A method and system for accurate replication of vehicle-road cooperative data

Abstract

This invention discloses a method and system for accurately replicating vehicle-road cooperative data. The method includes a data extraction step, a data fusion step, a data replication step, and a data sharing step. The data extraction step acquires raw roadside data and raw vehicle-side data, sends beyond-line-of-sight information data to the roadside, sends vehicle-road cooperative early warning information to the vehicles, and standardizes the raw roadside data and raw vehicle-side data. The data fusion step correlates and integrates the standardized roadside data and vehicle-side data to obtain fused perception data, and uses the perception data for scenario-based calculations. The data replication step replicates the raw vehicle-side data, raw roadside data, fused perception data, and scenario-based calculation results. The data sharing step transmits data internally and externally. This invention solves the problem of spatiotemporal misalignment of multi-source data, shortens the intersection early warning response delay, and significantly improves decision reliability.

Description

Technical Field

[0001] This invention relates to the field of cooperative perception for autonomous driving of vehicles, and specifically to a method and system for accurate replication of vehicle-road cooperative data. Background Technology

[0002] With the emergence of new business demands, particularly applications involving cross-intersections, beyond-line-of-sight, and based on comprehensive roadside traffic elements, as well as intelligent transportation data analysis, it is necessary to integrate regional or even global data before conducting related computational work. This data integration includes: spatiotemporal alignment of discrete data from various intersections, matching with high-precision map base maps to confirm their geographical relationships, and converting local coordinates of intersection data to global coordinates. This requires a new collaborative perception infrastructure business architecture oriented towards data flow to support it.

[0003] Existing vehicle-road cooperative systems typically employ a single-intersection independent calculation and local data aggregation model. Their core problems are as follows: First, the multi-source heterogeneous data (inconsistent coordinate systems, time bases, and format protocols) reported by roadside units (RSUs) and vehicle terminals are difficult to align across regions, leading to insufficient accuracy in global traffic situation projection. Second, traditional systems rely on customized application-layer processing for multi-source data fusion, lacking standardized middleware support. Different business modules (such as traffic light optimization and collision warning) require repeated development of data preprocessing logic, resulting in low resource utilization. Third, scenarios such as beyond-line-of-sight perception and cross-intersection path planning require the integration of global roadside perception data and vehicle dynamic trajectories, but the existing architecture cannot efficiently complete high-load calculations such as coordinate system conversion and local coordinate global mapping, hindering real-time business performance.

[0004] Therefore, there is an urgent need to propose a new type of precise replication system to realize the transformation of vehicle-road cooperative systems from single-point intelligence to global collaboration. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a method for accurate replication of vehicle-road cooperative data, comprising:

[0006] The data extraction step involves acquiring the original roadside data and original vehicle-side data reported by the roadside and vehicles respectively, sending beyond-line-of-sight information data to the roadside and sending vehicle-road cooperative early warning information to the vehicles, and performing standardization processing on the original roadside data and the original vehicle-side data. The standardization processing includes filtering out abnormal original roadside data and original vehicle-side data, and performing standardization processing on the remaining original roadside data and original vehicle-side data.

[0007] The data fusion step involves associating and integrating standardized roadside data and vehicle-side data to obtain fused perception data, and then using the perception data to perform perception calculations.

[0008] The data replication process involves replicating the original vehicle-side data, the original roadside data, the fused perception data, and the perception calculation results.

[0009] The data sharing process involves transmitting data both internally and externally.

[0010] In one embodiment, the data extraction step, which involves filtering out abnormal raw roadside data and raw vehicle-side data, includes:

[0011] Check the reporting frequency of the original roadside data and the original vehicle-side data. If the reporting frequency is lower than the base frequency and the difference between the reporting frequency and the base frequency is within the preset threshold, perform frame interpolation. If it exceeds the preset threshold, filter directly.

[0012] Check the reporting frequency of raw roadside data and raw vehicle-side data. If the reporting frequency is higher than the base frequency and the difference between the reporting frequency and the base frequency is within the preset threshold, perform frame extraction. If it exceeds the preset threshold, filter directly.

[0013] Check whether the original roadside data and original vehicle-side data are within the preset value range. If they exceed the preset value range, filter them directly.

[0014] Check whether the original roadside data and original vehicle data belong to vehicles and intersections in the service list. If they are not in the service list, filter them directly.

[0015] In one embodiment, the standardization process for the remaining raw roadside data and raw vehicle-side data in the data extraction step includes:

[0016] Spatiotemporal alignment includes spatial alignment and temporal alignment. Spatial alignment includes assigning a corresponding spatial partition number to each original roadside data and original vehicle-side data. Temporal alignment includes sorting the original roadside data and original vehicle-side data according to the timestamp, calculating the time of the earliest unprocessed data, and discarding the earliest unprocessed data if the difference between the current time and the time of the earliest unprocessed data exceeds a first time difference threshold.

[0017] Data timeout handling: When the difference between the time of data processing and the system time exceeds the second time difference threshold, the data being processed is discarded or the data being processed is forcibly aligned to the latest frame data.

