A digital road network cloud service platform architecture and system

Abstract

The application discloses a digital road network cloud service platform architecture and system, and belongs to the technical field of intelligent transportation. In order to solve the problem of improving the service range and service level of intelligent roads. The application provides distributed traffic big data processing services based on a multi-source heterogeneous data fusion engine based on space-time joint indexing. The multi-source heterogeneous data fusion engine based on space-time joint indexing realizes the global data integration of human, vehicle and road full-element physical devices. The heterogeneous data sources of real-time access vehicle-mounted sensors, roadside units and mobile terminals are accurately positioned and queried based on indexing. A multi-modal geographic space data intelligent fusion engine realizes the centimeter-level space registration and dynamic rendering of high-precision maps, high-precision maps, three-dimensional real scenes and three-dimensional model data, forms a computable digital twin base of virtual-real mapping, and realizes the millimeter-level simulation and intelligent decision verification of the urban traffic system based on the multi-modal geographic space data intelligent fusion engine.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, specifically relating to a digital road network cloud service platform architecture and system. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) technology and the autonomous driving industry, urban traffic flow density continues to rise, and the demand for multi-modal transportation coordination is becoming increasingly complex. Traffic management faces prominent problems such as data silos and delayed responses. Currently, the level of digitalization of traditional transportation infrastructure varies greatly, massive amounts of heterogeneous data lack effective integration, and the dynamic scheduling capability of road network resources is insufficient, making it difficult to meet the needs of real-time traffic situation awareness, accurate prediction, and collaborative management.

[0003] A patent application with publication number CN117671932A, entitled "A Method for Providing Traffic Services and a Cloud Service System," discloses a method for providing traffic services and a cloud service system. The method includes: acquiring a simulation task input or selected by a user on a cloud management platform, wherein the simulation task indicates a target road segment and a simulation requirement for traffic control on the target road segment; performing a simulation operation to control traffic on the target road segment according to the simulation task and at least one alternative control strategy corresponding to the target road segment, obtaining at least one simulation result, wherein the alternative control strategies characterize the traffic control on the target road segment; and selecting a target control strategy from the at least one alternative control strategy based on the at least one simulation result, wherein the target control strategy is used to implement traffic control on the target road segment. This method helps to adopt reasonable and effective traffic control strategies and improve road traffic efficiency. However, the accuracy of the simulation model in this technology is highly dependent on the quality and update frequency of real-time traffic data, and it lacks digital twin extrapolation capabilities. If data acquisition is delayed or there is noise (such as sensor failure or communication interruption), the simulation results may deviate from reality, thus affecting the effectiveness of strategy selection. Especially in sudden traffic incidents, the system may not be able to respond to dynamic changes in a timely manner. Moreover, the generation of alternative strategies and the configuration of simulation tasks in this method depend on user input and lack a traffic scenario library. If the user has insufficient understanding of traffic management (such as unreasonable weight allocation or ambiguous scenario definition), it may lead to the strategy selection deviating from the optimal solution. Summary of the Invention

[0004] The problem this invention aims to solve is to improve the service scope and service level of smart roads, and proposes a digital road network cloud service platform architecture and system.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A digital road network cloud service platform architecture includes a multi-source heterogeneous data fusion engine based on spatiotemporal joint index, a multimodal geospatial data intelligent fusion engine, and a high-precision simulation engine. The multi-source heterogeneous data fusion engine based on spatiotemporal joint index is connected to the multimodal geospatial data intelligent fusion engine, and the multimodal geospatial data intelligent fusion engine is connected to the high-precision simulation engine.

[0007] The multi-source heterogeneous data fusion engine based on spatiotemporal joint index is used to provide distributed traffic big data processing services. It integrates the full-domain data of physical devices of people, vehicles and roads through the multi-source heterogeneous data fusion engine based on spatiotemporal joint index, and obtains multi-source heterogeneous data based on spatiotemporal joint index. It can accurately locate and query heterogeneous data sources that are accessed in real time from vehicle sensors, roadside units and mobile terminals based on the index.

