High-precision map data storage medium and high-precision map automatic generation method

Abstract

The high-precision map data storage medium and automatic high-precision map data generation method disclosed in this invention eliminates manual parallel grouping and organizes the entire high-precision map data based on the lane centerline, the primary element used in autonomous driving to express traffic. This data storage system records the lateral relationships between lane centerlines, lane centerlines and lane boundaries, and lane centerlines and road boundaries, enabling functions such as lateral vehicle control. These lateral relationships are fully automatically calculated using a spatial matching algorithm, eliminating the need for manual intervention. This reduces labor costs, improves data production efficiency, shortens the data production cycle, and enhances data freshness.

Description

Technical Field

[0001] The present invention belongs to the field of autonomous driving, and specifically relates to a high-precision map data storage medium and a high-precision map automatic generation method. Background Art

[0002] With the rapid development of autonomous driving, the application of high-precision maps in this field has become increasingly prominent. An increasing number of automakers are incorporating high-precision maps into their development. However, challenges have arisen with the production of high-precision maps by traditional map vendors. Traditional high-precision map production relies on a centralized collection model, requiring a fleet of vehicles equipped with specialized equipment (such as laser point cloud, millimeter-wave radar, cameras, and inertial navigation) to collect and transmit data. High-precision maps are then generated through processes including point cloud preprocessing, image recognition, manual mapping, and data publishing. The entire data production process is lengthy and labor-intensive. Current mapping procedures limit high-precision map production to quarterly updates. While these maps offer high accuracy, they lack freshness, failing to meet the high freshness requirements of autonomous driving.

[0003] Patent publication number CN110083668A discloses a data management system, management method, terminal and storage medium for high-precision maps. It mainly divides the scene map into several local maps, and realizes accurate splicing of local maps through coordinate conversion relationships, so as to realize orderly management and classification of indoor multi-layer, cross-layer, underground and underground multi-layer high-precision map data. Then, based on the terminal positioning position, the local map near the terminal and the traffic sign information and traffic rule information contained in the local map are loaded to reduce the terminal loading and calculation amount. This mapping method can support the application of driving and parking, but it does not solve the problem that the data requires a lot of manual processing and the labor cost is high.

[0004] The patent with publication number CN114490910A discloses a map generation method, device, electronic device and storage medium. The method is applied to a mapping platform, including: obtaining vehicle-side collection data collected by a target autonomous driving vehicle; standardizing the vehicle-side collection data according to the target hardware equipment information of the target autonomous driving vehicle to obtain standardized collection data that can be recognized by the mapping platform; and generating an operational map based on the standardized collection data. Through this technical solution, the mapping platform can be decoupled from the collection vehicle, reducing the cost of data collection. However, the standardization processing still refers to conventional high-precision map processing, and it is impossible to achieve the purpose of automatically producing high-precision maps and reducing the cost of manual mapping. Summary of the Invention

[0005] The high-precision map data storage medium disclosed in the present invention simplifies the data storage model in the data production process and can automatically produce high-precision map data. While meeting the needs of autonomous driving applications, it improves data production efficiency, thereby reducing data production costs and improving data freshness.

[0006] The present invention also discloses a method for automatically generating high-precision maps, which simplifies the data storage model in the data production process and can automatically produce high-precision map data. While meeting the needs of autonomous driving applications, it improves data production efficiency, thereby reducing data production costs and improving data freshness.

[0007] The high-precision map data storage medium disclosed in the present invention includes a lane line module, which stores the lane centerline information required to realize the lateral vehicle control function, the vehicle lane boundary information, and the lateral left and right correlation information between the vehicle lane centerline and other lane centerlines, and between the vehicle lane centerline and lane boundary. The lane line module data is used at the downstream vehicle dispatching end to cooperate in realizing the vehicle lateral control function.

[0008] Furthermore, it also includes a road boundary module, which contains road boundary information, and the lane line module stores the lateral left and right correlation relationship information between the center line of the lane and the road boundary.