[0018] Vehicle driving area filtering: Calculate the area to which the vehicle's real-time location belongs; if the vehicle is not within the service area, the current data will be filtered.

[0019] Vehicle data transmission status output: Outputs the status of data packets transmitted per second, including the number of data packets transmitted and the maximum data latency;

[0020] Intersection data transmission status awareness, filtering data based on intersection facility maintenance status, intersection start / stop status, intersection data transmission version, and data channel latency;

[0021] Vehicle lifecycle management includes vehicle creation, activation, and deactivation. Creation refers to the initial registration of vehicle information within the system. The vehicle information includes the vehicle serial number, vehicle model, and certificate. After activation, the vehicle obtains data interaction permissions, which include communication with the vehicle gateway. After deactivation, the vehicle loses data interaction permissions. Once the vehicle gateway establishes a connection with the vehicle, it enters the activated state. If the vehicle gateway and the vehicle experience an abnormal disconnection and re-establish a connection, the activated state of the vehicle can be maintained. When a vehicle is actively taken offline, it enters the deactivated state.

[0022] Vehicle driving area management, dynamically enabling or disabling service areas, wherein the service area is the area where the vehicle-road cooperative data accurate replication method is applied.

[0023] In one embodiment, the data fusion step includes perception fusion, trajectory smoothing, perception computation, decision suggestion, and static map data update;

[0024] Static map data updates are used to send map update instructions for perception fusion, trajectory smoothing, and perception calculation. During perception fusion, trajectory smoothing, and perception calculation, high-precision map data is read and map update results are fed back. After the map data is updated, the static map data update provides services such as importing new version map OD format high-precision map data for inspection, map data format conversion, and map service data version synchronization.

[0025] The trajectory smoothing is to correct the vehicle position information that needs to be displayed in digital twins, including steering angle correction, road departure target correction, and red light violation correction.

[0026] The perception fusion includes ID tracking and lane tracking. ID tracking fuses and deduplicates raw roadside data from the roadside mobile edge computing cloud and raw vehicle-side data from the vehicle, assigning each data a unique serial number. Lane tracking identifies the lane to which the dynamic data belongs.

[0027] The perception computing provides vehicle-road cooperative early warning information, which includes intersection collision warning, pedestrian and non-motorized vehicle collision and blind spot warning, and left turn assistance service.

[0028] The decision recommendations are generated based on data from perception fusion, trajectory smoothing, and perception computing. These recommendations include lane change and speed suggestions for the vehicle.

[0029] In one embodiment, the lane tracking identification of the lane to which the dynamic data belongs includes the following steps:

[0030] Data preprocessing involves constructing an observation probability sequence for the driving path, and the process of constructing the observation probability sequence for the driving path.

[0031] Historical trajectory segmentation: For real-time reported vehicle trajectory data, the trajectory data is segmented by taking the trajectory data of the most recent n seconds as a group each time.

[0032] Based on historical information from dynamic data, lane tracking is optimized using a hidden Markov model, where the dynamic data refers to continuously changing dynamic data.

[0033] In one embodiment, the intersection collision warning in the perception calculation includes the following steps:

[0034] S31. Determine if there is a vehicle within a preset range in front of the current vehicle HV. If so, proceed to the next step; otherwise, end.

[0035] S32. Find the two closest vehicles (RVs) to the current vehicle, along with their speeds and distances, through conflict calculation;

[0036] S33. Calculate the warning time;

[0037] S34. Determine if the warning time exceeds the first-level warning time threshold. If yes, the warning ends; otherwise, proceed to the next step.

[0038] S35. Determine whether the warning time is greater than the second-level warning time threshold. If so, trigger the first-level warning and end the warning; otherwise, trigger the second-level warning and end the warning.

[0039] In one embodiment, step S32, the conflict calculation includes the following steps:

[0040] S321. Determine whether the current vehicle is within the preset range. If not, end; if yes, proceed to step S322.

[0041] S322. Determine whether the current vehicle's downstream intersection direction number N is a preset value. If not, end; otherwise, continue.

[0042] S323. Define the current vehicle entry direction ln1 and number each direction at the intersection clockwise;

[0043] S324. Determine the exit direction based on the current vehicle lane and turn signal or a known path;

[0044] S325. Call N's conflict table, which indicates which directions the two closest vehicles to the current vehicle are in conflict with the current vehicle;

[0045] S326. Filter the two RVs with a speed greater than 3km / h that are closest to the current vehicle, and calculate the time tfi = Sfi / Vfi for each RV to reach the conflict point, where Vfi is the speed of the multiple RVs in m / s, Sfi is the distance of the multiple RVs to the conflict point in m, and i is the vehicle number.

[0046] S327. Determine whether the current vehicle speed Vs is greater than 3km / h. If not, proceed to step S328. If yes, calculate the time ts = Ss / Vs for the current vehicle to reach the conflict point and proceed to step S329, where Ss is the distance from HV to the conflict point in meters and Vs is the speed of HV in meters per second.

