High-precision map-based local dynamic map implementation method

Abstract

A kind of local dynamic map implementation method based on high-precision map, first determine high-precision electronic map region to be processed, extract map information and construct local dynamic map static layer;Then consider the different characteristics and complex relationship of intelligent traffic participants, construct entity frame and knowledge graph world model respectively by LCM, store and update local dynamic map information using graph database;When the density of traffic participants is larger, based on 5GV2X, C-V2X technology realizes the LDM interaction between different traffic participants in a small range, realizes LDM construction, storage and update and local interaction.The application has the characteristics of high precision, storage is reasonable, can make full use of the multi-node nature of the local LDM information of each traffic participant, realizes LDM construction, storage and update, local interaction.

Description

Technical Field

[0001] This invention relates to a technology in the field of intelligent navigation, specifically a method for creating, storing, and updating interactive local dynamic maps based on centimeter-level maps. Background Technology

[0002] Local Dynamic Map (LDM) technology combines static digital maps (GIS maps) with various dynamic traffic information to form a comprehensive environmental perception description. It has become a key technology for integrating static, temporary, and dynamic information within a geographic environment. Its main function is to keep traffic conditions up-to-date and provide specific LDM information to specific stakeholders within a limited area. Currently, there is no fully mature solution for building, storing, and updating LDMs, and most LDMs are based on ordinary low-precision electronic maps and traditional relational databases, which cannot provide the accurate real-time information required for intelligent transportation (such as autonomous driving). Considering the problem of repeated perception of targets in intelligent transportation scenarios, it is necessary to select key perception nodes for data upload. Compared to uploading all data, the new selection method selects as few intelligent connected vehicle nodes as possible to reduce redundancy. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes a local dynamic map implementation method based on high-precision maps. This method features high precision and reasonable storage, and can fully utilize the multi-node nature of local LDM information of various traffic participants to achieve LDM construction, storage and updating, and local interaction.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a method for implementing local dynamic maps based on high-precision maps. First, the area to be processed in the high-precision electronic map is determined, map information is extracted, and a static layer of the local dynamic map is constructed. Then, considering the different characteristics and complex relationships of intelligent traffic participants, entity frames and knowledge graph world models are constructed respectively through LCM (Lightweight Communications and Marshalling), and the local dynamic map information is stored and updated using a graph database. When the density of traffic participants is high, LDM interaction between different traffic participants within a small area is realized based on 5G V2X and C-V2X technologies, realizing LDM construction, storage, updating, and local interaction.

[0006] This invention relates to a system for implementing the above-mentioned method, comprising: a static layer information extraction unit, an entity model construction and storage unit, and a node selection and information interaction unit, wherein: the static layer information extraction unit is connected to the entity model construction and storage unit and provides high-precision geographic information for LDM; the entity model construction and storage unit realizes real-time storage and processing of information; and the node selection and information interaction unit is a basic module for LDM construction, which is connected to the static layer information extraction unit and the entity model construction and storage unit respectively to cooperate in improving the construction efficiency and real-time performance of LDM.

[0007] Technical effect

[0008] This invention solves the technical problems of existing technologies, such as low positioning accuracy, limited ability to achieve lane-level information (e.g., roads, intersections) and inability to refine geographical environment, and inability to construct and store LDM information in complex traffic environments, such as road sections with high traffic density.

[0009] Compared to existing dynamic map construction methods based on ordinary electronic maps, centered on a central LDM system, and traditional relational databases, this approach introduces high-precision electronic maps, world model-based graph database storage, and adds a certain range of LDM interaction schemes. By constructing, storing, and updating interactive LDMs through this scheme, LDM accuracy can be significantly improved. Graph database storage and small-scale LDM interaction can reduce communication and data processing latency. Combined with LCM information interaction technology, dynamic and real-time sensing and updates can be achieved, improving system efficiency and saving resources. Attached Figure Description

[0010] Figure 1 This is a flowchart of the present invention;