[0009] Furthermore, the lane centerline information includes the lane centerline identifier of the vehicle, the lateral correlation between the lane centerline of the vehicle and other lane centerlines is determined by associating the lane centerline identifier of the vehicle with the left and / or right lane centerline identifiers, the lateral correlation between the lane centerline of the vehicle and the lane boundary is determined by associating the lane centerline identifier of the vehicle with the left and / or right lane dividing line identifiers, and the left-right lateral correlation between the lane centerline of the vehicle and the road boundary is determined by associating the lane centerline of the vehicle with the left and / or right road boundary identifiers.

[0010] Furthermore, after the lane centerline is automatically derived from the lane lines on both sides perceived and recognized, in scenarios without lane lines, point connections are automatically formed based on the lane centerline nodes through automated judgment of the vehicle trajectory.

[0011] Furthermore, the road boundary module further includes at least one of guardrail, curb, and paved edge road boundary data.

[0012] Furthermore, it also includes at least one of an aerial element module, a ground identification module, a lane driving attribute module and an interchange relationship module.

[0013] The aerial element module includes traffic lights, traffic sign information, etc.

[0014] The ground identification module includes ground pedestrian and vehicle driving prompt information, etc.

[0015] The lane driving attribute module includes lane driving restriction direction information, restriction time information and driving slope curvature information.

[0016] The interchange relationship module includes interchange point information and interchange level information.

[0017] The lane line module is associated with the corresponding road section information in the aerial element module, the ground identification module, the lane driving attribute module, and the interchange relationship module through the center line of the lane.

[0018] Furthermore, the lane driving attribute module calculates ADAS attributes and lane restriction attributes based on data from the lane line module, road boundary module, aerial element module, and ground sign module;

[0019] The interchange relationship module is obtained by calculating interchange relationship attributes based on data from the lane line module, the road boundary module, the aerial element module, and the ground identification module.

[0020] The present invention also discloses a method for automatically generating high-precision map data, comprising the following steps:

[0021] Step 1) The lane centerline is automatically generated from the lane lines on both sides of the perception recognition to form a discontinuous lane centerline. According to the trajectory, the discontinuous lane centerlines are connected longitudinally point by point.

[0022] Step 2) establishing associations between lane modules and between lane modules and road boundary modules. The lane modules store information on the centerline of the vehicle's lane required for lateral vehicle control, information on lateral associations between the centerline of the vehicle's lane and other lane centerlines, information on lateral associations between the centerline of the vehicle's lane and the lane boundary, and information on lateral associations between the centerline of the vehicle's lane and the road boundary. The lane module data is used by the dispatching end to implement the vehicle's lateral control function.

[0023] Step 3) establishing an association relationship between the lane line module, the aerial element module, and the ground identification module; the lane line module is associated with the corresponding road section information in the aerial element module and the ground identification module through the center line of the lane;

[0024] The aerial element module includes at least one of traffic light and traffic sign information;

[0025] The ground identification module includes ground pedestrian and vehicle driving prompt information.

[0026] Furthermore, step 4) is also included.

[0027] Step 4) establishing an association relationship between the lane line module, the lane driving attribute module, and the interchange relationship module;

[0028] The lane driving attribute module is obtained by calculating ADAS attributes and lane restriction attributes based on the data of the lane line module, road boundary module, aerial element module, and ground sign module;

[0029] The interchange relationship module is obtained by calculating the interchange relationship attributes based on the data of the lane line module, the road boundary module, the aerial element module, and the ground sign module;

[0030] The interchange relationship module includes interchange point and interchange level information; the lane driving attribute module includes the lane driving restriction direction, restriction time information and driving slope information;

[0031] The lane line module is associated with the corresponding road section information in the lane driving attribute module and the interchange relationship module through the center line of the lane.

[0032] Furthermore, discontinuous lane centerlines are automatically generated based on the lane lines and road boundaries perceived and recognized by the vehicle side. Based on the vehicle trajectory uploaded by the vehicle side, the discontinuous lane centerlines are connected longitudinally point by point.

