A cloud management method of automatic driving map data

Abstract

The present application relates to the technical field of geographic information database, specifically to a cloud management method of automatic driving map data, comprising the following steps: dividing map data into static basic layer, quasi-dynamic layer and high dynamic layer, extracting road network and topological structure of the static basic layer, and traffic rules and road speed limit of the quasi-dynamic layer, combining into the basic attribute of each automatic driving section, and establishing a space-time layered geographic database. The present application effectively improves the updating efficiency and real-time performance of automatic driving map data by hierarchical division and processing of map data, uses the road network structure in the static basic layer data and the traffic rules and road speed limit in the quasi-dynamic layer data to combine to form the section basic attribute, and establishes a space-time layered geographic database according to the section basic attribute, realizes high correlation between spatial information and traffic rules, reduces data redundancy, and improves the accuracy and processing speed of data calling.

Description

Technical Field

[0001] This invention relates to the field of geographic information database technology, and in particular to a cloud management method for autonomous driving map data. Background Technology

[0002] The field of geographic information database technology mainly involves the collection, storage, organization, management and application of spatial geographic information data, specifically covering spatial data modeling, data indexing, spatial data hierarchical and spatiotemporal correlation, rapid retrieval of map information, dynamic updating of spatial data, and integrated management methods for spatial and non-spatial information.

[0003] While existing technologies enable the collection, storage, management, and retrieval of spatial geographic information data, practical applications suffer from significant drawbacks. Data updates involve extensive duplication of data, and data queries require traversing large amounts of redundant information, reducing processing efficiency. Furthermore, the lack of effective methods for measuring the real-time combined impact of road geometry, traffic node structure, information density, and dynamic events hinders the real-time reliability and accuracy of navigation decisions, leading to problems such as slow vehicle decision-making or inappropriate route planning in real-world applications. Therefore, improvements are necessary. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a cloud-based management method for autonomous driving map data.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a cloud management method for autonomous driving map data, comprising the following steps:

[0006] The map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. The road network and topology of the static base layer, as well as the traffic rules and road speed limits of the quasi-dynamic layer, are extracted and combined to form the basic attributes of each autonomous driving road segment, and a spatiotemporal hierarchical geographic database is established.

[0007] Based on the spatiotemporal hierarchical geographic database, a multidimensional cognitive load factor set is established for each road segment in the autonomous driving road network. Based on the multidimensional cognitive load factor set, the road segment navigation cognitive load index is calculated.

[0008] Based on the spatiotemporal hierarchical geographic database and the road segment navigation cognitive load index, when high dynamic layer data is received, a unique timestamp is assigned to the data, dynamic event data with version identifier is generated, and the road segment navigation cognitive load index is recalculated according to the road segment associated with the dynamic event data with version identifier. The dynamic event data with version identifier and the road segment navigation cognitive load index are stored as a new version, and a multi-version map data snapshot is established.

[0009] Based on the multi-version map data snapshots, the system receives map data requests and request timestamps sent by the autonomous vehicle terminal. According to the request timestamps, it retrieves the latest static base layer, quasi-dynamic layer, and high-dynamic layer versions from the multi-version map data snapshots, combines them into a consistent map data view, and encapsulates the consistent map data view with the corresponding road segment navigation cognitive load index to generate a navigation path decision data package.

[0010] Preferably, the steps for obtaining the spatiotemporal hierarchical geographic database are as follows:

[0011] The map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. Based on the spatial location information and road network geometric features of the map data, the static base layer data is subjected to spatial structure identification and road connection relationship analysis. The road network information and topology of each road in the static base layer are extracted one by one. The high-dynamic layer includes real-time traffic accidents, road congestion events, temporary road construction, and temporary road closures, forming a set of road network topology structures for the static base layer.

[0012] Based on the static basic layer road network topology set, the quasi-dynamic layer data is matched and parsed segment by segment. The traffic rule information and road speed limit value corresponding to each road in the quasi-dynamic layer are extracted one by one. The road network topology, traffic rule information and road speed limit value are mapped and fused one by one to obtain the basic attributes of the autonomous driving road segment.

[0013] Based on the basic attributes of the autonomous driving road segments, the roads are spatially organized according to their geometric coordinates. By establishing a corresponding index between the road spatial coordinates and attributes, and using the spatial coordinate range of the road as the index rectangle, corresponding R-Tree index nodes are generated one by one to complete the spatial organization and association of all road segments, forming a spatiotemporal hierarchical geographic database.

[0014] Preferably, the steps for obtaining the multidimensional cognitive load factor set are as follows:

[0015] Based on the aforementioned spatiotemporal hierarchical geographic database, the road curvature value, road longitudinal slope angle, and lane lateral width value of each road segment in the autonomous driving road network are extracted one by one. The segments are identified by road segment number and their corresponding values ​​are recorded. By summarizing the three parameters and establishing a unified numbering table, a set of road geometric factor parameters is generated.

[0016] Based on the set of road geometric factor parameters, the number of intersections, lane merging, and roundabouts included in each road segment is extracted. At the same time, the actual length of the corresponding road segment is calculated by combining the number of traffic signs and traffic lights. Each type of quantity is divided by the corresponding length to calculate the density per unit distance. Meanwhile, a reference standard value is introduced to make each density parameter dimensionless, resulting in a multidimensional cognitive load factor set.

[0017] Preferably, the step of obtaining the road segment navigation cognitive load index is as follows:

[0018] Based on the multidimensional cognitive load factor set, the navigation cognitive load index of the road segment is calculated.