[0011] Figure 2a and Figure 2b This is a diagram of the overall LDM architecture of the present invention;

[0012] Figure 3 This is a schematic diagram of LDM data layering according to the present invention;

[0013] Figure 4a This is a flowchart of the LDM high-precision map data processing of the present invention. Figure 4b This invention relates to the LDM high-precision map data extraction and storage process;

[0014] Figure 5a This is a flowchart of the LDM data access process. Figure 5b This is a diagram illustrating data storage entities and relationships. Figure 5c A knowledge graph-based world model entity relationship graph;

[0015] Figure 6a This is a flowchart illustrating the LDM (Local Traffic Management) interaction process among traffic participants on road sections with high traffic density, as described in this invention. Figure 6a This is a schematic diagram of the interaction. Detailed Implementation

[0016] In this embodiment, the intelligent transportation participants are universal, such as various vehicles equipped with various sensors, roadside equipment, various traffic infrastructures, network monitoring centers, and even pedestrian handheld devices. These can use this invention to construct, store, update, and interact with the LDM (Local Data Model) under a wireless network connection. Device terminals interact with information through mobile communication or various wireless communication devices, sending data collected by their various sensors to upper-layer data processors, such as roadside equipment and data centers, to construct the LDM. Simultaneously, they subscribe to real-time LDM messages through the LCM (Local Data Management System) and share LDM information within a certain range via short-range communication methods such as 5G V2X.

[0017] like Figure 2b As shown, this embodiment relates to a method for constructing, storing, updating, and locally interacting with an LDM (Local Descriptor Map) based on high-precision maps and graph database storage technology, including the following steps:

[0018] 1) Determine the area to be processed in the high-precision electronic map, extract map information, and construct a local dynamic map static layer, specifically including:

[0019] 1.1) Select the SHP (shapefile) file of a high-precision map of a local area, and use PostGIS to obtain the data information of each entity object on the map from the SHP file;

[0020] like Figure 4a As shown, the high-precision map data processing flow is divided into two directions: the first direction is information extraction and storage, and the second direction is pure map publishing for use by various display terminals.

[0021] 1.2) As Figure 5b As shown, after obtaining the SHP file, the static map entity data is stored in the graph database in the form of nodes as entity objects and edges as the relationships between entity objects. At the same time, the SHP file is synchronized to the map server for rendering and publishing.

[0022] The physical objects mentioned include, but are not limited to, lane lines.

[0023] The node attributes include, but are not limited to, color, position, and shape.

[0024] The edges mentioned include, but are not limited to, membership relationships and inclusion relationships.

[0025] The aforementioned release is intended for use by traffic participants, such as for display in traffic monitoring centers or on vehicle displays.

[0026] like Figure 4bAs shown, this is a schematic diagram of high-precision map information extraction and storage in the database. The storage stage also uses LCM-gen to construct entities. The SHP file is a high-precision map file of a certain intersection. The map information extraction tool is PostgreSQL, the electronic map publishing server is GeoServer, and the storage database is Neo4j. The entity construction is implemented using LCM programming. Nodes A, B, and C are three map data entities. A is the lane marking arrow, B is the lane obstacle, and C is the center point of the road intersection.

[0027] 2) As shown in Figure 5, considering the different characteristics and complex relationships of intelligent transportation participants, entity frames and knowledge graph world models are constructed respectively through LCM programming. A graph database is used to store and update local dynamic map information, specifically including:

[0028] 2.1) Monitor multi-process data in real time through LCM notification / subscription function, construct the frame structure of the corresponding entity based on the monitored traffic participant object information, and save the acquired information as an intermediate state.

[0029] The LCM is a collection of libraries and tools for messaging and data marshalling, which can be obtained by those skilled in the art through the following means: http: / / lcm-proj.github.io / .

[0030] The aforementioned multi-process data refers to the communication process data uploaded by traffic participants.