[0008] The multimodal geospatial data intelligent fusion engine, based on a spatiotemporal joint index multi-source heterogeneous data fusion engine, achieves centimeter-level spatial registration and dynamic rendering of standard precision maps, high precision maps, 3D real scenes and 3D model data, forming a computable digital twin base for virtual-real mapping;

[0009] The high-precision simulation engine is based on a multimodal geospatial data intelligent fusion engine. By constructing a three-dimensional technology system of behavior modeling, scene construction, and virtual-real feedback, it achieves millimeter-level simulation and intelligent decision verification of urban transportation systems.

[0010] Furthermore, the heterogeneous data from vehicle-mounted sensors, roadside units, and mobile terminals specifically includes the following data:

[0011] Vehicle-mounted sensor data includes vehicle GPS data, camera data, V2X communication data, and radar data; roadside unit data includes traffic flow monitoring data, road condition detection data, meteorological data, and traffic signal data; mobile terminal data includes GPS positioning data, travel behavior data, crowdsourced road condition data, and user profile data.

[0012] Furthermore, the method for constructing the multi-source heterogeneous data fusion engine based on spatiotemporal joint index includes the following steps:

[0013] Step 1. Mathematically define the composite code of traffic elements to obtain the traffic entity set E, where each entity in the traffic entity set... Having spatial coordinates timestamp and category identifier The spatiotemporal attribute triples are then encoded to obtain the industry coding function, spatial coding function, and temporal coding function.

[0014] Then, a composite coding function with industry coding function, spatial coding function, and time coding function is defined. The expression is:

[0015]

[0016] in, For industry-specific coding functions, For spatial coding functions, For time coding functions, This is a bit string concatenation operation;

[0017] Step 2. Based on the spatial encoding function and temporal encoding function obtained in Step S1, design a spatiotemporal proximity query method, and perform joint filtering based on a combination of spatial and temporal filtering to obtain the expression:

[0018]

[0019] in, For the candidate set of traffic elements, The encoding range for spatial filtering. The encoding range for time filtering;

[0020] Step 3. For the candidate set of traffic elements obtained in Step 2, design a multi-level index structure. The first-level index is a B+ tree index based on the element code prefix, the second-level index is an R-tree index based on the spatiotemporal range, and the third-level index is an inverted index based on the semantic category of traffic elements. Define the index query complexity. ,in, For the total amount of all data, The total amount of data for the candidate set of traffic elements;

[0021] Step 4. Combining the multi-level index structure obtained in Step 3, a sharding function based on spatiotemporal correlation is used for distributed storage of traffic elements. The expression is:

[0022]

[0023] in, For a piecewise function based on spatiotemporal correlation, For the number of fragments, The MurmurHash3 algorithm is used; then, the node load of traffic elements is balanced based on consistent hashing, and the construction of a multi-source heterogeneous data fusion engine based on spatiotemporal joint index is completed, resulting in multi-source heterogeneous data based on spatiotemporal joint index.

[0024] Furthermore, the method for constructing the multimodal geospatial data intelligent fusion engine includes the following steps:

[0025] Step 1: Perform data fusion on the obtained multi-source heterogeneous data based on spatiotemporal joint index, and establish coordinate system standardization and traffic element semantic structure standardization extraction for data features;

[0026] Coordinate system standardization is based on the GDAL library to build a coordinate system standardization process, and traffic element semantic structure standardization extraction is based on a deep learning model to build a topology graph.

[0027] Step 2: Perform data visualization rendering to obtain visualized traffic spatial elements, and define the semantic model of the traffic spatial elements as follows:

[0028]

[0029] in, This represents the semantic model of the i-th traffic spatial element. Represents the geometric description of the i-th traffic spatial element. This represents the topological connection of the i-th traffic spatial element. Represents the rule constraints for the i-th traffic spatial element. Represents the semantic attribute of the i-th traffic spatial element;

[0030] Step 3: Perform computable spatial co-verification on the obtained visualized traffic elements, verifying them from four aspects: topological consistency, geometric consistency, temporal consistency, and semantic consistency.

[0031] The method for topological consistency verification is to check any spatial edge element. At least two valid topology nodes must be connected; otherwise, it is considered a topology break.

[0032] The method for geometric consistency verification is to use spatial Boolean relations to determine whether the layout relationship of traffic space elements in geometric space has overlapping, intersecting, or crossing errors.

[0033] The method for verifying temporal consistency is to determine whether traffic spatial elements are consistent in time and spatial sequence.