[0019] Preferably, the steps for obtaining the dynamic event data with version identifier are as follows:

[0020] Based on the spatiotemporal hierarchical geographic database and the road segment navigation cognitive load index, the event occurrence time, location coordinates and event type in the received high dynamic layer data are analyzed. A unique timestamp based on the event occurrence time is assigned to each dynamic event, and a corresponding version identifier is added to the original dynamic event record to form dynamic event data with version identifier.

[0021] Preferably, the steps for obtaining the multi-version map data snapshot are as follows:

[0022] Based on the event location coordinates of the dynamic event data with version identifiers, each event is spatially matched with the spatial range coordinates of road segments in the spatiotemporal hierarchical geographic database. The road segment number affected by the dynamic event is determined according to the matching of spatial coordinates, forming a set of road segment factor parameters to be updated.

[0023] Based on the set of road segment factor parameters to be updated, the road segment navigation cognitive load index of the road segments affected by dynamic events is recalculated one by one. The recalculated road segment navigation cognitive load index is then merged with the dynamic event data with version identifiers to generate multi-version data records including dynamic events and road segment navigation cognitive load indices. Multi-version map data snapshots are formed using the event version identifier as an index.

[0024] Preferably, the steps for obtaining the consistent map data view are as follows:

[0025] Based on the multi-version map data snapshots, the map data requests sent by the autonomous vehicle terminal are parsed, and the request timestamps carried in each map data request are extracted one by one. The request timestamps are used as the reference retrieval time to determine the latest data version number in each version of data that is less than or equal to the request timestamp. The data versions corresponding to the static base layer, quasi-dynamic layer and high dynamic layer are retrieved respectively to form a set of map data versions with timestamp matching.

[0026] Based on the set of map data versions matched by the timestamp, spatial coordinate consistency is compared for each static base layer data version, quasi-dynamic layer data version, and high-dynamic layer data version. By unifying the road segment numbers, the road network topology, traffic rule information, road speed limit values, and dynamic event description information of the same road segments are merged one by one to form a consistent map data view with consistent matching of road elements and event information.

[0027] Preferably, the steps for obtaining the navigation path decision data packet are as follows:

[0028] Based on the consistent map data view, the road segment navigation cognitive load index of the corresponding road segment is called one by one. According to the unified data encapsulation protocol, the road network topology, traffic rule information, road speed limit value, dynamic event description information and road segment navigation cognitive load index of each road segment in the map data view are encapsulated one by one to form a navigation path decision data package containing navigation path, road information and cognitive load information.

[0029] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0030] This invention effectively improves the update efficiency and real-time performance of autonomous driving map data by hierarchically dividing and processing map data. It utilizes the road network structure in the static base layer data and the traffic rules and speed limits in the quasi-dynamic layer data to form basic road segment attributes. Based on these attributes, a spatiotemporal hierarchical geographic database is established, achieving a high degree of correlation between spatial information and traffic rules, reducing data redundancy, and improving the accuracy and processing speed of data retrieval. By extracting multi-dimensional factors such as road segment curvature, slope, lane width, node density, and information density to establish a cognitive load index, the comprehensiveness of road feature description is enhanced, improving the reliability and adaptability of road risk assessment and navigation path planning. Furthermore, by combining dynamic event data and assigning unique timestamps and version identifiers, the navigation cognitive load index of the corresponding road segment is recalculated when a dynamic event occurs, enabling real-time versioned updates and multi-version snapshot management of map data, improving dynamic tracking capabilities and the timeliness of accident response. Based on the timestamps provided by vehicle terminal requests, the latest versions of each layer of data are accurately retrieved and merged into a consistent view, encapsulated into a navigation decision data package. This ensures that autonomous vehicles can obtain the latest road condition information and road safety index in real time, improving the accuracy, safety, and stability of navigation decisions. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0033] Please see Figure 1 This invention provides a technical solution, a cloud management method for autonomous driving map data, comprising the following steps:

[0034] The map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. The road network and topology of the static base layer, as well as the traffic rules and road speed limits of the quasi-dynamic layer, are extracted and combined to form the basic attributes of each autonomous driving road segment, and a spatiotemporal hierarchical geographic database is established.

[0035] Based on a spatiotemporal hierarchical geographic database, a multidimensional cognitive load factor set is established for each road segment in the autonomous driving road network. Based on the multidimensional cognitive load factor set, the road segment navigation cognitive load index is calculated.

[0036] Based on the spatiotemporal hierarchical geographic database and the road segment navigation cognitive load index, when high dynamic layer data is received, a unique timestamp is assigned to the data, dynamic event data with version identifier is generated, and the road segment navigation cognitive load index is recalculated according to the road segment associated with the dynamic event data with version identifier. The dynamic event data with version identifier and the road segment navigation cognitive load index are stored as a new version, and a multi-version map data snapshot is established.

[0037] Based on multiple versions of map data snapshots, the system receives map data requests and request timestamps sent by autonomous vehicle terminals. According to the request timestamps, it retrieves the latest static base layer, quasi-dynamic layer, and high-dynamic layer versions from the multiple versions of map data snapshots, combines them into a consistent map data view, and encapsulates the consistent map data view with the corresponding road segment navigation cognitive load index to generate a navigation path decision data package.

[0038] The steps for obtaining a spatiotemporal hierarchical geodatabase are as follows:

[0039] The map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. Based on the spatial location information and road network geometric features of the map data, the static base layer data is subjected to spatial structure identification and road connection relationship analysis. The road network information and topology of each road in the static base layer are extracted one by one. The high-dynamic layer includes real-time traffic accidents, road congestion events, temporary road construction, and temporary road closures, forming a set of road network topology structures for the static base layer.

[0040] Based on the static base layer road network topology set, the quasi-dynamic layer data is matched and parsed segment by segment. The traffic rule information and road speed limit value corresponding to each road in the quasi-dynamic layer are extracted one by one. The road network topology, traffic rule information and road speed limit value are mapped and fused one by one to obtain the basic attributes of the autonomous driving road segment.