[0031] like Figure 1 As shown, the entity frame structure includes: the name, attributes, and status of the entity object, as well as the association relationships with other entities and a list of status records; furthermore, for different types of data entities, different names, attributes, statuses, association relationships, and status records may be included.

[0032] The attributes mentioned include, but are not limited to, features such as the shape, size, dimensions, and color of the entity;

[0033] The states mentioned include, but are not limited to, information such as the entity's position, velocity, acceleration, and direction of motion;

[0034] 2.2) Based on the static map entity data stored in the graph database in step 1.2), define and record the relationships between entities using pointers.

[0035] The relationships mentioned include, but are not limited to, hierarchical relationships, inclusion relationships, and distance relationships.

[0036] like Figure 5bThe diagram shows a schematic of graph database storage. A, B, and C are three node entities, representing a moving vehicle A, a moving vehicle B, and a working traffic light C, respectively. Each node has its own attributes and state, and edges represent the relationships between nodes. In the actual storage code, these relationships are represented by pointers.

[0037] 2.3) Generate and maintain a time-dimension-based list of event information for each entity, such as... Figure 1 The information shown is at time 0.

[0038] The event information list stores the corresponding status of the entity within a certain time range.

[0039] The event information list preferably further records future states predicted by a prediction algorithm. In this embodiment, a target state prediction algorithm based on time series is used, but not limited to this one.

[0040] 2.4) Construct a knowledge graph world model using LCM-gen, store entity information in a graph database, and access and maintain this information using LCM-gen. Specifically, use pointers to store nodes in a graph format, representing entities and relationships (i.e., connections between entities). This involves building a knowledge graph from the bottom-level nodes based on the relationships between them, while simultaneously implementing data layering. Figure 3 As shown, the first layer uses high-precision map information, while information for other layers is dynamically obtained from LCM-IPC.

[0041] Preferably, in the process of constructing a knowledge graph world model, when entity information is unclear, it is stored as a concept object or its information is inferred by calling a knowledge graph algorithm.

[0042] like Figure 5c The diagram illustrates an example of world model construction based on a knowledge graph, including data from each layer of the LDM (Local World Model). The lower layers of the knowledge graph serve as the basis for higher-level computation and reasoning. In the knowledge graph, the top-level entities are general world model objects, possessing a standard object library that stores typical objects based on various rules summarized from experience. The sub-entities of the top-level entities include static objects, semi-dynamic objects, and dynamic objects. These entities are objects identified by various sensor signals and acquired through the LCM, and then stored. They may be irregular and atypical, but they are objects that the current vehicle needs to focus on and track; they are actual existing entities, such as other vehicles, traffic lights, and pedestrians.

[0043] The aforementioned ambiguity in entity information refers to the following: when the perceived signal is significantly interfered with, there may be no corresponding external entity after processing. In this case, it is represented by a "conceptual object." An uncertainty model is used to represent the uncertainty in classification. Based on the relationships in the knowledge graph, algorithms can be triggered to merge, calculate, and infer the attributes of relevant entities and relationships. Furthermore, the attributes in entities and relationships can be modified based on the identification results, thereby achieving dynamic and real-time perception and knowledge updates.

[0044] The knowledge graph algorithm deduction mentioned refers to: reasoning based on traditional rules, reasoning based on distributed representation, reasoning based on neural networks, and a hybrid reasoning algorithm based on the above methods. The embodiment uses a modeling algorithm based on DKRL multi-source information fusion.

[0045] 2.5) By using LCM-IPC to obtain dynamic and semi-dynamic information from multi-process data, the associations between nodes are organized according to time, and the attributes in entities and relationships are modified, thereby realizing dynamic and real-time perception and knowledge updates, and completing the construction of LDM.

[0046] The dynamic and semi-dynamic information mentioned refers to, but is not limited to, vehicle information, traffic light information, pedestrian information, traffic conditions, and weather.

[0047] The time-based organization refers to the corresponding event information list generated and maintained for each entity according to the time dimension in step 2.3).