[0034] The method for semantic consistency verification is to determine the rationality of the business logic or rules between traffic space elements and avoid semantic conflicts or contradictions.

[0035] Furthermore, the method for constructing the high-precision simulation engine includes the following steps:

[0036] S1. Conduct multi-granularity traffic behavior modeling, including micro-level traffic behavior modeling and macro-level traffic behavior modeling;

[0037] The micro-traffic behavior model uses joint modeling of vehicles and drivers to describe the motion state of vehicles, expressed as:

[0038]

[0039] in, This represents the IDM driver model. This represents a vehicle lane-changing model. It is the vehicle speed at time t. It is the acceleration at time t. It is the speed difference between the vehicle and the vehicle in front at time t. It is the distance to the car in front at time t. This represents the direction control angle at time t. It is the current position at time t. These are the driver's behavioral preference parameters at time t;

[0040] The macro-level traffic behavior model is based on a continuous model of traffic flow, and is described as follows:

[0041]

[0042] in, It's traffic density. It is a flow function, where x is the location of the traffic flow;

[0043] Using the LWR model for the flow function, the expression is:

[0044]

[0045] in, It's traffic density. It is the free-flow velocity; It is the saturation density;

[0046] S2. Traffic scene generation;

[0047] The generation of the scene library is divided into two stages: scene extraction and scene generation. The scene extraction stage is achieved by combining template matching with a temporal behavior recognition network.

[0048] In the scene extraction phase, for typical scene behavior sequences, an event detector is constructed, expressed as:

[0049]

[0050] in, for Vehicle location at any given time for The feature parameters at each time point are calculated, and the category identifier is output. , It is a long short-term memory network;

[0051]

[0052] in, For normalized exponential functions, and These are the first model parameters and the second model parameters, respectively. It is a parameter function that represents the maximum value;

[0053] The scene generation stage uses a method combining finite structures with infinite variables to combine intersections, main and auxiliary roads, and ramps of a finite road topology with infinite traffic flow variables to generate simulation scenes. The resulting expression is:

[0054]

[0055] in, It is a road network topology template; It is a set of vehicle behavior strategies; It is an environment variable;

[0056] Then, combined with the Wasserstein GAN network, a realistic traffic operation scenario is generated. The generator adopts the Transformer architecture, and the recursive formula for generating time-series traffic flow yields the state at time t+1 as follows:

[0057] in Let t be the state at time t. Let be random noise at time t. This refers to the conditional information at time t;

[0058] S3. Constructing a fully networked, real-world intelligent closed-loop decision-making system:

[0059] The bidirectional interactive channel is composed of a virtual-real feedback mapping system. The real-to-virtual mapping uses sensors to acquire the real-time state of traffic participants and reconstruct the simulation environment, resulting in:

[0060]

[0061] in, It is a digital transportation element. It is a virtualized transportation pipeline. It is a physical transportation element;

[0062] Then, the virtual-to-real feedback is tested and optimized in simulation before being deployed to the actual control system, resulting in:

[0063]

[0064] in, It is an analog control strategy. It is a real control strategy. It is a platform for deploying and distributing tasks;

[0065] Then, based on reinforcement learning algorithms, the objective function is optimized to obtain the optimal policy. ,in For strategy The utility value is obtained as follows:

[0066]

[0067] in, It is the mathematical expectation of the strategy; It is a strategy The utility value.

[0068] A digital road network cloud service platform system includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the digital road network cloud service platform architecture as described above.

[0069] The beneficial effects of this invention are:

[0070] The digital road network cloud service platform architecture described in this invention improves data integration and processing efficiency: through a self-developed heterogeneous data fusion engine, it realizes full-domain data integration of all physical devices of people, vehicles and roads, and has the ability to access multiple heterogeneous data sources in real time, with a daily processing capacity of up to PB level.

[0071] The digital road network cloud service platform architecture described in this invention enhances the training of intelligent driving algorithms and the verification of traffic control strategies: by constructing a large-scale, high-quality traffic scenario library, it provides rich data support for the training of intelligent driving algorithms and the verification of traffic control strategies.

