targeted operational scenarios (such as open-pit mines) are typically located in remote areas, relying on self-built base stations for communication, resulting in poor network transmission conditions. This application implements targeted optimizations in data processing and transmission. Specifically, local operational data is first processed into local raster terrain data using the computing module of the unmanned vehicle, and then only data relevant to the current operation is transmitted using a preset range upload method, eliminating irrelevant and redundant data, significantly reducing the amount of data transmitted. Point cloud maps are thinned and pre-processed by tiling to reduce data storage and transmission pressure. The initial raster map is first partitioned and calculated before stitching, avoiding the inefficiency caused by directly processing the entire raster data. Through this series of layered processing and precise transmission strategies, the dependence on network bandwidth is reduced, enabling efficient data transmission even in environments with poor network conditions and ensuring the timeliness of map data updates.

[0102] In some embodiments of this application, the global basic data includes the original point cloud file and the lidar calibration parameters, and the local operation data includes the original point cloud file and the lidar calibration parameters after being filtered by preset rules; the downtime is the shift change time, inspection time or maintenance time of the target operation scenario. The global basic data is collected at a preset speed and route and then uploaded through a high-speed network, and the local operation data is collected synchronously with the operation process of the target operation scenario.

[0103] In this embodiment, the raw point cloud file refers to the original data file composed of a large number of discrete three-dimensional point coordinates acquired by acquisition devices such as LiDAR. Each three-dimensional point contains spatial location information (X, Y, Z coordinates) and can directly reflect the spatial form of the acquisition area, such as terrain and features. It is the core raw data for generating point cloud maps and raster maps. LiDAR calibration parameters are key parameters used to calibrate the acquisition accuracy of LiDAR, including LiDAR installation position offset parameters, angle deviation parameters, and ranging accuracy correction parameters. These parameters can correct system errors during the LiDAR acquisition process, ensuring that the acquired raw point cloud file accurately corresponds to the actual spatial location. Preset rule filtering refers to the pre-set rules for filtering local operation data based on the operational needs and data processing requirements of the target operation scenario. These filtering rules include data validity rules (removing invalid data acquired abnormally by sensors), correlation rules (retaining data directly related to the current operation area), and accuracy threshold rules (retaining data that meets preset accuracy requirements), aiming to improve the quality and relevance of local operation data. Shift change periods refer to the time when workers in the target work scenario hand over responsibilities. During this period, production operations are temporarily suspended, there is no large-scale movement of equipment, and the environment is relatively stable, making it suitable for large-scale data collection. Inspection periods refer to the time periods for regular inspections and maintenance of production equipment and infrastructure in the target work scenario. During this period, production operations are suspended, facilitating the movement of data collection equipment within the scenario without affecting inspection work. Maintenance periods refer to the time periods for repairing faulty equipment and damaged facilities in the target work scenario. During this period, production operations in the relevant areas are suspended, allowing for the use of maintenance gaps to conduct comprehensive basic data collection, achieving coordination between data collection and equipment maintenance. Preset speed and route refer to the pre-set movement speed and route of the data collection equipment for comprehensive basic data collection. The preset speed must balance collection accuracy and efficiency, avoiding incomplete or inaccurate data collection due to excessive speed. The preset route must cover the entire area of ​​the target work scenario to ensure comprehensive data coverage. High-speed networks refer to networks with high transmission rates and good stability that are available during downtime in the target work scenario. These are typically dedicated communication networks built by the scenario itself, capable of meeting the needs for rapid and complete transmission of basic data across the entire domain (such as large-capacity raw point cloud files). Operational processes refer to the standardized procedures for carrying out core production activities in the target work scenario. For open-pit mines, this is the "loading-transporting-discharging" process (the complete process of loading, transporting, and discharging ore). The collection of localized operational data needs to be synchronized with this process, collecting data from the corresponding areas simultaneously when the equipment performs loading, transporting, and discharging operations.

[0104] This application further refines the technical solution by clearly defining the specific types of overall basic data and localized operational data, the specific scope of downtime periods, and the methods for collecting and uploading both types of data. It clarifies data types and filtering rules, ensuring the effectiveness and relevance of the collected data. Raw point cloud files and LiDAR calibration parameters are core data for generating high-precision maps; preset filtering rules can eliminate invalid and redundant information in localized operational data, improving data processing efficiency. Secondly, it clarifies the specific scenarios of downtime periods, providing an operable time window for overall basic data collection and avoiding conflicts between collection time and production operations. Simultaneously, by standardizing collection and uploading methods, preset speeds and routes ensure the overall coverage and accuracy of overall basic data collection, high-speed network uploading ensures the transmission efficiency of large-capacity overall basic data, and synchronous collection with operational processes ensures the real-time nature of localized operational data.