[0048] 2.6) Push the LDM constructed in step 2.5) to traffic participants via LCM in the form of subscription / notification.

[0049] 3) As shown in Figure 6, for traffic flow density K = N / L, where N is the number of vehicles and L is the road segment length, when K is greater than a set threshold, LDM interaction between different traffic participants within a small area is achieved based on 5G V2X and C-V2X technologies to reduce dynamic data redundancy and improve information update efficiency and real-time performance; otherwise, each intelligent traffic participant is set to directly upload / subscribe to data, specifically including:

[0050] like Figure 6a The diagram shown illustrates the LDM (Local Traffic Management) interaction process among traffic participants on road sections with high traffic flow density. Figure 6b The diagram shows an LDM interaction diagram, where A, B, C, and D are all traffic participant nodes (vehicles); the selected center nodes in the diagram are node B and node F.

[0051] 3.1) Based on 5G V2X and C-V2X communication technologies, a central node is selected from the traffic flow according to the geographical location of traffic participants to construct a small-scale communication network. Specifically, each vehicle first determines whether it is the head vehicle of the traffic flow based on its position and heading sensed by its onboard equipment. Figure 6b The lead vehicle is vehicle A. It sends its own position information and sensing range information to the vehicles behind it. The following vehicles determine if the distance between them is less than R, and select the vehicle furthest away as the first selected vehicle. Since vehicle B is the furthest away from vehicle A within range R, the first critical vehicle is within the lead vehicle's communication range. Figure 6b Vehicle B is selected as the first selected vehicle, and the central node uploads data. The availability of the central node is determined by whether the vehicle is equipped with V2X equipment and whether it is working properly. If it is not available, the next furthest vehicle is selected as the central node, and so on. Figure 6b If node B is unavailable, then node A must be selected as the central node. This process is repeated to select the next critical node, F.

[0052] The communication network refers to a circular area with a radius R equal to the communication distance of the on-board unit (OBU) or the roadside unit (RSU). Each participant in the communication network is a node. Taking node B as an example, at this moment, its communication coverage area includes three nodes: A, C, and E.

[0053] 3.2) The central node within the communication range uploads sensing information and subscribes to LDM information. Other nodes within the communication range of the central node upload sensing information and subscribe to the LDM information generated in steps 1) and 2) to the central node, thereby realizing real-time and fast LDM transmission, sharing and updating.

[0054] Taking node B as an example, the central node communicates bidirectionally with the data center directly through roadside equipment, wireless base stations and networks to access data center LDM resources. Alternatively, taking the coverage area of ​​node B as an example, nodes A, C and E directly interact bidirectionally with node B to achieve real-time LDM updates.

[0055] The LDM transmission includes: uplink, which is the perception information transmitted from other nodes to the central node, and the central node transmitting its own and other nodes' perception information to the LDM; downlink, which is the LDM transmission to other nodes of the central node and the LDM information subscribed to by the central node, and the central node transmitting the LDM information subscribed to by other nodes of other nodes.

[0056] The update refers to the LDM updating the data in the graph database based on the data uploaded by the central node.

[0057] 3.3) When a new traffic participant enters the communication network, the traffic participant is added to the communication network and the shared LDM in the network is updated and uploaded to the roadside unit and data center.

[0058] Taking the coverage area of ​​node B as an example, determine whether there are any new nodes joining the communication range of this node. In the figure, node D is the node that will be added soon. Add this node to the node network, establish communication with the central node, update LDM and synchronize it to the data center.

[0059] The aforementioned update of the shared LDM in the network refers to: modifying / updating the entities in the LDM and their association information with other entities.

[0060] The roadside units mentioned include, but are not limited to, RSUs, APs, LTE-V, cameras, etc.

[0061] The data center refers to a cloud server used for generating, maintaining, and publishing LDMs.

[0062] 3.4) When an existing traffic participant leaves the communication network, the traffic participant is removed from the communication network and the shared LDM in the network is updated and uploaded to the roadside unit and control center.