[0072] The digital road network cloud service platform architecture described in this invention enables precise modeling and simulation of intelligent transportation systems: relying on a multimodal geospatial data intelligent fusion engine, it achieves centimeter-level spatial registration and dynamic rendering of standard-precision maps, high-precision maps, 3D real scenes and 3D model data, forming a computable digital twin base that maps the virtual and the real. Attached Figure Description

[0073] Figure 1 This is a schematic diagram of the architecture of a digital road network cloud service platform according to the present invention. Detailed Implementation

[0074] 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 specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0075] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0076] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 Detailed explanation is as follows:

[0077] Example 1:

[0078] A digital road network cloud service platform architecture includes a multi-source heterogeneous data fusion engine based on spatiotemporal joint index, a multimodal geospatial data intelligent fusion engine, and a high-precision simulation engine. The multi-source heterogeneous data fusion engine based on spatiotemporal joint index is connected to the multimodal geospatial data intelligent fusion engine, and the multimodal geospatial data intelligent fusion engine is connected to the high-precision simulation engine.

[0079] The multi-source heterogeneous data fusion engine based on spatiotemporal joint index is used to provide distributed traffic big data processing services. It integrates the full-domain data of physical devices of people, vehicles and roads through the multi-source heterogeneous data fusion engine based on spatiotemporal joint index, and obtains multi-source heterogeneous data based on spatiotemporal joint index. It can accurately locate and query heterogeneous data sources that are accessed in real time from vehicle sensors, roadside units and mobile terminals based on the index.

[0080] The multimodal geospatial data intelligent fusion engine, based on a spatiotemporal joint index multi-source heterogeneous data fusion engine, achieves centimeter-level spatial registration and dynamic rendering of standard precision maps, high precision maps, 3D real scenes and 3D model data, forming a computable digital twin base for virtual-real mapping;

[0081] The high-precision simulation engine is based on a multimodal geospatial data intelligent fusion engine. By constructing a three-dimensional technology system of behavior modeling, scene construction, and virtual-real feedback, it achieves millimeter-level simulation and intelligent decision verification of urban transportation systems.

[0082] Furthermore, the heterogeneous data from vehicle-mounted sensors, roadside units, and mobile terminals specifically includes the following data:

[0083] Vehicle-mounted sensor data includes vehicle GPS data, camera data, V2X communication data, and radar data; roadside unit data includes traffic flow monitoring data, road condition detection data, meteorological data, and traffic signal data; mobile terminal data includes GPS positioning data, travel behavior data, crowdsourced road condition data, and user profile data.

[0084] Furthermore, the method for constructing the multi-source heterogeneous data fusion engine based on spatiotemporal joint index includes the following steps:

[0085] Step 1. Mathematically define the composite code of traffic elements to obtain the traffic entity set E, where each entity in the traffic entity set... Having spatial coordinates timestamp and category identifier The spatiotemporal attribute triples are then encoded to obtain the industry coding function, spatial coding function, and temporal coding function.

[0086] Then, a composite coding function with industry coding function, spatial coding function, and time coding function is defined. The expression is:

[0087]

[0088] in, For industry-specific coding functions, For spatial coding functions, For time coding functions, This is a bit string concatenation operation;

[0089] Industry coding functions ;

[0090] Spatial encoding function (complete H3 encoding) ;

[0091] Time encoding function: ;

[0092] Step 2. Based on the spatial encoding function and temporal encoding function obtained in Step S1, design a spatiotemporal proximity query method, and perform joint filtering based on a combination of spatial and temporal filtering to obtain the expression:

[0093]

[0094] in, For the candidate set of traffic elements, The encoding range for spatial filtering. The encoding range for time filtering;

[0095] Furthermore, the spatial filtering stage calculates the query region. Covered H3 mesh set: The time filtering stage is the calculation time range. Corresponding encoding range: ,in, It is any one of the H3 grids. It's the start time. It is the end time;

[0096] Step 3. For the candidate set of traffic elements obtained in Step 2, design a multi-level index structure. The first-level index is a B+ tree index based on the element code prefix, the second-level index is an R-tree index based on the spatiotemporal range, and the third-level index is an inverted index based on the semantic category of traffic elements. Define the index query complexity. ,in, For the total amount of all data, The total amount of data for the candidate set of traffic elements;

[0097] Step 4. Combining the multi-level index structure obtained in Step 3, a sharding function based on spatiotemporal correlation is used for distributed storage of traffic elements. The expression is:

[0098]

[0099] in, For a piecewise function based on spatiotemporal correlation, For the number of fragments, The MurmurHash3 algorithm is used; then, the node load of traffic elements is balanced based on consistent hashing, and the construction of a multi-source heterogeneous data fusion engine based on spatiotemporal joint index is completed, resulting in multi-source heterogeneous data based on spatiotemporal joint index.