[0041] Based on the basic attributes of autonomous driving road segments, the road is spatially organized according to its geometric coordinates. By establishing a corresponding index between road spatial coordinates and attributes, and using the spatial coordinate range of the road as an index rectangle, corresponding R-Tree index nodes are generated one by one to complete the spatial organization and association of all road segments, forming a spatiotemporal hierarchical geographic database.

[0042] Specifically, the map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. This division is based on the data update frequency. Data with an update cycle longer than six months is classified as the static base layer, such as road centerlines, lane numbers, and permanent facilities like bridges and tunnels. Data with an update cycle between one day and six months is classified as the quasi-dynamic layer, such as traffic rules and speed limit changes. Data with an update cycle less than one day is classified as the high-dynamic layer, which includes real-time traffic accidents, road congestion events, temporary road construction, and temporary road closures. Subsequently, the system automatically retrieves the original vector map data from the Geographic Information System (GIS) database. For the static base layer data, a spatial structure identification and road connectivity analysis process is initiated. This process first uses the DBSCAN algorithm to cluster the endpoints of road centerlines to identify intersection nodes. Specifically, the starting and ending points of all road centerline geometries are extracted as a point dataset, and the neighborhood radius of the DBSCAN algorithm is set. The minimum number of neighborhood points required for the core object is 5 meters. A value of 2 means that any two or more road endpoints converging within a 5-meter radius are identified as an intersection or connection point and assigned a unique node ID. Endpoints not clustered are considered road ends and also assigned unique node IDs. After identifying all nodes, the system traverses each original road centerline, recording the node IDs connecting its two ends, forming a graph structure with nodes as vertices and the original road centerlines as edges. For each edge, i.e., a road segment, the system calculates its core geometric properties. For example, the segment length is obtained by summing the Euclidean distances of the segments forming the centerline polygon, while the road curvature... The system performs equidistant sampling along the centerline (e.g., one sampling point every 5 meters) and applies the curvature formula to calculate the curvature for every three consecutive sampling points. Finally, the average of all calculated values ​​is taken as the representative curvature of the road segment. For the construction of the road topology, the system generates an adjacency list. Each entry in the list records a node ID and a list of all road segment IDs directly connected to that node. In this way, the relationship between whether any two road segments are connected and how they are connected is clearly defined and solidified. Finally, the geometric attribute information of all road segments is integrated with the network topology connection relationship to form a static basic layer road network topology set.

[0043] Based on the static base layer road network topology set, the system initiates the matching and parsing process for quasi-dynamic layer data. Quasi-dynamic layer data typically exists in the form of a rule list with geographic coordinates and effective time, such as ("G001", (39.9°N, 116.3°E), "Speed ​​Limit", "80km / h", "All Vehicles"). The system first spatially locates each quasi-dynamic layer data record and matches it with road segments in the static base layer. The matching algorithm adopts a method based on buffer and overlap analysis. Specifically, for a point-like quasi-dynamic information (such as a traffic sign at a specific location), the system generates a circular buffer with a radius of 15 meters centered on its coordinates. Then, it retrieves all road segments in the static base layer road network topology set whose geometric center lines intersect with this buffer and associates the quasi-dynamic information with these candidate road segments. For linear quasi-dynamic information (such as traffic restriction rules on a road segment), the system directly determines the degree of overlap between its geometric line segment and the geometric center line of the static road segment. An overlap length threshold is set, which is determined by referencing the statistical distribution of the average road segment length. For example, a baseline value is set as 1.5 times the average length of all road segments plus one standard deviation. If the length of a road segment is less than this baseline value, the overlap length threshold is set to 50% of the segment's own length. If it is greater than the baseline value, it is set to a fixed 100 meters. Only when the overlap length exceeds this threshold is a match considered successful. After completing spatial matching, the system performs structured parsing on the matched quasi-dynamic data content, extracting key fields such as "traffic rule type," "specific value," "effective time," and "applicable vehicle type." For example, "no left turn" is parsed as {rule_type: The code `'turn_restriction', direction: 'left', value: 'forbidden'` parses "road speed limit 80" into `{rule_type: 'speed_limit', value: 80, unit: 'km / h'}`. Finally, the system maps and merges these parsed structured rule information with the attributes of the previously static road segments. Specifically, it adds or updates attribute fields such as "traffic rules" and "road speed limit" in the corresponding road segment records of the static basic layer road network topology set, filling in the parsed rule list and speed limit value. This combines the static geometric topology attributes of the road with the quasi-dynamic traffic rule attributes to obtain the basic attributes of the autonomous driving road segment.

[0044] Based on the fundamental attributes of autonomous driving road segments, the system initiates a spatial organization and index construction process to achieve efficient spatial querying of road segment data. The core of this process is to establish a spatial index for each road segment, specifically using an R-Tree index structure. First, the system traverses the fundamental attribute records of each autonomous driving road segment, reads the geometric coordinate sequence of its road centerline, and calculates the minimum bounding rectangle (MBR) of the road segment based on this coordinate sequence. This rectangle is the smallest rectangle that can completely contain the geometry of the road segment and whose sides are parallel to the coordinate axes. It is determined by the coordinates of its lower left and upper right corners (…). , , , The MBR is defined as the index rectangle for that road segment in the R-Tree. Subsequently, the system inserts these MBRs one by one into a newly created R-Tree. The node capacity parameters of the R-Tree are set based on empirical data; for example, the maximum number of entries for non-leaf nodes is set to 50, and the maximum number of entries for leaf nodes is set to 100. During the insertion process, whenever a new MBR is submitted, the R-Tree insertion algorithm recursively selects a child node starting from the root node, based on the "minimum area increment" principle. That is, it selects the child node whose own MBR area increases the least due to accommodating the new MBR, until it reaches a leaf node. If the selected leaf node has enough space, it directly stores the new MBR and the corresponding road segment's unique identifier; if the leaf node is full... If the node splits, the system uses a quadratic split algorithm to perform the split. This algorithm aims to distribute all MBRs (including the new MBRs) in the original node to two new nodes, minimizing the total area of ​​the MBRs in the two new nodes. This split process may propagate upwards to the parent node and eventually to the root node. After the insertion of all road segment MBRs is completed, a complete R-Tree index is formed. Each leaf node contains several road segment MBRs and their unique IDs. These IDs point to an independent attribute database with the road segment ID as the primary key. This database stores the complete basic attributes of autonomous driving road segments. Finally, this R-Tree spatial index and the attribute database together constitute a spatiotemporal hierarchical geographic database that allows upper-layer applications to perform fast spatial retrieval and attribute queries.