[0105] In one embodiment, taking an open-pit mine scenario as an example, the downtime periods include twice-daily shift change periods (30 minutes each time), once-weekly inspection periods (2 hours each time), and maintenance periods for equipment failures (lasting 1-4 hours depending on the failure). During shift change periods, the manned map data acquisition platform unit is activated, traveling at a preset speed (5km / h) and along a preset route covering the entire mine, collecting raw point cloud files and LiDAR calibration parameters for the entire mine area as global basic data. The collected global basic data is then uploaded to the server-side computing unit via the mine's self-built high-speed wireless network (transmission rate ≥100Mbps). During the normal operation of the mine and the "loading-transporting-discharging" process, the unmanned mining truck simultaneously collects the original point cloud files and lidar calibration parameters of the current working area while performing ore loading, transportation and unloading operations. The collected data is filtered according to preset rules (removing invalid data with a ranging error greater than 0.5cm and retaining relevant data within 100 meters of the working equipment) to obtain local operation data, ensuring that the local operation data can accurately reflect the terrain and boundary changes of the current working area.

[0106] In some embodiments of this application, obtaining local grid terrain data corresponding to local operation data based on local operation data includes: calculating the local operation data using the computing power module of an unmanned operation vehicle in the target operation scenario to determine the local grid terrain data; transmitting the local grid terrain data to the computing terminal using a preset range upload method; wherein, the preset range upload method is to upload local grid terrain data related to the current operation of the target operation scenario.

[0107] In this embodiment, the unmanned operation vehicle refers to a vehicle that has been modified for unmanned operation and has the ability to drive and operate autonomously. In the open-pit mine scenario, it is an unmanned mining truck. Its core components include a domain controller, a gateway controller, a vehicle controller, sensors that can support map acquisition (such as LiDAR), a gateway, etc., which can complete production operations and data acquisition tasks without human driving.

[0108] The computing module refers to a hardware module integrated on an unmanned operating vehicle that has data processing and computing capabilities. It can quickly process and analyze the local operating data collected by the vehicle, and can generate local grid terrain data without transmitting the raw data to a remote computing device, thereby improving data processing efficiency.

[0109] Solving refers to the process of analyzing, transforming, and calculating local operation data through a preset algorithm model and calculation logic, converting discrete raw point cloud files and lidar calibration parameters into structured, rasterized local raster terrain data. The solving process includes key steps such as coordinate transformation, elevation calculation, and raster division.

[0110] The preset range upload method refers to the pre-defined data upload range rules, which only upload local raster terrain data that is directly related to the current operation activities of the target work scenario, and remove redundant terrain data that is not related to the current operation. The core purpose is to reduce the amount of data transmitted and reduce network transmission pressure.

[0111] The computing end refers to the core computing device that receives and processes uploaded data. In this application, it is the server computing unit, which has powerful data analysis, fusion and storage capabilities. It can integrate and process the global basic data and the uploaded local raster terrain data to generate key data such as target raster maps and global terrain data.

[0112] This application further improves the technical solution by clarifying the calculation subject, calculation method, and upload method of local raster terrain data. It enables local calculation using the computing module built into the unmanned operating vehicle, generating local raster terrain data without transmitting the original local operation data to a remote computing terminal. This reduces bandwidth consumption and transmission latency of the original data transmission, improving the real-time performance of data processing. Furthermore, the use of a preset range upload method transmits only data relevant to the current operation, significantly reducing the amount of uploaded data, effectively reducing network transmission pressure, adapting to the harsh network environment of the target operation scenario, and avoiding data loss or update delays due to network congestion or poor transmission quality. Simultaneously, the calculation and upload processes are performed synchronously with the operation flow during operating hours, without affecting production efficiency.

[0113] In one embodiment, in an open-pit mine scenario, the unmanned mining truck, acting as an unmanned operating vehicle, is equipped with a computing module (CPU clock speed ≥ 2.5GHz, GPU memory ≥ 8GB) capable of rapid data processing. When the unmanned mining truck performs ore transportation operations during operating hours, it simultaneously collects local operational data (filtered raw point cloud files and LiDAR calibration parameters) of the current transportation route and operating area. The computing module calls a preset processing algorithm (a rasterization algorithm based on triangular mesh interpolation) to process the local operational data into local raster terrain data with a resolution of 0.1m × 0.1m. Following preset upload rules, only the local raster terrain data of the unmanned mining truck's current operating area (a radius of 150 meters centered on the vehicle) is uploaded. This range covers the key areas required for the current operation, eliminating redundant terrain data from other unoperated areas of the mine. The amount of uploaded data is reduced by more than 80% compared to a full upload. This data is rapidly transmitted to the server-side computing unit via the mine's wireless communication network, effectively reducing network transmission pressure and ensuring that the data reaches the computing end in a timely manner for subsequent fusion processing.