[0100] Furthermore, a load balancing strategy is constructed based on consistent hashing, and the node load calculation formula is as follows:

[0101] in These are the first weight parameter and the second weight coefficient, respectively. For data volume, This section provides standardized, low-latency multi-source heterogeneous traffic data fusion capabilities for subsequent applications, and provides a data foundation for the subsequent digital twin platform.

[0102] Furthermore, the method for constructing the multimodal geospatial data intelligent fusion engine includes the following steps:

[0103] Step 1: Perform data fusion on the obtained multi-source heterogeneous data based on spatiotemporal joint index, and establish coordinate system standardization and traffic element semantic structure standardization extraction for data features;

[0104] Coordinate system standardization is based on the GDAL library to build a coordinate system standardization process, and traffic element semantic structure standardization extraction is based on a deep learning model to build a topology graph.

[0105] Furthermore, coordinate system standardization is based on the GDAL library, and the coordinate system standardization process involves setting up a set of traffic entities. Coordinate systems from WGS-84, CGCS2000, UTM, etc., are uniformly converted to the target coordinate system. Let the coordinates of the set of traffic entities be... The conversion method is as follows:

[0106]

[0107] in, To unify coordinate systems, For data source coordinates, It is a three-axis rotation matrix. These are roll angle, pitch angle, and yaw angle, respectively. It is a translation vector;

[0108] Traffic element semantic structure standardization extraction is based on a deep learning model to construct a topology graph, given a set of lane points. , construct graph edge set This indicates the topological connections between lanes (predecessor, successor, left turn, right turn, etc.).

[0109] Step 2: Perform data visualization rendering to obtain visualized traffic spatial elements, and define the semantic model of the traffic spatial elements as follows:

[0110]

[0111] in, This represents the semantic model of the i-th traffic spatial element. Represents the geometric description of the i-th traffic spatial element. This represents the topological connection of the i-th traffic spatial element. Represents the rule constraints for the i-th traffic spatial element. Represents the semantic attribute of the i-th traffic spatial element;

[0112] Traffic elements are visualized using the real-time high-fidelity Unreal Engine and the WebGL-based Cesium client-side 3D rendering engine. Road segment-level networks are loaded and rendered via web map services and web map tile services. The web map service generates maps using geospatial data and can return map imagery in different formats, such as raster formats like PNG and JPEG, or vector formats like SVG and WEBCG.

[0113] Step 3: Perform computable spatial co-verification on the obtained visualized traffic elements, verifying them from four aspects: topological consistency, geometric consistency, temporal consistency, and semantic consistency.

[0114] The method for topological consistency verification is to check any spatial edge element. At least two valid topological nodes must be connected; otherwise, it is considered a topological break. Topological consistency ensures that all elements are meaningfully connected in the network topology, preventing the existence of islands, dangling lines, or breaks. Construct a spatial topology graph. , where nodes Represents physical or logical connection points (intersections, lane ends, junctions, etc.) edges This represents spatial geometric objects such as road segments, lanes, and pipeline segments. For any spatial edge element... At least two valid topological nodes must be connected; otherwise, it is considered a topological break, represented as:

[0115] ;

[0116] The method for geometric consistency verification is to use spatial Boolean relations to determine whether the layout relationship of traffic space elements in geometric space overlaps, intersects, or crosses; two elements The corresponding spatial geometry is ,but:

[0117]

[0118] in, This represents the minimum tolerance error.

[0119] The method for verifying temporal consistency is to determine whether traffic spatial elements maintain consistency in both temporal and spatial sequence; (object) Different time versions Encoded as It must meet the following requirements:

[0120]

[0121] in, This represents a measure of geometric or attribute change. This is the threshold for tolerable changes.