[0045] The steps for obtaining the multidimensional cognitive load factor set are as follows:

[0046] Based on a spatiotemporal hierarchical geographic database, the road curvature value, longitudinal slope angle and lane lateral width value of each road segment in the autonomous driving road network are extracted one by one. The road segments are identified by road segment number and their corresponding values ​​are recorded. By summarizing the three parameters and establishing a unified numbering table, a set of road geometric factor parameters is generated.

[0047] Based on the set of road geometric factor parameters, the number of intersections, lane merging, and roundabouts included in each road segment is extracted. At the same time, the actual length of the corresponding road segment is calculated by combining the number of traffic signs and traffic lights. Each type of quantity is divided by the corresponding length to calculate the density per unit distance. Meanwhile, reference standard values ​​are introduced to make each density parameter dimensionless, resulting in a multidimensional cognitive load factor set.

[0048] Specifically, based on a spatiotemporally layered geographic database, the system initiates a geometric factor extraction process for each autonomous driving road segment. First, for the extraction of road curvature values, the system queries the database for the centerline geometric data corresponding to each road segment identifier. This data consists of a series of ordered (X, Y, Z) coordinate points. The system resamples along the centerline at fixed intervals of 5 meters, generating a new set of equidistant sampling points. For every three consecutive sampling points P1, P2, and P3, the system calculates the curvature of the circle passing through these three points using a curvature formula. Specifically, it calculates four times the area of ​​the triangle formed by these three points and divides it by the product of the lengths of the three sides. This process is repeated along all sampling point sequences throughout the entire road segment, generating a series of curvature values. Finally, the system calculates the average of these curvature values ​​as the overall road curvature value for that road segment and records it. Second, for the extraction of the road's longitudinal slope angle, the system uses the Z-value (elevation) from the same set of centerline coordinate points. For every two consecutive sampling points, the system calculates the distance between the two points... The horizontal distance and elevation difference are used to calculate the slope angle of the small segment using the arctangent function. Similarly, the system calculates the slope angle of all small segments in the entire road segment and takes the average as the longitudinal slope angle of the road segment. For the lane lateral width value, the system directly reads the attribute field associated with the road segment number from the spatiotemporal hierarchical geographic database. This value is standardized data that has been entered during the data database construction phase. If a road segment contains multiple lanes with different widths, the width of the rightmost lane is extracted as the representative value. The system associates the three extracted parameters, namely the road curvature value, the road longitudinal slope angle, and the lane lateral width value, with the corresponding road segment number to form a record, for example (Road segment ID: "S001", curvature: 0.005, slope: 1.2, width: 3.5). By performing the above operations on all road segments in the road network and summarizing all records, a set of road geometric factor parameters is generated.

[0049] Based on the set of road geometric factor parameters, the system further expands the cognitive load-related factors for each road segment. First, it extracts the number of complex nodes contained in each road segment from the topological structure information of the spatiotemporal hierarchical geographic database. Specifically, the system checks the starting and ending nodes of the road segment; if a node connects to three or more roads, it is counted as an intersection. The system also analyzes lane geometry information; when the number of exit lanes in a road segment is less than the number of entrance lanes, and there are clear lane merging markings, it is counted as a lane merging. For roundabouts, the system identifies a group of one-way road segments that are connected end-to-end and form a closed loop, marking each road segment constituting a roundabout as containing one roundabout. Then, the system queries the quasi-dynamic layer data in the database to count the number of traffic signs and traffic lights that overlap with the spatial range of the current road segment. Simultaneously, it directly reads the precise length of the road segment in kilometers from the static base layer data. Next, the system performs density calculations, dividing the number of intersections, lane merging points, roundabouts, traffic signs, and traffic lights by the road segment length to obtain their respective unit distance densities. For example, a 2-kilometer-long road segment with 4 traffic signs has a traffic sign density of 2 per kilometer. To eliminate dimensional differences, the system introduces reference standard values ​​to process each density parameter. These reference standard values ​​are set based on statistical analysis of over 100,000 kilometers of roads of various levels across the country. For example, the reference density for traffic signs on urban expressways is set at 3 per kilometer, and the reference density for traffic lights is set at 0.5 per kilometer. The system divides the actual calculated density value by the corresponding reference standard value to obtain a dimensionless relative density value. Finally, the geometric factors of the road segment, the original number of nodes, the road segment length, and the dimensionless information density and node density factors are integrated to form a multidimensional cognitive load factor set.