[0114] In some embodiments of this application, the first operation boundary feature data is the terrain edge and operation point boundary feature information of the local operation area, and the second operation boundary feature data is the terrain boundary and fixed facility boundary feature information of the entire area; determining the work boundary reference data based on the first operation boundary feature data and the second operation boundary feature data includes: performing fusion verification on the first operation boundary feature data and the second operation boundary feature data, removing abnormal feature information, and obtaining the work boundary reference data.

[0115] In this embodiment, terrain edge refers to the boundary line formed by terrain undulations, such as the boundary line between a hillside and flat land, or the boundary line between a depression and the ground. It reflects the spatial morphological changes of the terrain and is an important component of the operational boundary. Operational point boundary refers to the boundary range of a specific operational location within the target operational scenario, such as the boundary of an ore loading point, the boundary of an unloading point, or the boundary of a transportation route. It directly relates to the scope definition and safety control of autonomous driving operations. Fixed facility boundary refers to the boundary range of fixed facilities and equipment within the target operational scenario, such as the boundary of a mine wall, the boundary of an equipment parking area, the boundary of an office area, or the boundary of a road. These facility boundaries are relatively stable and are an important basis for constructing the overall boundary framework of the scenario. Fusion verification refers to the process of integrating, comparing, and verifying the first operational boundary feature data and the second operational boundary feature data. Through preset verification rules (such as spatial location consistency verification, boundary continuity verification, and data credibility verification), the consistency and accuracy of the two types of data are judged. Abnormal feature information refers to feature information in the two types of operational boundary feature data that does not conform to the actual scenario, contains errors or contradictions, such as false terrain edges caused by sensor failure, boundary overlap or breakage caused by data acquisition deviation, and invalid fixed facility boundaries caused by facility dismantling and failure to update.

[0116] This application clarifies the specific content of two types of operational boundary feature data and the generation method of operational boundary reference data, thereby defining the specific types of boundary feature data. This ensures that the boundary information extracted from the two data sources is targeted and complementary. Terrain edges and operational point boundaries reflect dynamically changing operational boundaries, while fixed facility boundaries provide a stable overall boundary framework. The combination of these two types of data achieves comprehensive coverage of boundary information. Simultaneously, through a fusion verification process, abnormal feature information can be effectively eliminated, addressing potential errors and unreliability issues associated with single data sources and ensuring the accuracy and continuity of the operational boundary reference data. Furthermore, the generated operational boundary reference data includes both a globally stable boundary framework and supplements the dynamic boundary changes of local operational areas, providing precise range guidance for autonomous driving operations and improving the safety and accuracy of autonomous driving operations.

[0117] In one embodiment, in an open-pit mine scenario, first operational boundary feature data is extracted from local operational data collected by unmanned mining trucks. This includes the operational point boundary of the current ore loading point (a circular boundary with a radius of 20 meters centered on the loading equipment) and the terrain edges surrounding the loading point (the boundary between the hillside and the flat land where the loading point is located). Second operational boundary feature data is extracted from the overall basic data collected by manned map acquisition units. This includes the terrain boundaries of the entire mine (the boundary between the mine's mountain and the surrounding flat land) and the boundaries of fixed facilities (mine perimeter wall boundaries, main road boundaries, and equipment parking area boundaries). After transmitting the two types of boundary feature data to the server-side computing unit, a fusion verification is performed: First, a spatial coordinate consistency verification ensures that the coordinate systems of the two types of data are consistent, and that the spatial positions of the operational point boundaries match the mine's terrain boundaries and fixed facility boundaries. Second, a boundary continuity verification checks whether the connection between the loading point's terrain edge and the overall mine terrain boundary is continuous, avoiding breaks or overlaps. Finally, through data credibility verification, false terrain edges caused by lidar obstruction (such as false boundaries formed by tree obstruction) and invalid fixed facility boundaries of dismantled equipment are eliminated. After fusion verification and elimination of abnormal feature information, continuous and complete working boundary reference data is generated, clarifying the safety boundary of the current working area, guiding unmanned mining trucks to carry out loading and transportation operations within the designated boundaries, and avoiding safety accidents caused by exceeding the working range.