[0033] Further,

[0034] Step 1) includes the following steps before,

[0035] The vehicle perceives and identifies lanes and related mapping information in real time, combines vehicle posture and combined inertial navigation information to create a local semantic map on the vehicle side and upload it to the cloud;

[0036] Step 4 is followed by the following steps:

[0037] Build a high-precision map data quality inspection library to automatically check the high-precision map data; the high-precision map data that has passed the quality inspection will be sent to the vehicle end after data compilation and compliance processing.

[0038] The beneficial technical effects of the present invention are as follows: the high-precision map data storage medium and automatic high-precision map data generation method disclosed in the present invention do not manually interrupt parallel grouping. Instead, they organize the entire high-precision map data storage based on the lane centerline element—the main element used to express traffic in autonomous driving. By recording the lateral left-right relationships between lane centerlines, lane centerlines and lane boundaries, and lane centerlines and road boundaries, functions such as lateral vehicle control can be achieved. This lateral relationship can be fully automatically calculated based on a spatial matching algorithm, without manual intervention, reducing labor costs, improving data production efficiency, shortening the data production cycle, and improving data freshness. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flow chart for automatically generating high-precision maps of the present invention;

[0040] Figure 2This is a schematic diagram of a high-precision map data storage medium according to the present invention;

[0041] Figure 3 This is a schematic diagram of the information relationship of the high-precision map data storage medium of the present invention;

[0042] Figure 4 A schematic diagram comparing the high-precision map production results of the present invention and the traditional high-precision map production results;

[0043] Figure 5 This is a schematic diagram of the centerline generation process in the element processing process of the high-precision map making link of the present invention;

[0044] Figure 6 A schematic diagram is established for the relationship between various elements in the element processing process of the high-precision map making link of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be described in detail below with reference to the accompanying drawings.

[0046] like Figures 1 to 6 As shown, the method for automatically generating high-precision maps includes the following steps:

[0047] The first step is real-time perception and recognition on the vehicle side

[0048] For commercial vehicles in mass production, autonomous vehicles capture images during operation and leverage the vehicle's real-time perception and recognition capabilities to identify lanes and map-related information in real time. This approach, leveraging the autonomous vehicle's inherent perception and recognition capabilities, eliminates the need to upload images to the cloud for recognition, as is done in conventional high-precision map production. This reduces data transmission, the cost of cloud-based image recognition, and the cycle time required.

[0049] Step 2: Generate and upload the vehicle-side local semantic map

[0050] With the help of the real-time perception and recognition results in the first step, combined with the vehicle posture and combined inertial navigation information, mapping is carried out on the vehicle side to obtain a local semantic map; the mapping results are uploaded to the cloud through the vehicle-cloud protocol.

[0051] Step 3: Cloud map learning

[0052] Based on the results of multiple real-time mapping, map learning is carried out, which mainly includes fully automated processing such as dirty data cleaning, cluster analysis, and aggregation analysis to form object data.

[0053] Step 4: Cloud Mapping

[0054] Based on the object data produced by map learning, high-precision map data is automatically generated according to the storage medium requirements designed by the present invention. The specific steps are as follows:

[0055] Step 1) Figure 5 As shown, the lane line module, road boundary module, aerial element module, ground identification module, and lane driving attribute module are initialized.

[0056] The lane line module includes lane lines and lane center lines. Lane lines are obtained by vehicle-side perception and recognition, and basic lane line attribute data is assigned. The basic lane line attribute data includes lane line shape, color, road boundary type, etc. The lane center line is currently empty.

[0057] The road boundary module also includes at least one of the guardrail, curb, and paved edge road boundary data, and the road boundary module is obtained by vehicle-side perception and recognition.

[0058] The aerial element module includes at least one of traffic light and traffic sign information.

[0059] The ground sign module includes at least one of the ground pedestrian and vehicle driving prompt information. The vehicle driving prompt information includes ground traffic signs such as stop lines, speed bumps, and ground guide arrows.

[0060] The lane driving attribute module includes the lane driving restriction direction, restriction time information and driving slope curvature information.

[0061] Step 2) Automatically generate discontinuous lane centerlines based on the lane lines and road boundaries perceived by the vehicle. According to the vehicle trajectory uploaded by the vehicle, the discontinuous lane centerlines are connected longitudinally point by point.