[0050] The steps for obtaining the road segment navigation cognitive load index are as follows:

[0051] Based on a multidimensional cognitive load factor set, the navigation cognitive load index of a road segment is calculated using the following formula:

[0052] ;

[0053] in, For the first Navigation cognitive load index for each road segment For the first The road curvature value for each section. This serves as a reference standard value for road curvature. For the first The absolute value of the road slope angle for each section is used in the calculation. This is a reference standard value for road slope. For the first The lane width values ​​for each road segment. For standard comfort lane width, A constant for adjusting the sensitivity of the lane width index. The total weight coefficient of the geometric factor. These are the sub-factor weighting coefficients for curvature, slope, and width, respectively. The first Traffic sign density and traffic light density for each road segment (unit: signs / km). These are the reference density values ​​for traffic signs and traffic lights, respectively. For the overall weight of information factors, The first The number of intersections, lane merges, and roundabouts on each road segment. For the first The length of each road segment (in km). Reference density value for complex nodes (unit nodes / km). The weighting parameters are the structural complexity exponential amplification factors.

[0054] Specifically, the formula:

[0055] ;

[0056] The advantage of this formula lies in its comprehensive and precise quantification of the navigation cognitive load faced by autonomous vehicles on specific road sections through a multi-level, non-linear structure. The first part of the formula quantifies the basic physical load caused by road geometric features (curvature, slope, and width). The square terms of curvature and slope amplify the impact of adverse geometric conditions, while the exponential term of width precisely depicts the sharp increase in driving pressure when the lane narrows. The second part of the formula quantifies the information processing load and uses this load as the base. The exponential part uses the road structural complexity (density of intersections, lane changes, and roundabouts) as an amplification factor on the information load, simulating the real experience of human drivers: at complex intersections, even if the number of traffic signs is not large, the cognitive pressure of drivers will be amplified sharply, enabling more accurate identification of potentially high-risk, high-cognitive-load road sections and providing a safer decision-making basis for route planning.

[0057] , and The overall weight coefficients for geometric factors, information factors, and structural complexity are set using the Analytic Hierarchy Process (AHP). Ten senior traffic engineering and ergonomics experts were invited to conduct pairwise comparisons of the relative importance of the three factors (geometric, information, and structural) to driving cognitive load, constructing a judgment matrix. For example, if the experts generally believe that the importance of the information factor lies between that of the geometric factor and the structural complexity factor, then a judgment matrix can be obtained. By calculating the largest eigenvalue and its corresponding eigenvector of this matrix and performing normalization, the weights of each factor are obtained. Finally, the geometric mean of the results from the ten experts is taken. For example, the overall weight of the geometric factor is calculated. The overall weight of the information factor is 0.4. The weighting parameter of the structural complexity exponential amplification factor is 0.35. The sum of the three is 1, which is 0.25.

[0058] The sub-factor weight coefficients for curvature, slope, and width are determined similarly to the overall weights. Under the geometric factor category, experts compare the relative impacts of curvature, slope, and width on driving load pairwise, constructing a judgment matrix. For example, experts might consider changes in road curvature to have the greatest impact, followed by lane width, with slope having the least. The weights of the three factors are obtained by solving the normalized eigenvectors of the judgment matrix, and these weights must satisfy… For example, the sub-factor weighting coefficients of curvature are obtained by calculation. The sub-factor weighting coefficient for slope is 0.5. The sub-factor weight coefficient is 0.2 for width. It is 0.3.

[0059] , , , , , These serve as reference standard values ​​for various factors. These values ​​are derived from statistical analysis of large-scale actual road network data. Specifically, road data covering major cities nationwide was collected, and the values ​​of each parameter (curvature, slope, etc.) were statistically analyzed. The 85th percentile of its distribution was used as the reference standard value. This value represents a relatively common level that has begun to exert some stress on ordinary drivers. For example, the reference standard value for road curvature was obtained through statistical analysis. It is 0.01 (Corresponding to a radius of curvature of 100 meters), the reference standard value for road slope. 4.0 degrees, standard comfort lane width The reference density value for traffic signs is set at 3.75 meters. The reference density value for traffic lights is 5 per km. The reference density value for complex nodes is 2 per km. It is 2.5 units / km.

[0060] To adjust for the constant of lane width exponential sensitivity, this parameter was calibrated through a driving simulator experiment. In the experiment, 30 drivers with valid licenses were invited to drive in the simulator. All other road parameters were kept constant, only the lane width was changed. Simultaneously, the drivers' field skin response (GSR) was recorded to quantify their stress level. The GSR value was used as the dependent variable, and the lane width as the independent variable. An exponential function was fitted to obtain the exponential coefficient that best matched the experimental data. For example, through fitting the experimental data, the exponential coefficient was determined. The value is 2.5, which means that when the lane width decreases from the standard value, the driver's stress will increase at a faster exponential rate.

[0061] Calculation process:

[0062] A specific section of road in the city For example, the parameters obtained through the aforementioned steps are as follows:

[0063] Road curvature value ;

[0064] Road slope angle ;

[0065] Lane width values ;

[0066] Traffic sign density Units / km;

[0067] Traffic light density Units / km;

[0068] Number of intersections indivual;

[0069] Number of lane changes indivual;

[0070] Number of roundabouts indivual;

[0071] Road segment length ;

[0072] Substitute the pre-set weights and reference values:

[0073] , , , ;

[0074] , ;

[0075] , , ;

[0076] Units / km Units / km Units / km;

[0077] ;

[0078] The calculation process is as follows:

[0079] Calculate the geometric factor part :

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] ;

[0085] ;

[0086] Calculating information factors :

[0087] ;

[0088] ;

[0089] Calculate the structural complexity index :

[0090] ;

[0091] ;

[0092] Calculate the final navigation cognitive load index :

[0093] ;

[0094] ;

[0095] ;

[0096] The results indicate that the navigation cognitive load index for this road segment is 4.846. This value is a comprehensive, dimensionless index used to quantify the amount of "cognitive" resources required by the autonomous driving system to operate on this road segment. The system establishes a cognitive load level classification standard. For example, an index less than 1.5 indicates low load, meaning good road conditions, simple information, and easy system handling. An index between 1.5 and 3.5 indicates medium load, meaning there are certain driving challenges, and the system needs to allocate more computing resources for perception and decision-making. An index greater than 3.5 indicates high load, meaning complex road conditions, high risks, and the system needs to enter a state of high alert, and may need to adopt a conservative driving strategy. The current calculation result of 4.846 belongs to the high load level, indicating that this road segment is a significant challenge for the autonomous driving system.