[0118] In some embodiments of this application, determining a target raster map covering the entire target work scene based on multiple initial raster maps includes: calculating the terrain elevation and slope of each initial raster map to obtain multiple calculated initial raster maps; and performing edge matching and stitching of the multiple calculated initial raster maps according to the spatial coordinate association relationship of the target work scene to obtain a target raster map covering the entire target work scene.

[0119] In this embodiment, terrain elevation refers to the vertical height of a point on the terrain relative to a preset reference surface (such as sea level). It is a core parameter reflecting terrain undulation and can directly affect the driving safety and operational efficiency of autonomous vehicles.

[0120] Slope refers to the degree of inclination of the terrain surface, usually expressed as a percentage or angle. It is a key parameter reflecting the steepness of the terrain, and the slope directly determines the passability and operational feasibility of autonomous vehicles.

[0121] Spatial coordinate correlation refers to the spatial positional correspondence of various local areas in the target operation scene, that is, the relative position, overlap range, and other relationships of the partitions corresponding to different initial raster maps in the global coordinate system. It is the core basis for realizing raster map stitching.

[0122] Edge matching refers to the process of comparing and aligning the features of the edge parts of the initial raster maps of adjacent partitions. By identifying the consistency of features such as terrain elevation and slope in the edge areas, it ensures that adjacent raster maps can be accurately connected, avoiding splicing gaps or overlapping conflicts.

[0123] Stitching refers to the process of integrating multiple initial raster maps that have undergone edge matching into a complete map. By merging the overlapping area data of adjacent raster maps, a continuous and complete target raster map covering the entire target operation scene is formed.

[0124] This application, by clearly defining the calculation content of the initial raster map and the stitching method of the target raster map, enables the calculation of terrain elevation and slope for each initial raster map, supplementing the core terrain parameters of the raster map. This allows the raster map to more accurately reflect terrain features, providing high-quality foundational data for the subsequent construction of global terrain data. Simultaneously, edge matching and stitching through spatial coordinate relationships ensure precise connection between multiple initial raster maps, avoiding gaps, overlaps, or abrupt terrain changes in the stitched map, thus guaranteeing the integrity and consistency of the target raster map. Furthermore, the generated target raster map covers the entire target operational scenario, providing a complete terrain framework for the construction of global terrain data, while also improving the practicality and reliability of the map data.

[0125] In the open-pit mine scenario, the global basic data is processed to generate eight initial raster maps covering the entire mine (each initial raster map corresponds to a section of the mine, with a 10% overlap between sections). First, terrain elevation and slope are calculated for each initial raster map: based on the original point cloud data in the initial raster map, an interpolation algorithm is used to calculate the terrain elevation of each raster unit (accuracy ±0.1m), and the slope of each raster unit is calculated (accuracy ±1°) based on the elevation difference between adjacent raster units, resulting in eight initial raster maps containing terrain elevation and slope parameters. Subsequently, the spatial coordinate relationships of each initial raster map are determined according to the mine's global coordinate system, and overlapping areas between adjacent initial raster maps (each overlapping area contains 50 raster units) are identified. Edge matching is performed on the terrain elevation and slope features of the overlapping areas, and the spatial positions of adjacent initial raster maps are adjusted to ensure that the terrain parameter deviation in the overlapping areas is less than 0.2m (elevation deviation) and 2° (slope deviation). Finally, the eight initial raster maps after matching are stitched together, and the terrain data of the overlapping areas are merged (the elevation and slope values ​​of the overlapping areas are calculated using a weighted average method). The connection traces of the partition boundaries are eliminated, and a target raster map covering the entire mine is generated. This map can completely and continuously reflect the overall terrain features of the mine, providing a reliable foundation for the construction of subsequent overall terrain data.

[0126] In some embodiments of this application, after obtaining the point cloud map based on the global basic data, the method further includes: performing thinning and slicing preprocessing on the point cloud map to obtain a preprocessed point cloud map.

[0127] In this embodiment, thinning refers to the process of removing redundant point data that have little impact on terrain features from the point cloud map by using a preset thinning algorithm (such as uniform thinning method, feature point retention thinning method, etc.), thereby reducing the total amount of point cloud data without losing key terrain information.

[0128] Slicing refers to dividing a point cloud map into multiple independent point cloud tiles of the same size according to a preset spatial range (such as square tiles). Each point cloud tile contains point cloud data of the corresponding spatial range, which is convenient for storage, transmission and parallel processing.

[0129] Preprocessed point cloud maps refer to point cloud maps obtained after thinning and slicing. They consist of multiple point cloud tiles, each containing thinned key terrain point data. This process preserves the core features of the terrain while significantly reducing the amount of data, making it more efficient for subsequent construction of global terrain data.