[0122] The method for semantic consistency verification is to determine the rationality of the business logic or rules between traffic space elements, avoiding semantic conflicts or contradictions. For example, if a semantic rule states: "If a lane is a bus lane, it must be a two-way lane or have a right-turn access lane," then the definition is:

[0123]

[0124] Furthermore, verification is performed using a set of first-order logic rules. This part outputs a digital twin base with centimeter-level precision for virtual-to-real mapping, providing a high-fidelity operating environment for the simulation engine.

[0125] Furthermore, the method for constructing the high-precision simulation engine includes the following steps:

[0126] S1. Conduct multi-granularity traffic behavior modeling, including micro-level traffic behavior modeling and macro-level traffic behavior modeling;

[0127] The micro-traffic behavior model uses joint modeling of vehicles and drivers to describe the motion state of vehicles, expressed as:

[0128]

[0129] in, This represents the IDM driver model. This represents a vehicle lane-changing model. It is the vehicle speed at time t. It is the acceleration at time t. It is the speed difference between the vehicle and the vehicle in front at time t. It is the distance to the car in front at time t. This represents the direction control angle at time t. It is the current position at time t. These are the driver's behavioral preference parameters at time t;

[0130] The macro-level traffic behavior model is based on a continuous model of traffic flow, and is described as follows:

[0131]

[0132] in, It's traffic density. It is a flow function, where x is the location of the traffic flow;

[0133] Using the LWR model for the flow function, the expression is:

[0134]

[0135] in, It's traffic density. It is the free-flow velocity; It is the saturation density;

[0136] S2. Traffic scene generation;

[0137] The generation of the scene library is divided into two stages: scene extraction and scene generation. The scene extraction stage is achieved by combining template matching with a temporal behavior recognition network.

[0138] In the scene extraction phase, for typical scene behavior sequences, an event detector is constructed, expressed as:

[0139]

[0140] in, for Vehicle location at any given time for The feature parameters at each time point are calculated, and the category identifier is output. , It is a long short-term memory network;

[0141]

[0142] in, For normalized exponential functions, and These are the first model parameters and the second model parameters, respectively. It is a parameter function that represents the maximum value;

[0143] The scene generation stage uses a method combining finite structures with infinite variables to combine intersections, main and auxiliary roads, and ramps of a finite road topology with infinite traffic flow variables to generate simulation scenes. The resulting expression is:

[0144]

[0145] in, It is a road network topology template; It is a set of vehicle behavior strategies; It is an environment variable;

[0146] Then, combined with the Wasserstein GAN network, a realistic traffic operation scenario is generated. The generator adopts the Transformer architecture, and the recursive formula for generating time-series traffic flow yields the state at time t+1 as follows:

[0147] in Let t be the state at time t. Let be random noise at time t. This refers to the conditional information at time t;

[0148] The formula for calculating attention mechanism is:

[0149] ;

[0150] S3. Constructing a fully networked, real-world intelligent closed-loop decision-making system:

[0151] The bidirectional interactive channel is composed of a virtual-real feedback mapping system. The real-to-virtual mapping uses sensors to acquire the real-time state of traffic participants and reconstruct the simulation environment, resulting in:

[0152]

[0153] in, It is a digital transportation element. It is a virtualized transportation pipeline. It is a physical transportation element;

[0154] Then, the virtual-to-real feedback is tested and optimized in simulation before being deployed to the actual control system, resulting in:

[0155]

[0156] in, It is an analog control strategy. It is a real control strategy. It is a platform for deploying and distributing tasks;

[0157] Then, based on reinforcement learning algorithms, the objective function is optimized to obtain the optimal policy. ,in For strategy The utility value is obtained as follows:

[0158]

[0159] in, It is the mathematical expectation of the strategy; It is a strategy The utility value.

[0160] Example 2:

[0161] A digital road network cloud service platform system includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements a digital road network cloud service platform architecture as described in Embodiment 1.

[0162] Furthermore, based on the constructed "data-foundation-simulation" technology framework, it provides user-friendly simulation task configuration, execution, and evaluation services, achieving full-link coverage from technology R&D to business implementation. A cloud service system is developed to support user simulation task generation and scenario testing, outputting user-defined multi-dimensional evaluation metrics such as trajectory offset and emergency braking frequency through a configurable rule engine. Users can create any simulation task on the platform, supporting custom task names, inputting task descriptions, selecting task maps, and choosing task models. Specific messages are retrieved from the simulation system, such as vehicle location messages, traffic participant perception messages, high-precision maps, traffic light status, and planned trajectories, to monitor and evaluate specific metrics selected by the client, providing corresponding detailed evaluation data.