[0097] The steps for obtaining dynamic event data with version identifiers are as follows:

[0098] Based on a spatiotemporal hierarchical geographic database and a road segment navigation cognitive load index, the event occurrence time, location coordinates, and event type in the received high dynamic layer data are analyzed. A unique timestamp based on the event occurrence time is assigned to each dynamic event, and a corresponding version identifier is added to the original dynamic event record to form dynamic event data with version identifier.

[0099] Specifically, based on a spatiotemporal hierarchical geographic database and a road segment navigation cognitive load index, the system automatically parses highly dynamic layer data streams in XML or JSON format received in real time from data sources such as traffic management centers and vehicle sensor networks. The system has a pre-set set of parsing rule templates that extract key fields for different data source formats, including the UTC timestamp of the event occurrence time, the WGS-84 latitude and longitude coordinates of the event location, and a predefined event type code. For example, "ACC01" represents a traffic accident, "CON02" represents moderate congestion, and "WRK01" represents road construction. The system immediately generates a globally unique timestamp accurate to nanoseconds for each successfully parsed dynamic event. This timestamp is generated by the Unix timestamp of the event occurrence time. The version identifier is formed by concatenating the timestamp (in seconds), the nanosecond portion of the current system time, and the unique identifier of the server node, for example, "1672531200.123456789-SVR05". Subsequently, the system generates a version identifier based on this unique timestamp and event content. The format of the version identifier is "unique timestamp-event type-content hash value". The content hash value is a SHA-256 hash digest of core information such as event type, location, and description, ensuring that even if the same type of event occurs at the same time, as long as there are slight differences in the content, the version identifier will never be the same. The system adds this generated unique timestamp and version identifier as new fields to the original, parsed dynamic event record to form dynamic event data with version identifier.

[0100] The steps to obtain multi-version map data snapshots are as follows:

[0101] Based on the event location coordinates of dynamic event data with version identifiers, each event is spatially matched with the spatial range coordinates of road segments in the spatiotemporal hierarchical geographic database. The road segment number affected by the dynamic event is determined according to the matching of spatial coordinates, forming a set of road segment factor parameters to be updated.

[0102] Based on the set of road segment factor parameters to be updated, the road segment navigation cognitive load index of the road segments affected by dynamic events is recalculated one by one. The recalculated road segment navigation cognitive load index is then merged with the dynamic event data with version identifiers to generate multi-version data records including dynamic events and road segment navigation cognitive load indices. Multi-version map data snapshots are formed using the event version identifier as an index.

[0103] Specifically, based on the event location coordinates of dynamic event data with version identifiers, the system initiates an efficient spatial matching process to accurately associate dynamic events with specific road segments. This process first utilizes the pre-built R-Tree spatial index in the spatiotemporal hierarchical geodatabase. For point-like events, such as traffic accidents, the system initiates a range query in the R-Tree, centered on the event's location coordinates. The query range is a circular area with a radius dynamically set according to the event type and road level. The radius is set based on ensuring coverage of the road itself and its adjacent areas. For example, for accidents on urban arterial roads, the query radius is set to 25 meters, while for highways, it is set to 50 meters. This initial query... The system quickly filters out all road segments whose minimum bounding rectangles (MBRs) intersect with the circular area as candidate segments. Then, the system performs precise geometric verification on each road segment in the candidate set, calculating the vertical projection distance from the event point to the centerline of the road segment. Only when this distance is less than half the width of the road segment plus a safety margin of 5 meters is the road segment confirmed as an affected road segment. For linear or area events, such as road construction or congestion, the system directly uses the geometric polygon of the event to perform an intersection query with the R-Tree. All road segments that have geometric overlap with the polygon are identified as affected road segments. The system collects the unique numbers of all confirmed affected road segments to form a set of road segment factor parameters to be updated.

[0104] Based on the set of road segment factor parameters to be updated, the system recalculates the road segment navigation cognitive load index for each road segment in the set. The core of this process lies in adjusting specific parameters in the road segment navigation cognitive load index calculation formula according to the type of dynamic event. The system has a built-in event impact mapping rule base, which was established by traffic engineering experts based on a large number of simulations and empirical studies. For example, when the event type is "real-time traffic accident", the rule base indicates the number of intersections on the road segment. Add 1 temporarily and adjust traffic sign density. A fixed value is added, for example, 5 lane widths per km, to simulate the additional attention and information processing demands of an accident scene. When the event type is "temporary road construction" and includes lane width information, the system directly updates the lane width value. This represents the actual width after construction, and the traffic sign density is updated based on the number of construction signs. If the event is "temporary road closure", the system skips complex calculations and directly sets the road segment navigation cognitive load index to a predefined maximum value, such as 9999.9, to indicate that it is not passable. After recalculating the road segment navigation cognitive load index for each affected road segment, the system merges the newly generated index value with the version-identified dynamic event data that triggered the update to create a new data record. This record contains the event version identifier, event details, affected road segment number, and the updated road segment navigation cognitive load index. Finally, the system stores this new record in a dedicated versioned data storage with the event version identifier as the primary key, forming a multi-version map data snapshot.