[0130] This application clarifies the preprocessing method for point cloud maps, achieving the elimination of redundant data through thinning, reducing the total amount of point cloud data, lowering data storage pressure and subsequent computational load, and improving data processing efficiency. Simultaneously, tiling divides the point cloud map into multiple independent tiles, facilitating parallel processing and precise data retrieval. When constructing full-domain terrain data, the corresponding area's point cloud tiles can be called as needed, eliminating the need to load the complete point cloud map, further improving processing efficiency. Furthermore, the preprocessed point cloud map achieves data lightweighting while preserving key terrain features, ensuring the accuracy of full-domain terrain data construction while improving data processing and transmission efficiency.

[0131] In the open-pit mine scenario, the point cloud map generated from the global baseline data contains 100 million 3D points, amounting to 50GB. Directly using this data for subsequent processing would result in low computational efficiency and excessive storage pressure. Therefore, the point cloud map is preprocessed: First, a feature point thinning method is used to thin the data, setting a thinning threshold (one feature point is retained when the distance between adjacent points is less than 0.2m). Redundant point data is removed, reducing the point cloud data size to 1GB, only 2% of the original data size, while fully preserving key features of the mine terrain such as peaks, valleys, and pits. Then, 20m×20m square tiles are used to slice the thinned point cloud map, dividing it into 1000 independent point cloud tiles, each with a data size of approximately 1MB. When constructing global terrain data, the preprocessed point cloud map can accurately call the point cloud tiles of the corresponding area according to the partition range of the target raster map, and merge them with the local raster terrain data and the target raster map. This avoids the inefficiency caused by loading the full point cloud data, ensures the accuracy of terrain data construction, and significantly reduces the pressure of data storage and transmission.

[0132] In some embodiments of this application, generating autonomous driving map data for a target work scenario based on global terrain data and work boundary reference data includes: aligning spatial coordinates based on the global terrain data and work boundary reference data to obtain coordinate-matched global terrain data and work boundary reference data; correcting terrain elevation deviations and processing areas with abrupt changes in terrain slope based on the coordinate-matched global terrain data to obtain accuracy-optimized global terrain data; performing fusion encoding based on the accuracy-optimized global terrain data and work boundary reference data to obtain a map data prototype containing three-dimensional terrain information and work boundary markers; and supplementing road network topology and work point labeling information according to map data specifications recognizable by the autonomous driving system based on the map data prototype to obtain autonomous driving map data for the target work scenario.

[0133] In this embodiment, spatial coordinate alignment refers to aligning the coordinate systems of the global terrain data and the working boundary reference data to the same preset coordinate system (such as the local coordinate system or global coordinate system of the target working scene), and adjusting the spatial positions of the two types of data to ensure that the working boundary reference data can accurately match the corresponding spatial position of the global terrain data. The core is to eliminate the coordinate deviation between the two types of data.

[0134] The coordinate-matched global terrain data and working boundary reference data refer to the global terrain data and working boundary reference data that have been processed by spatial coordinate alignment and whose coordinate system and spatial location are accurately matched. The two types of data can correspond to each other within the same spatial framework, laying the foundation for subsequent fusion coding.

[0135] Topographic elevation deviation refers to the difference between the topographic elevation value in the global topographic data and the actual topographic elevation value. The causes of deviation include sensor acquisition error, data processing error, etc., and it needs to be corrected to ensure the accuracy of topographic elevation.

[0136] Abrupt changes in terrain slope refer to areas in the overall terrain data where the slope value changes abruptly and significantly. These areas are usually abnormal regions that occur during data processing and do not conform to the natural variation patterns of actual terrain, thus requiring smoothing processing.

[0137] Accuracy-optimized full-domain terrain data refers to full-domain terrain data that has undergone terrain elevation deviation correction and slope change abrupt area processing, resulting in more accurate terrain parameters and terrain changes that better reflect the actual situation, thus possessing higher accuracy and reliability.

[0138] Fusion coding refers to the process of integrating the optimized global terrain data and the working boundary reference data into a whole, and formatting the data according to the preset coding rules to transform the three-dimensional terrain information and the working boundary markers into structured data that can be recognized and stored by computers.

[0139] Map data prototypes refer to the initial map data generated after fusion encoding, containing 3D terrain information and work boundary markers. This is the basic version of autonomous driving map data, but it lacks supplementary information such as road network topology and work point annotations. Map data specifications recognizable by autonomous driving systems refer to the technical specifications such as map data format, data structure, and parameter indicators preset by the autonomous driving system. Different types of autonomous driving systems may have different specification requirements; map data must conform to these specifications to be recognized and used by the autonomous driving system. Road network topology refers to the connection relationships and traffic rules of roads in the target work scenario, including information such as the start and end points of roads, intersections, traffic directions, and number of lanes. This is the core information guiding the autonomous vehicle's driving route. Work point annotations refer to the identification information of key work locations in the target work scenario, including the name, type (e.g., loading point, unloading point, maintenance point), coordinates, and work area of ​​the work point, facilitating accurate identification of work locations by autonomous vehicles.