[0163] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0164] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A digital road network cloud service platform architecture, characterized in that, It includes a multi-source heterogeneous data fusion engine based on spatiotemporal joint index, a multimodal geospatial data intelligent fusion engine, and a high-precision simulation engine. The multi-source heterogeneous data fusion engine based on spatiotemporal joint index is connected to the multimodal geospatial data intelligent fusion engine, and the multimodal geospatial data intelligent fusion engine is connected to the high-precision simulation engine. The multi-source heterogeneous data fusion engine based on spatiotemporal joint index is used to provide distributed traffic big data processing services. It integrates the full-domain data of physical devices of people, vehicles and roads through the multi-source heterogeneous data fusion engine based on spatiotemporal joint index, and obtains multi-source heterogeneous data based on spatiotemporal joint index. It can accurately locate and query heterogeneous data sources that are accessed in real time from vehicle sensors, roadside units and mobile terminals based on the index. The multimodal geospatial data intelligent fusion engine, based on a spatiotemporal joint index multi-source heterogeneous data fusion engine, achieves centimeter-level spatial registration and dynamic rendering of standard precision maps, high precision maps, 3D real scenes and 3D model data, forming a computable digital twin base for virtual-real mapping; The high-precision simulation engine is based on a multimodal geospatial data intelligent fusion engine. By constructing a three-dimensional technology system of behavior modeling, scene construction, and virtual-real feedback, it can achieve millimeter-level simulation and intelligent decision verification of urban transportation systems. The method for constructing the multi-source heterogeneous data fusion engine based on spatiotemporal joint index includes the following steps: Step 1. Mathematically define the composite code of traffic elements to obtain the traffic entity set E, where each entity in the traffic entity set... Having spatial coordinates timestamp and category identifier The spatiotemporal attribute triples are then encoded to obtain the industry coding function, spatial coding function, and temporal coding function. Then, a composite coding function with industry coding function, spatial coding function, and time coding function is defined. The expression is: ； in, For industry-specific coding functions, For spatial coding functions, For time coding functions, This is a bit string concatenation operation; Step 2. Based on the spatial encoding function and temporal encoding function obtained in Step S1, design a spatiotemporal proximity query method, and perform joint filtering based on a combination of spatial and temporal filtering to obtain the expression: ； in, For the candidate set of traffic elements, The filtering range for spatial encoding. The filtering range for spatial encoding; Step 3. For the candidate set of traffic elements obtained in Step 2, design a multi-level index structure. The first-level index is a B+ tree index based on the element code prefix, the second-level index is an R-tree index based on the spatiotemporal range, and the third-level index is an inverted index based on the semantic category of traffic elements. Define the index query complexity. ,in, For the total amount of all data, The total amount of data for the candidate set of traffic elements; Step 4. Combining the multi-level index structure obtained in Step 3, a sharding function based on spatiotemporal correlation is used for distributed storage of traffic elements. The expression is: ； in, For a piecewise function based on spatiotemporal correlation, For the number of fragments, The MurmurHash3 algorithm is used; then, the node load of traffic elements is balanced based on consistent hashing, and the construction of a multi-source heterogeneous data fusion engine based on spatiotemporal joint index is completed, resulting in multi-source heterogeneous data based on spatiotemporal joint index.

2. The digital road network cloud service platform architecture according to claim 1, characterized in that, The heterogeneous data from vehicle-mounted sensors, roadside units, and mobile terminals specifically includes the following data: Vehicle-mounted sensor data includes vehicle GPS data, camera data, V2X communication data, and radar data; roadside unit data includes traffic flow monitoring data, road condition detection data, meteorological data, and traffic signal data; mobile terminal data includes GPS positioning data, travel behavior data, crowdsourced road condition data, and user profile data.