[0063] Taking the coverage area of ​​node B as an example, determine whether any nodes have left the communication range of this node. In the figure, node E is a node that is about to be added. Remove this node from the network of this node, disconnect the communication of the central node, update LDM and synchronize it to the data center.

[0064] Preferably, the node density is periodically determined to be below the disbanding threshold, and the network is disbanded when the condition is met. The nodes can directly communicate bidirectionally with the data center through roadside equipment, wireless base stations and the network to access the data center's LDM resources.

[0065] As shown in Figure 2, the data center obtains traffic participant information from the outside through the LCM and performs storage and retrieval operations on the graph database. External devices also subscribe to LDM information through the LCM.

[0066] Based on specific practical experiments, the simulation platform was MATLAB+PRESCAN+Neo4j, while the comparison scheme used MATLAB+PRESCAN+MySQL. The simulation process included:

[0067] A. Scene Parameter Settings

[0068] Table 1 Scene Parameter Settings Table

[0069] parameter numerical values Intelligent connected vehicle data 86 Number of ordinary vehicles 14 Number of pedestrians 6 Number of lanes 8*4 lane width 3.5m Number of vehicle-mounted V2X devices 86 Vehicle radar number 86 V2X, radar sensing radius R 20m Number of roadside cameras 4 Number of traffic lights 4*2 Traffic flow density threshold 40 units / km

[0070] B. Map

[0071] A high-precision map shapefile of a certain intersection (accuracy: error less than 10cm). The comparison scheme uses a regular electronic map of the same intersection, OpenStreetMap.

[0072] C. Test Results

[0073] Table 2 Test Results

[0074]

[0075] In summary, the significant advancements of this invention compared to the prior art include:

[0076] 1. Provides more accurate underlying static map information. Using high-precision maps as static map information, the accuracy can reach the centimeter level. Moreover, the data of various elements in high-precision maps are more detailed, and the location and shape information is more accurate, which is conducive to providing lane-level navigation information.

[0077] 2. Adapting to the characteristics of various data in intelligent transportation and the complex relationships between traffic participants, this approach abandons traditional relational databases and adopts a more realistic graph database for LDM storage, facilitating expansion and maintenance. By introducing a world model and constructing entity frames, the relationships between data are stored in nodes and edges, eliminating the concepts of database tables and fields. If the relationships between data are complex, and data exists in multiple tables with intermediate tables, traditional databases require various join operations to query data, resulting in complex SQL statements that are difficult to maintain and have low performance. In contrast, graph databases can achieve query and modification (entity attribute, state, pointer modification) functions with simple statements, and the execution speed is much faster, ensuring the real-time performance of LDM.

[0078] 3. Organizing the relationships between nodes by time can depict the changes in data over time in real time while significantly reducing data redundancy: as shown in the test results, the additional overhead ratio is as follows. The data redundancy is directly reduced by constructing the nodes by selecting nodes.

[0079] 4. Scalability. Data obtained through LCM can be used to trigger various algorithms for merging, calculation, and inference based on relationships within the knowledge graph, thereby acquiring the attributes of relevant entities and relationships. This solves the problem of inaccurate sensor entity data obtained through LCM.

[0080] 5. Fully utilize information exchange among traffic participants to ensure the real-time nature and effectiveness of LDM information. By constructing local communication networks in road sections with complex traffic conditions, LDM sharing can be achieved within a small area. Participants do not need to request LDM from the intelligent center, thereby reducing the pressure on the central LDM, improving efficiency, reducing the risk of high concurrency latency, and saving resources.