[0140] This application achieves the following technical effects by clearly defining the steps for generating autonomous driving map data: First, spatial coordinate alignment ensures the spatial consistency between the overall terrain data and the working boundary reference data, laying the foundation for the fusion of the two types of data and avoiding map data errors caused by coordinate deviations. Second, terrain accuracy optimization corrects elevation deviations and abrupt slope changes, improving the accuracy and rationality of the overall terrain data and ensuring that the map data truly reflects the actual terrain. Third, fusion coding integrates terrain data and boundary data into a structured map data prototype. After supplementing road network topology and work point annotations, it generates final map data that meets the requirements of autonomous driving systems, realizing the transformation from basic data to practical map data and ensuring that the generated map data can directly support autonomous driving operations.

[0141] In an open-pit mine scenario, the specific process for generating autonomous driving map data is as follows: First, the global terrain data (based on the mine's local coordinate system) and the work boundary reference data (initially based on the global coordinate system) are spatially aligned. The work boundary reference data is then converted to the mine's local coordinate system. A coordinate transformation algorithm (such as the seven-parameter transformation method) is used to eliminate coordinate deviations between the two types of data, resulting in coordinate-matched global terrain data and work boundary reference data. This ensures that the work boundary markers accurately correspond to the corresponding positions in the global terrain data. Second, the accuracy of the coordinate-matched global terrain data is optimized. By comparing it with the elevation control point data measured in the field, terrain elevation deviations are corrected (areas with elevation deviations greater than 0.3m are adjusted to the actual elevation values). A moving average method is used to process areas with abrupt changes in terrain slope (areas with slope changes greater than 10° are smoothed to make the slope changes conform to the actual terrain patterns of the mine), resulting in accuracy-optimized global terrain data. Next, the optimized global terrain data and work boundary reference data are fused and encoded. The three-dimensional terrain information (elevation, slope) and work boundary identifiers (boundary coordinates, boundary type) are integrated into structured data using binary encoding format to generate a map data prototype. Finally, in accordance with the map data specifications for unmanned mining systems (supported map format is .pcd, road network topology data is stored in XML format, and work point labels are stored in JSON format), the mine's road network topology information (including the starting point, ending point, intersection, and one-way traffic rules of the three main transportation roads) and work point label information (including the names, coordinates, and work areas of the five loading points and three unloading points) are supplemented. Finally, autonomous driving map data for the open-pit mine scene is generated. This map data can be directly recognized and used by the autonomous driving system of unmanned mining trucks to guide unmanned mining trucks to safely and efficiently complete the "loading-transporting-unloading" operation within the defined boundaries according to the preset route.

[0142] like Figure 2 As shown, Figure 2This diagram illustrates the overall system for crowdsourced data collection, processing, and map generation for autonomous driving maps in open-pit mines. It details the entire data acquisition chain: data collection, transmission, preprocessing, fusion calculation, and map output. It clearly reflects the core system architecture and data flow relationships of "manned map acquisition platform unit (platform mapping vehicle) + unmanned mining truck unit (crowdsourced mapping mining truck) + server-side map calculation unit (cloud-based map platform)." During downtime, the platform mapping vehicle collects comprehensive basic data (raw point cloud files, LiDAR calibration parameters, etc.) and transmits it back via a high-speed network. The crowdsourced mapping mining truck... During operation, local operation data is collected synchronously and processed into local raster terrain by its own computing module before being uploaded in a small area. The cloud map platform receives the data and performs processing, thinning and tiling preprocessing, raster map calculation and stitching, multi-source data fusion and feature extraction to generate global terrain data and working boundary data, which are then synchronized to the map editing module as a base map for optimization and adjustment. Finally, a high-precision autonomous driving map is formed, which intuitively demonstrates the core solution of "dual-source acquisition, layered processing and multi-source fusion", as well as how to achieve the technical effect of no additional equipment, no impact on production and operation and reduced network pressure through the collaboration of existing equipment.