[0062] Since the lane lines and road boundaries recognized by the vehicle are incomplete, the generated lane center lines may be discontinuous, as shown in the following figure. Figure 5 Traditional high-precision map storage requires manual drawing of virtual lane lines in locations without actual lane lines to ensure lane groups can be formed, which incurs significant labor costs. The storage medium designed in this invention eliminates the need for manual drawing of virtual lane lines in locations without actual lane lines and allows lane lines to be disconnected based on the collected data.

[0063] According to the uploaded vehicle trajectory, an example of the center line of the discontinuous lanes connected longitudinally is shown in the attached figure. Figure 5 As shown in Figure 1, through trajectory mining, it can be found that it is passable from point A to point B and from point C to point D. The connectivity relationship is stored according to the storage method in Table 1. This ensures that lane-level path planning can be achieved.

[0064] Table 1

[0065]

[0066] Step 3) Establish associations between lane modules and between lane modules and road boundary modules. The lane module stores the lane centerline information required to implement the lateral vehicle control function, the lateral association relationship information between the lane centerline of the vehicle and the centerlines of other lanes, the lateral association relationship information between the lane centerline and the lane boundary, and the lateral association relationship information between the lane centerline and the road boundary; the road boundary module includes at least one of the guardrail, curb, and paved edge road boundary data; the lane module data is used at the dispatching end to cooperate in implementing the vehicle lateral control function;

[0067] As attached Figure 6 As shown, after the association relationship is established, the lane line module stores the association relationship according to the storage method in Table 2. This makes it easier for the vehicle to use this relationship to change lanes during autonomous driving.

[0068] Table 2

[0069]

[0070] Step 3) Establishing an association relationship between the lane line module, the aerial element module, and the ground sign module; the lane line module is associated with the corresponding road section information in the aerial element module and the ground sign module through the center line of the lane;

[0071] The aerial element module includes at least one of traffic signal light and traffic sign information;

[0072] The ground identification module includes pedestrian and vehicle driving prompt information on the ground;

[0073] like Figure 6 As shown, through the spatial intersection algorithm, the association between ground arrows, crosswalks, etc. and lane centerlines is maintained. Ground arrows and crosswalks are included in the ground pedestrian and vehicle driving prompt information; by calculating the lane centerline closest to the traffic sign, the association between the traffic sign and the lane centerline is maintained; based on behavioral analysis, the traffic light is associated with the lane centerline on which it acts.

[0074] Step 4) Establish an association between the lane marking module, the lane driving attribute module, and the interchange relationship module. The lane driving attribute module includes the lane restriction information module and the ADAS attribute module. The lane restriction information module includes information about the restricted direction and time of driving in the lane; the ADAS attribute module includes information about driving slope and curvature. The interchange relationship module includes information about interchange points and interchange levels.

[0075] The lane driving attribute module calculates ADAS attributes and lane restriction attributes based on data from the lane line module, road boundary module, aerial element module, and ground sign module. The interchange relationship module calculates interchange relationship attributes based on data from the lane line module, road boundary module, aerial element module, and ground sign module.

[0076] The lane line module associates the corresponding road section information in the lane driving attribute module and the interchange relationship module through the center line of the lane;

[0077] Step 5: Fully automated data quality inspection

[0078] Build a map data quality inspection case library to ensure the quality of the generated high-precision map data by performing automated logical checks on the generated high-precision map data.

[0079] Step 6: Data compilation and release

[0080] After the data passes quality inspection and data compilation and compliance processing, it is sent to the vehicle through the SDK and then sent to the autonomous driving control, positioning, and rendering module applications through the vehicle-side map middleware.

[0081] The present invention also discloses a high-precision map data storage medium, including a lane line module and a road boundary module; the road boundary module contains road boundary information; the lane line module stores the lane centerline information required to realize the lateral vehicle control function, the vehicle lane boundary information, the lateral left and right correlation relationship information between the vehicle lane centerline and other lane centerlines, the lane centerline and lane boundary, and the lateral correlation relationship information between the lane centerline and the road boundary. The lane line module data is used at the downstream vehicle dispatching end to cooperate in realizing the vehicle lateral control function.