[0081] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A high-precision map-based local dynamic map implementation method, characterized in that, Firstly, the high-precision electronic map to be processed is determined, the map information is extracted, and the local dynamic map static layer is constructed; then, considering the different characteristics and complex relationships of intelligent traffic participants, the entity frame and knowledge graph world model are respectively constructed through LCM, and the local dynamic map information is stored and updated using a graph database; when the traffic participant density is large, the LDM interaction between different traffic participants in a small range is realized based on 5G V2X and C-V2X technology, and the LDM construction, storage, updating and local interaction are realized; The local dynamic map static layer is constructed by selecting the SHP file of the high-precision map of the local area, using Post GIS to obtain the data information of each entity object of the map from the SHP file, taking the node as the entity object, and using LCM-gen to construct the entity in the form of the relationship between the entity objects and storing the static map entity data into the graph database; The entity frame includes the name, attribute and state of the entity object, and the association relationship with other entities and the state record list; The entity frame and the knowledge graph world model are respectively constructed through LCM, specifically including: real-time monitoring of multi-process data through the LCM notification / subscription function, constructing the frame structure of the corresponding entity according to the monitored object information of each traffic participant, and saving the obtained information as the intermediate state; defining and recording the association relationship between entities through a pointer according to the static map entity data stored in the graph database; generating and maintaining an event information list of a time dimension for each entity event information; constructing the knowledge graph world model through LCM-gen, storing the entity information into the graph database, and accessing and maintaining the entity information through LCM-gen; obtaining dynamic and semi-dynamic information from multi-process data through LCM-IPC, organizing the association between nodes according to time, modifying the attributes in the entity and the relationship, thereby realizing dynamic and real-time perception and knowledge update, and completing the construction of the LDM.

2. The high-precision map-based local dynamic map implementation method according to claim 1, characterized in that, When the traffic flow density is greater than a set threshold, the LDM interaction between different traffic participants in a small range is realized based on 5G V2X and C-V2X technology, specifically: each vehicle judges whether the vehicle is the head vehicle of the traffic flow through the position and heading perceived by the vehicle-mounted device, the head vehicle sends the position information and perception range information of the vehicle to the rear vehicle, the rear vehicle judges whether the distance between the head vehicle and the rear vehicle is less than the communication distance range, and selects the vehicle farthest from the head vehicle, i.e. the vehicle closest to the communication distance range, as the first selected vehicle, the first selected vehicle is used as the center node for data uploading, the center node uploads the perception information and subscribes to the LDM information, and other nodes in the communication range of the center node upload the perception information and subscribe to the LDM information of the center node, thereby realizing real-time and rapid LDM transmission sharing and updating.

3. The high-precision map-based local dynamic map implementation method according to claim 2, characterized in that, Whether the first selected vehicle is equipped with a V2X device and whether the V2X device is working normally are used to judge whether the center node is available, and when the center node is not available, the vehicle farthest from the center node is further selected as the center node.

4. The high-precision map-based local dynamic map implementation method according to claim 1, characterized in that, The LDM constructed is pushed to the traffic participants in the form of subscription / notification through LCM.

5. The high-precision map-based local dynamic map implementation method according to claim 2, characterized in that, The updating refers to: when a new traffic participant enters the communication network, the traffic participant is added to the communication network and the LDM shared in the network is updated and uploaded to the roadside unit and the data center; when an existing traffic participant leaves the communication network, the traffic participant is deleted from the communication network and the LDM shared in the network is updated and uploaded to the roadside unit and the control center.

6. The high-precision map-based local dynamic map implementation method according to claim 2, characterized in that, Periodically judge whether the node density is lower than the dissolving threshold, and dissolve the network when the condition is met, the node directly communicates with the data center through the roadside device, the wireless base station and the network, and accesses the LDM resource of the data center.

7. A high-definition map based local dynamic map system implementing the method of any one of claims 1-6, characterized in that, Comprise: The static layer information extraction unit, the entity model construction storage unit and the node selection and information interaction unit, wherein: the static layer information extraction unit is connected with the entity model construction storage unit and provides high-precision geographic information for the LDM, the entity model construction storage unit realizes real-time storage and processing of information, the node selection and information interaction unit constructs the basic module for the LDM, and is connected with the static layer information extraction unit and the entity model construction storage unit respectively to cooperate to improve the construction efficiency and real-time performance of the LDM.

Images