[0143] like Figure 3 As shown, an embodiment of this application provides an autonomous driving map data determination device 200, including: a first acquisition module 210, a second acquisition module 220, a third acquisition module 230, a first determination module 240, a second determination module 250, a third determination module 260, and a generation module 270. The first acquisition module 210 is used to acquire the overall basic data of the target operation scene during the downtime period and the local operation data of the target operation scene during the operation period; the second acquisition module 220 is used to acquire, based on the local operation data, the local raster terrain data corresponding to the local operation data and the first operation boundary feature data within the local area; the third acquisition module 230 is used to acquire, based on the overall basic data of the target operation scene during the downtime period and the local operation data during the operation period; the second acquisition module 220 is used to acquire, based on the local operation data, the local raster terrain data corresponding to the local operation data and the first operation boundary feature data within the local area; the third acquisition module 230 is used to acquire, based on the local basic data, the local raster terrain data corresponding to the local operation data and the first operation boundary feature data within the local area; the third acquisition module 230 is used to acquire, based on the local basic data, the local operation ... operation data, the local operation data, and the first operation boundary feature data within the local area; the third acquisition module 240 is used to acquire, based on the local basic data, the local operation data, the local operation data, and the first operation boundary feature data within the local area; the third acquisition module 240 is used to acquire, based on the local basic data, the local operation data, and the first operation boundary feature data within the local area; the third acquisition module 240 is used to acquire, based on the local basic data, the local operation data, and the first operation boundary feature data within the local area; the The system acquires basic domain data, including point cloud maps, initial raster maps of multiple local areas within the entire domain, and second operational boundary feature data within the entire domain. A first determining module 240 determines a target raster map covering the entire target operational scenario based on the multiple initial raster maps. A second determining module 250 determines the entire domain terrain data based on local raster terrain data, the preprocessed point cloud map, and the target raster map. A third determining module 260 determines operational boundary reference data based on the first and second operational boundary feature data. A generation module 270 generates autonomous driving map data for the target operational scenario based on the entire domain terrain data and the operational boundary reference data.

[0144] The autonomous driving map data determination device 200 of this application acquires full-domain basic data and local operation data in time periods through the first acquisition module 210, which can supplement the data source without the need for additional equipment, avoiding frequent downtime caused by relying on dedicated data collection vehicles. The second acquisition module 220 uses the computing power module of the unmanned operation vehicle to locally solve local raster terrain data and uploads it with a preset range. The third acquisition module 230 performs thinning and slicing preprocessing on the point cloud map, which greatly reduces the network transmission pressure and is suitable for the harsh network environment of remote scenarios such as open-pit mines. The first determination module 240 generates a full-domain target raster map through partition calculation and stitching. The second determination module 250 integrates multi-source terrain data to construct accurate full-domain terrain data. The third determination module 260 integrates and verifies the two types of operation boundary feature data to ensure boundary accuracy. Finally, the generation module 270 generates map data that meets the requirements of the autonomous driving system. The multi-module collaboration realizes efficient acquisition, transmission and generation of map data, which not only improves the operational efficiency of target operation scenarios such as open-pit mines, but also enhances the accuracy and reliability of map data, effectively solving the pain point of high dependence on data collection vehicles and network environment in existing technologies.

[0145] like Figure 4 As shown, an embodiment of this application provides an autonomous driving map data determination device 300, including a processor 302 and a memory 304. The memory 304 stores a program or instructions. When the processor 302 executes the program or instructions in the memory 304, it implements the steps of the autonomous driving map data determination method as described in any of the above embodiments. Therefore, the autonomous driving map data determination device 300 possesses all the beneficial effects of the autonomous driving map data determination method as described in any of the above embodiments.

[0146] This application provides a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps of the autonomous driving map data determination method as described in any of the above embodiments. Therefore, the readable storage medium possesses all the beneficial effects of the autonomous driving map data determination method as described in any of the above embodiments.

[0147] In the claims, description, and accompanying drawings of this application, the term "plural" refers to two or more objects. Unless otherwise explicitly defined, the terms "upper," "lower," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used solely for the convenience of describing this application and simplifying the descriptive process, and are not intended to indicate or imply that the device or element referred to must have the described specific orientation, or be constructed and operated in a specific orientation. Therefore, these descriptions should not be construed as limitations on this application. The terms "connection," "installation," "fixing," etc., should be interpreted broadly. For example, "connection" can be a fixed connection between multiple objects, a detachable connection between multiple objects, or an integral connection; it can be a direct connection between multiple objects or an indirect connection between multiple objects through an intermediate medium. For those skilled in the art, the specific meaning of the above terms in this application can be understood based on the specific circumstances described above.