[0082] The lane centerline information includes the lane centerline marking of the vehicle, the lateral correlation between the lane centerline of the vehicle and the centerlines of other lanes is determined by the association between the lane centerline marking of the vehicle and the left and / or right lane centerline markings, the lateral correlation between the lane centerline and the lane boundary is determined by the association between the lane centerline marking of the vehicle and the left and / or right lane dividing line markings, and the left-right lateral correlation between the lane centerline and the road boundary is determined by the association between the lane centerline of the vehicle and the left and / or right road boundary markings.

[0083] The lane centerline is automatically formed by connecting lane centerline nodes through the automatic judgment of vehicle trajectory. The road boundary module also includes at least one of guardrail, curb, and paved edge road boundary data.

[0084] The high-precision map data storage medium also includes at least one of an aerial element module, a ground identification module, a lane driving attribute module, and an interchange relationship module.

[0085] The aerial element module includes traffic lights and traffic sign information;

[0086] The ground identification module includes pedestrian and vehicle driving prompt information on the ground;

[0087] The lane driving attribute module includes the lane driving restriction direction information, restriction time information and driving slope information; the interchange relationship module includes interchange point information and interchange level information;

[0088] The lane line module is associated with the corresponding road section information in the aerial element module, ground marking module, lane driving attribute module, and interchange relationship module through the center line of the lane.

[0089] The lane driving attribute module calculates ADAS attributes and lane restriction attributes based on data from the lane line module, road boundary module, aerial element module, and ground sign module;

[0090] The interchange relationship module is obtained by calculating the interchange relationship attributes based on the data from the lane line module, road boundary module, aerial element module, and ground sign module.

[0091] Conventional high-precision map storage media organize lane-level data elements based on lane group elements. Lane group elements require manual maintenance and interruption, and all lane lines, lane center lines, and road boundaries on the corresponding road form a lane group. In current high-precision map production, it is difficult to achieve automated leveling and interruption to generate lane group elements. This element requires a lot of manual maintenance, high production and update costs, and low manual production efficiency, resulting in a long overall update cycle. The high-precision map data storage medium and high-precision map automatic generation method disclosed in the present invention reduce manual maintenance elements, automatically generate the relationship between the lane center line and other lane line modules, road boundary modules, lane driving attribute modules, and interchange relationship modules, and issue downstream autonomous driving regulations and controls. With the help of the horizontal relationship between each lane and between the lane and the road driving environment, the lateral vehicle control function is completed. Based on the automatic generation of high-precision map targets that can be used for autonomous driving applications, the elements that require manual operation in conventional high-precision map production are removed, the efficiency of map data production is improved, and the data production and storage costs are reduced.

Claims

1. High-precision map data storage medium, characterized by: include, A lane line module stores information on the centerline and lane boundaries of the vehicle's lane required to implement the lateral vehicle control function, as well as information on the lateral left-right correlation between the centerline of the vehicle's lane and other lane centerlines, and between the centerline of the vehicle's lane and the lane boundary; a road boundary module contains road boundary information, and the lane line module stores information on the lateral left-right correlation between the centerline of the vehicle's lane and the road boundary; the module also includes an aerial element module, a ground identification module, a lane driving attribute module, and an interchange relationship module. The lane driving attribute module calculates ADAS attributes and lane restriction attributes based on data from the lane line module, road boundary module, aerial element module, and ground identification module; and the interchange relationship module calculates interchange relationship attributes based on data from the lane line module, road boundary module, aerial element module, and ground identification module. Automatically generate discontinuous lane centerlines based on lane lines and road boundaries perceived by the vehicle side. Connect the discontinuous lane centerlines point by point longitudinally based on the vehicle trajectory uploaded by the vehicle side. The horizontal left-right association relationship is automatically generated based on a spatial matching algorithm and dynamically updated based on the trajectory data uploaded by the vehicle; The lane line module data is used at the downstream vehicle dispatching end to realize the vehicle lateral control function.