[0105] The steps to obtain a consistent map data view are as follows:

[0106] Based on multiple versions of map data snapshots, the map data requests sent by autonomous vehicle terminals are analyzed, and the request timestamps carried in each map data request are extracted one by one. Using the request timestamp as the reference retrieval time, the latest data version number that is less than or equal to the request timestamp in each version of data is determined, and the data versions corresponding to the static base layer, quasi-dynamic layer and high dynamic layer are retrieved respectively, forming a set of map data versions with timestamp matching.

[0107] Based on the set of map data versions matched by timestamps, spatial coordinate consistency is compared for each static base layer data version, quasi-dynamic layer data version, and high-dynamic layer data version. By unifying the road segment numbers, the road network topology, traffic rule information, road speed limit values, and dynamic event description information of the same road segments are merged one by one to form a consistent map data view with consistent matching of road elements and event information.

[0108] Specifically, based on multiple versions of map data snapshots, the system parses the map data requests sent by the autonomous vehicle terminal via HTTPS protocol. This request is a binary data stream in Protobuf format. The system first deserializes it into a structured object according to a predefined .proto file, extracting the core request parameters one by one. These parameters include the vehicle's current WGS-84 latitude and longitude coordinates, the requested map area (e.g., 5 kilometers ahead, 500 meters to the left and right), the road segment number the vehicle is currently traveling on, and the crucial request timestamp. This timestamp is the UTC time recorded by the vehicle terminal when initiating the request, synchronized via NTP service, accurate to milliseconds. The system uses this timestamp as the baseline retrieval time to perform time-slice queries on the data stored in the multiple versions of map data snapshots. For the static base layer and the quasi-dynamic layer, since updates are infrequent and the number of versions is relatively small, the system directly performs a binary search in their version history to quickly locate the latest version whose timestamp is less than or equal to the requested timestamp. For the highly dynamic layer, due to the large number of versions, the system uses a timestamp-based B+ tree index for efficient querying. The query condition is also "version timestamp less than or equal to the requested timestamp," and it returns the record with the largest timestamp among all version records that meet the condition. By performing the above version retrieval operation on these three layers of data respectively, the system obtains a list of version numbers containing the latest versions of the static base layer, the quasi-dynamic layer, and the highly dynamic layer. Then, based on these version numbers, the corresponding data entities are extracted from their respective storage areas to form a set of map data versions with matching timestamps.

[0109] Based on the map data version set matched with timestamps, the system initiates a data fusion and consistency verification process. The core of this process is to ensure that information retrieved from different data layers is spatially and logically fully aligned. First, the system uses the road segment list in the static base layer data version as a basic framework, traversing each road segment number. Then, for each road segment number, the system uses that number as a key to search in both the quasi-dynamic layer data version and the high-dynamic layer data version for associated traffic rule information, road speed limits, and dynamic event descriptions. During the merging process, the system performs strict spatial coordinate consistency comparisons. For example, the location coordinates of a dynamic event must precisely fall within the geometric centerline buffer of its associated road segment. The width of the collision zone is set to 1.5 times the width of the road segment. If the location of a dynamic event does not match the road segment geometry, for example, if an accident reported on a bridge has coordinates that deviate from the geometric range of the bridge segment by more than 10 meters, the system will trigger an anomaly flag, exclude the event from the current view, and log the data quality problem. After merging the attribute information, the system will aggregate all relevant information for the same road segment, including the road network topology obtained from the static base layer, traffic rule information and road speed limit values ​​obtained from the quasi-dynamic layer, and dynamic event description information obtained from the high-dynamic layer, into a unified data structure. This process is repeated for all road segments within the request range to form a consistent map data view that matches road features and event information.

[0110] The steps for obtaining the navigation path decision data packet are as follows:

[0111] Based on a consistent map data view, the road segment navigation cognitive load index is called for each corresponding road segment. Following a unified data encapsulation protocol, the road network topology, traffic rule information, road speed limit values, dynamic event description information, and road segment navigation cognitive load index corresponding to each road segment in the map data view are encapsulated to form a navigation path decision data package containing navigation path, road information, and cognitive load information.

[0112] Specifically, based on the consistent map data view, the system enters the data encapsulation stage, generating the final data packet for the autonomous vehicle terminal. The system iterates through each road segment record in the consistent map data view, and based on the road segment number, retrieves the latest version of the road segment navigation cognitive load index corresponding to the current state of that road segment (i.e., after incorporating the impact of high-dynamic events) from multiple version map data snapshots. This retrieval process is implemented through an efficient key-value lookup, where the key consists of the road segment number and the latest high-dynamic event version identifier. If the current road segment is not affected by any dynamic events, its baseline road segment navigation cognitive load index is directly retrieved. Subsequently, the system follows a predefined data encapsulation protocol based on Protocol Buffers, which specifies in detail the overall structure of the data packet and the details of each field. The system creates a new data packet instance, specifying the data type, name, and number. Following the protocol definition, it fills in each road segment's information line by line. Specifically, it encapsulates the road network topology (e.g., a list of connection node IDs), traffic rule information (e.g., turning restrictions, lane types), road speed limits, dynamic event descriptions (e.g., event type, severity, and impact range), and the calculated road segment navigation cognitive load index as attributes of a road segment object. This encapsulation process is repeated for all road segments within the entire request range, adding all road segment objects to a list. Finally, this list, along with the data packet header information (including timestamps, version numbers, and other metadata), is serialized into a binary byte stream, forming a navigation path decision data packet containing navigation path, road information, and cognitive load information.