[0148] In the claims, description, and accompanying drawings of this application, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In the claims, description, and accompanying drawings of this application, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0149] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for determining autonomous driving map data, characterized in that, include: Acquire the overall basic data of the target work scenario during the downtime period, as well as the local work data of the target work scenario during the operation period; Based on the local operation data, obtain the local raster terrain data corresponding to the local operation data, and the first operation boundary feature data within the local area; Based on the aforementioned global basic data, a point cloud map, initial raster maps of multiple local areas within the global domain, and second operation boundary feature data within the global domain are obtained. Based on the multiple initial grid maps, a target grid map covering the entire target operation scenario is determined. Based on the local raster terrain data, the preprocessed point cloud map, and the target raster map, determine the global terrain data; Based on the first work boundary feature data and the second work boundary feature data, determine the work boundary reference data; Based on the global terrain data and the working boundary reference data, autonomous driving map data for the target operation scenario is generated.

2. The method for determining autonomous driving map data according to claim 1, characterized in that, The overall basic data includes the original point cloud file and the lidar calibration parameters, and the local operation data includes the original point cloud file and the lidar calibration parameters after being filtered by preset rules. The downtime periods are shift change periods, inspection periods, or maintenance periods of the target work scenario. The overall basic data is collected at a preset speed and route and then uploaded via a high-speed network. The local work data is collected synchronously with the operation process of the target work scenario.

3. The method for determining autonomous driving map data according to claim 1, characterized in that, The step of obtaining local raster terrain data corresponding to the local operation data based on the local operation data includes: The local operation data is determined by processing the local operation data using the computing power module of the unmanned operation vehicle in the target operation scenario; The local raster terrain data is transmitted to the computing terminal using a preset range upload method; The preset range upload method involves uploading local raster terrain data that is currently related to the operation of the target work scenario.

4. The method for determining autonomous driving map data according to claim 1, characterized in that, The first operation boundary feature data consists of terrain edge and operation point boundary feature information of a local operation area, while the second operation boundary feature data consists of terrain boundary and fixed facility boundary feature information of the entire area. The step of determining the work boundary reference data based on the first work boundary feature data and the second work boundary feature data includes: The first and second job boundary feature data are fused and verified to remove abnormal feature information and obtain job boundary reference data.

5. The method for determining autonomous driving map data according to claim 1, characterized in that, The step of determining the target grid map covering the entire target operation scene based on multiple initial grid maps includes: For each of the initial raster maps, calculate the terrain elevation and slope to obtain multiple calculated initial raster maps; The calculated initial grid maps are edge-matched and stitched together according to the spatial coordinate relationship of the target operation scene to obtain a target grid map covering the entire target operation scene.

6. The method for determining autonomous driving map data according to claim 1, characterized in that, After obtaining the point cloud map based on the global basic data, the method further includes: The point cloud map is thinned and sliced ​​to obtain a preprocessed point cloud map.

7. The method for determining autonomous driving map data according to claim 1, characterized in that, The process of generating autonomous driving map data for the target operation scenario based on the global terrain data and the work boundary reference data includes: Based on the global terrain data and the working boundary reference data, spatial coordinate alignment is performed to obtain the coordinate-matched global terrain data and working boundary reference data; Based on the coordinate-matched global terrain data, the terrain elevation deviation is corrected, areas with abrupt changes in terrain slope are processed, and the global terrain data with optimized accuracy is obtained. Based on the precision-optimized full-domain terrain data and the work boundary reference data, fusion encoding is performed to obtain a map data prototype containing three-dimensional terrain information and work boundary markers; Based on the map data prototype, and in accordance with the map data specifications that can be recognized by the autonomous driving system, supplement the road network topology and work point labeling information to obtain the autonomous driving map data for the target work scenario.

8. A device for determining autonomous driving map data, characterized in that, include: The first acquisition module is used to acquire the overall basic data of the target work scenario during the downtime period, as well as the local work data of the target work scenario during the operation period. The second acquisition module is used to acquire, based on the local operation data, local raster terrain data corresponding to the local operation data, and first operation boundary feature data within the local area; The third acquisition module is used to acquire point cloud map, initial raster map of multiple local areas within the whole domain, and second operation boundary feature data within the whole domain based on the whole domain basic data. The first determining module is used to determine a target grid map covering the entire target operation scene based on the multiple initial grid maps; The second determining module is used to determine the global terrain data based on the local raster terrain data, the preprocessed point cloud map, and the target raster map. The third determining module is used to determine the work boundary reference data based on the first work boundary feature data and the second work boundary feature data; The generation module is used to generate autonomous driving map data for the target operation scenario based on the global terrain data and the working boundary reference data.

9. A device for determining autonomous driving map data, characterized in that, include: processor; A memory storing programs or instructions, wherein the processor, when executing the programs or instructions in the memory, implements the steps of the method for determining autonomous driving map data as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the method for determining autonomous driving map data as described in any one of claims 1 to 7.

Images