3. The digital road network cloud service platform architecture according to claim 2, characterized in that, The construction method of the multimodal geospatial data intelligent fusion engine includes the following steps: Step 1: Perform data fusion on the obtained multi-source heterogeneous data based on spatiotemporal joint index, and establish coordinate system standardization and traffic element semantic structure standardization extraction for data features; Coordinate system standardization is based on the GDAL library to build a coordinate system standardization process, and traffic element semantic structure standardization extraction is based on a deep learning model to build a topology graph. Step 2: Perform data visualization rendering to obtain visualized traffic spatial elements, and define the semantic model of the traffic spatial elements as follows: ； in, This represents the semantic model of the i-th traffic spatial element. Represents the geometric description of the i-th traffic spatial element. This represents the topological connection of the i-th traffic spatial element. Represents the rule constraints for the i-th traffic spatial element. Represents the semantic attribute of the i-th traffic spatial element; Step 3: Perform computable spatial co-verification on the obtained visualized traffic elements, verifying them from four aspects: topological consistency, geometric consistency, temporal consistency, and semantic consistency. The method for topological consistency verification is to check any spatial edge element. At least two valid topology nodes must be connected; otherwise, it is considered a topology break. The method for geometric consistency verification is to use spatial Boolean relations to determine whether the layout relationship of traffic space elements in geometric space has overlapping, intersecting, or crossing errors. The method for verifying temporal consistency is to determine whether traffic spatial elements are consistent in time and spatial sequence. The method for semantic consistency verification is to determine the rationality of the business logic or rules between traffic space elements and avoid semantic conflicts or contradictions.

4. The digital road network cloud service platform architecture according to claim 3, characterized in that, The method for constructing the high-precision simulation engine includes the following steps: S1. Conduct multi-granularity traffic behavior modeling, including micro-level traffic behavior modeling and macro-level traffic behavior modeling; The micro-traffic behavior model uses joint modeling of vehicles and drivers to describe the motion state of vehicles, expressed as: ； in, This represents the IDM driver model. This represents a vehicle lane-changing model. It is the vehicle speed at time t. It is the acceleration at time t. It is the speed difference between the vehicle and the vehicle in front at time t. It is the distance to the car in front at time t. This represents the direction control angle at time t. It is the current position at time t. These are the driver's behavioral preference parameters at time t; The macro-level traffic behavior model is based on a continuous model of traffic flow, and is described as follows: ； in, It's traffic density. It is a flow function, where x is the location of the traffic flow; Using the LWR model for the flow function, the expression is: ； in, It's traffic density. It is the free-flow velocity; It is the saturation density; S2. Traffic scene generation; The generation of the scene library is divided into two stages: scene extraction and scene generation. The scene extraction stage is achieved by combining template matching with a temporal behavior recognition network. In the scene extraction phase, for typical scene behavior sequences, an event detector is constructed, expressed as: ； in, for Vehicle location at any given time for The feature parameters at each time point are calculated, and the category identifier is output. , It is a long short-term memory network; ； in, For normalized exponential functions, and These are the first model parameters and the second model parameters, respectively. It is a parameter function that represents the maximum value; The scene generation stage uses a method combining finite structures with infinite variables to combine intersections, main and auxiliary roads, and ramps of a finite road topology with infinite traffic flow variables to generate simulation scenes. The resulting expression is: ； in, It is a road network topology template; It is a set of vehicle behavior strategies; It is an environment variable; Then, combined with the Wasserstein GAN network, a realistic traffic operation scenario is generated. The generator adopts the Transformer architecture, and the recursive formula for generating time-series traffic flow yields the state at time t+1 as follows: ; in, Let t be the state at time t. Let be random noise at time t. This refers to the conditional information at time t; S3. Construct a fully networked, real-world intelligent closed-loop decision-making system: The bidirectional interactive channel is composed of a virtual-to-real feedback mapping system. The real-to-virtual mapping uses sensors to acquire the real-time state of traffic participants and reconstruct the simulation environment, resulting in: ； in, It is a digital transportation element. It is a virtualized transportation pipeline. It is a physical transportation element; Then, the virtual-to-real feedback is tested and optimized in simulation before being deployed to the actual control system, resulting in: ； in, It is an analog control strategy. It is a real control strategy. It is a platform for deploying and distributing tasks; Then, based on reinforcement learning algorithms, the objective function is optimized to obtain the optimal policy. ,get: ； in, It is the mathematical expectation of the strategy; It is a strategy The utility value.

5. A digital road network cloud service platform system, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed, implements a digital road network cloud service platform architecture as described in any one of claims 1-4.

Images