[0113] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A cloud-based management method for autonomous driving map data, characterized in that, Includes the following steps: The map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. The road network and topology of the static base layer, as well as the traffic rules and road speed limits of the quasi-dynamic layer, are extracted and combined to form the basic attributes of each autonomous driving road segment, and a spatiotemporal hierarchical geographic database is established. Based on the spatiotemporal hierarchical geographic database, a multidimensional cognitive load factor set is established for each road segment in the autonomous driving road network. Based on the multidimensional cognitive load factor set, the road segment navigation cognitive load index is calculated. Based on the spatiotemporal hierarchical geographic database and the road segment navigation cognitive load index, when high dynamic layer data is received, a unique timestamp is assigned to the data, dynamic event data with version identifier is generated, and the road segment navigation cognitive load index is recalculated according to the road segment associated with the dynamic event data with version identifier. The dynamic event data with version identifier and the road segment navigation cognitive load index are stored as a new version, and a multi-version map data snapshot is established. Based on the multi-version map data snapshots, the system receives map data requests and request timestamps sent by the autonomous vehicle terminal. According to the request timestamps, it retrieves the latest static base layer, quasi-dynamic layer, and high-dynamic layer versions from the multi-version map data snapshots, combines them into a consistent map data view, and encapsulates the consistent map data view with the corresponding road segment navigation cognitive load index to generate a navigation path decision data package.

2. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the spatiotemporal hierarchical geographic database are as follows: The map data is divided into a static base layer, a quasi-dynamic layer, and a high-dynamic layer. Based on the spatial location information and road network geometric features of the map data, the static base layer data is subjected to spatial structure identification and road connection relationship analysis. The road network information and topology of each road in the static base layer are extracted one by one. The high-dynamic layer includes real-time traffic accidents, road congestion events, temporary road construction, and temporary road closures, forming a set of road network topology structures for the static base layer. Based on the static basic layer road network topology set, the quasi-dynamic layer data is matched and parsed segment by segment. The traffic rule information and road speed limit value corresponding to each road in the quasi-dynamic layer are extracted one by one. The road network topology, traffic rule information and road speed limit value are mapped and fused one by one to obtain the basic attributes of the autonomous driving road segment. Based on the basic attributes of the autonomous driving road segments, the roads are spatially organized according to their geometric coordinates. By establishing a corresponding index between the road spatial coordinates and attributes, and using the spatial coordinate range of the road as the index rectangle, corresponding R-Tree index nodes are generated one by one to complete the spatial organization and association of all road segments, forming a spatiotemporal hierarchical geographic database.

3. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the multidimensional cognitive load factor set are as follows: Based on the aforementioned spatiotemporal hierarchical geographic database, the road curvature value, road longitudinal slope angle, and lane lateral width value of each road segment in the autonomous driving road network are extracted one by one. The segments are identified by road segment number and their corresponding values ​​are recorded. By summarizing the three parameters and establishing a unified numbering table, a set of road geometric factor parameters is generated. Based on the set of road geometric factor parameters, the number of intersections, lane merging, and roundabouts included in each road segment is extracted. At the same time, the actual length of the corresponding road segment is calculated by combining the number of traffic signs and traffic lights. Each type of quantity is divided by the corresponding length to calculate the density per unit distance. Meanwhile, a reference standard value is introduced to make each density parameter dimensionless, resulting in a multidimensional cognitive load factor set.

4. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the navigation cognitive load index of the aforementioned road segment are as follows: Based on the multidimensional cognitive load factor set, the navigation cognitive load index of the road segment is calculated.

5. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the dynamic event data with version identifier are as follows: Based on the spatiotemporal hierarchical geographic database and the road segment navigation cognitive load index, the event occurrence time, location coordinates and event type in the received high dynamic layer data are analyzed. A unique timestamp based on the event occurrence time is assigned to each dynamic event, and a corresponding version identifier is added to the original dynamic event record to form dynamic event data with version identifier.

6. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the multi-version map data snapshot are as follows: Based on the event location coordinates of the dynamic event data with version identifiers, each event is spatially matched with the spatial range coordinates of road segments in the spatiotemporal hierarchical geographic database. The road segment number affected by the dynamic event is determined according to the matching of spatial coordinates, forming a set of road segment factor parameters to be updated. Based on the set of road segment factor parameters to be updated, the road segment navigation cognitive load index of the road segments affected by dynamic events is recalculated one by one. The recalculated road segment navigation cognitive load index is then merged with the dynamic event data with version identifiers to generate multi-version data records including dynamic events and road segment navigation cognitive load indices. Multi-version map data snapshots are formed using the event version identifier as an index.

7. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the consistent map data view are as follows: Based on the multi-version map data snapshots, the map data requests sent by the autonomous vehicle terminal are parsed, and the request timestamps carried in each map data request are extracted one by one. The request timestamps are used as the reference retrieval time to determine the latest data version number in each version of data that is less than or equal to the request timestamp. The data versions corresponding to the static base layer, quasi-dynamic layer and high dynamic layer are retrieved respectively to form a set of map data versions with timestamp matching. Based on the set of map data versions matched by the timestamp, spatial coordinate consistency is compared for each static base layer data version, quasi-dynamic layer data version, and high-dynamic layer data version. By unifying the road segment numbers, the road network topology, traffic rule information, road speed limit values, and dynamic event description information of the same road segments are merged one by one to form a consistent map data view with consistent matching of road elements and event information.

8. The cloud management method for autonomous driving map data according to claim 1, characterized in that, The steps for obtaining the navigation path decision data packet are as follows: Based on the consistent map data view, the road segment navigation cognitive load index of the corresponding road segment is called one by one. According to a unified data encapsulation protocol, the road network topology, traffic rule information, road speed limit value, dynamic event description information and road segment navigation cognitive load index of each road segment in the map data view are encapsulated one by one to form a navigation path decision data package containing navigation path, road information and cognitive load information.

Images