2. The high-precision map data storage medium according to claim 1, wherein: The lane centerline information includes the lane centerline identifier of the vehicle, the lateral correlation between the lane centerline of the vehicle and the centerlines of other lanes is determined by associating the lane centerline identifier of the vehicle with the left and / or right lane centerline identifiers, the lateral correlation between the lane centerline and the lane boundary is determined by associating the lane centerline identifier of the vehicle with the left and / or right lane dividing line identifiers, and the left-right lateral correlation between the lane centerline and the road boundary is determined by associating the lane centerline of the lane with the left and / or right road boundary identifiers.

3. The high-precision map data storage medium according to any one of claims 1 or 2, characterized in that: The lane centerline is automatically generated by the lane line perceived and recognized. If there is no lane line in the field, it is automatically formed by connecting the existing lane centerline nodes through automatic judgment of the vehicle trajectory.

4. The high-precision map data storage medium according to claim 3, wherein: The road boundary module further includes at least one of guardrail, curb, and paved edge road boundary data.

5. The high-precision map data storage medium according to claim 4, wherein: The aerial element module includes traffic lights and traffic sign information; The ground identification module includes ground pedestrian and vehicle driving prompt information; The lane driving attribute module includes the lane driving restriction direction information, restriction time information and driving slope curvature information; The interchange relationship module includes interchange point information and interchange level information; The lane line module is associated with the corresponding road section information in the aerial element module, the ground identification module, the lane driving attribute module, and the interchange relationship module through the center line of the lane.

6. A method for automatically generating high-precision map data includes the following steps: Step 1) Automatically generate discontinuous lane centerlines by identifying lane lines on both sides of the lane, and connect the discontinuous lane centerlines point by point longitudinally according to the trajectory; Step 2) Establishing associations between lane modules and between lane modules and road boundary modules. The lane modules store information required for implementing the lateral vehicle control function, including the lane centerline information, the lateral association information between the lane centerline of the vehicle and the centerlines of other lanes, the lateral association information between the lane centerline of the vehicle and the lane boundary, and the lateral association information between the lane centerline of the vehicle and the road boundary. The lane module data is used by the dispatching end to implement the vehicle lateral control function. The lateral association information is automatically generated based on a spatial matching algorithm and dynamically updated based on the trajectory data uploaded by the vehicle. Step 3) establishing an association relationship between the lane module, the aerial element module, and the ground identification module; the lane module is associated with the corresponding road section information in the aerial element module and the ground identification module through the center line of the lane; The aerial element module includes at least one of traffic light and traffic sign information; The ground identification module includes ground pedestrian and vehicle driving prompt information; Step 4) Establish the association relationship between the lane line module, the lane driving attribute module, and the interchange relationship module; The lane driving attribute module is obtained by calculating ADAS attributes and lane restriction attributes based on the data of the lane line module, road boundary module, aerial element module, and ground sign module; The interchange relationship module is obtained by calculating the interchange relationship attributes between the lane center lines in the lane line module; The interchange relationship module includes interchange point and interchange level information; the lane driving attribute module includes the lane driving restriction direction, restriction time information and driving slope information; The lane line module is associated with the corresponding road section information in the lane driving attribute module and the interchange relationship module through the center line of the lane; Discontinuous lane centerlines are automatically generated based on the lane lines and road boundaries perceived by the vehicle side. Based on the vehicle trajectory uploaded by the vehicle side, the discontinuous lane centerlines are connected longitudinally point by point.

7. The method for automatically generating high-precision map data according to claim 6, wherein: Discontinuous lane centerlines are automatically generated based on the lane lines and road boundaries perceived by the vehicle side. Based on the vehicle trajectory uploaded by the vehicle side, the discontinuous lane centerlines are connected longitudinally point by point.

8. The method for automatically generating high-precision map data according to claim 6, wherein: Step 1) includes the following steps before, The vehicle perceives and identifies lanes and related mapping information in real time, combines vehicle posture and combined inertial navigation information to create a local semantic map on the vehicle side and upload it to the cloud; Step 4 is followed by the following steps: Build a high-precision map data quality inspection library to automatically check the high-precision map data; the high-precision map data that has passed the quality inspection will be sent to the vehicle end after data compilation and compliance processing.

Images