[0047] S328. Determine whether the distance between the current vehicle and the conflict point is greater than the preset distance dm. If not, end. If yes, determine whether the conflict point time tfi of the RV is less than the time difference threshold t_Threshold. If yes, issue an intersection warning and end. Otherwise, end directly. Here, t_Threshold is the time difference threshold in seconds.

[0048] Step S329. Determine whether the absolute value of the difference between the conflict point time tfi and ts is less than the time difference threshold t_Threshold. If so, issue an intersection warning and end the process; otherwise, end the process directly.

[0049] In one embodiment, step S33, calculating the warning time includes the following steps:

[0050] If the acceleration, constant speed, and deceleration of the two closest vehicles to the current vehicle are greater than the negative threshold, then the following calculations are performed:

[0051]

[0052] If the deceleration of the two vehicles closest to the current vehicle is less than the negative threshold, then the following calculation is performed:

[0053]

[0054] Warning time

[0055] Where d0 is the minimum safe distance when stationary, in meters, A_set is the deceleration threshold, and S d1 The distance between the first vehicle in front of the RV and the HV is in meters (m) and s. d2 Vs is the actual distance between the second vehicle before the RV and the HV, in meters; Vs is the speed of the HV, in meters per second (m / s); Vs ≠ 0. f1 V represents the speed of the first vehicle in the RV, measured in m / s. f2A is the speed of the second vehicle in the RV, in m / s. f1 The deceleration of the first vehicle is expressed in m / s². 2 A f2 The deceleration of the second-to-first vehicle is expressed in m / s². 2 .

[0056] This invention also proposes a vehicle-road cooperative data accurate replication system, which executes the vehicle-road cooperative data accurate replication method as described above, including: a data extraction module, a data fusion module, a data replication module, and a data sharing module. The data extraction module is connected to the data fusion module and the data replication module respectively. The data replication module is connected to the data fusion module. The data sharing module is connected to the data extraction module, the data fusion module, and the data replication module respectively.

[0057] The data extraction module acquires the original roadside data and original vehicle-side data reported by the roadside and vehicles respectively. The data extraction module sends beyond-line-of-sight information data to the roadside and vehicle-road cooperative early warning information to the vehicles. The data extraction module performs standardization processing on the original roadside data and the original vehicle-side data. The standardization processing includes filtering out abnormal original roadside data and original vehicle-side data, and performing standardization processing on the remaining original roadside data and original vehicle-side data.

[0058] The data fusion module associates and integrates standardized roadside data and vehicle-side data to obtain fused perception data, and uses the perception data to perform perception calculations.

[0059] The data replication module replicates the original vehicle-side data, the original roadside data, the fused perception data, and the perception calculation results.

[0060] The data sharing module is responsible for data transmission between the data extraction module, the data fusion module, and the data replication module, as well as the external data transmission of the data extraction module, the data fusion module, and the data replication module.

[0061] In one embodiment, the data extraction module includes a roadside gateway cluster, a vehicle gateway cluster, and a data standardization processing unit. The roadside gateway cluster is a group of multiple roadside gateways that work together through networking. The roadside gateways communicate with the roadside mobile edge computing cloud and roadside units respectively. The roadside gateways obtain the raw roadside data reported by the roadside mobile edge computing cloud and roadside units, and send beyond-line-of-sight information data to the roadside mobile edge computing cloud and roadside units. The vehicle gateway cluster is a group of multiple vehicle gateways that work together through networking. The vehicle gateways communicate with vehicles, obtain the raw vehicle-side data reported by the vehicles, and send vehicle-road cooperative early warning information to the vehicles. The data standardization processing unit performs data standardization processing on the raw roadside data and the raw vehicle-side data.

[0062] The original roadside data includes traffic participant, traffic event, and traffic light data reported by the roadside mobile edge computing cloud, as well as basic safety messages, vehicle information reports, and traffic light phase and timing information reported by the roadside units; the beyond-line-of-sight information data includes dynamic roadside information and traffic light phase and timing information; the vehicle-side data includes vehicle information data and vehicle navigation data, with the vehicle information data uploaded in real time according to a fixed cycle, and the vehicle navigation data uploaded when the vehicle starts navigation or when the navigation path changes; the vehicle-road cooperative early warning information includes green wave speed guidance information, traffic event information, hazard warning information, roadside sign information, and traffic light prompt information.

[0063] The data replication module includes an original data replication layer, a fused data replication layer, and a calculated data replication layer. The original data replication layer replicates the original vehicle-side data and the original roadside data. The fused data replication layer replicates the fused perception data. The calculated data replication layer replicates the perception calculation results.

[0064] In the data replication module, the original perception replication layer will stop updating the data when the following situations occur: the replication layer identifies that a traffic participant has left the perception area of ​​the current location, or the replication layer identifies that a traffic participant has been lost in tracking.

[0065] The step of stopping the updating of the replicated data includes: stopping the updating of the traffic participant object, all data of the traffic participant object will become historical data and will be gradually discarded in a first-in-first-out manner until the last frame of data is discarded, after which the traffic participant object will completely disappear from the data replication layer.

[0066] The vehicle-road cooperative data accurate replication method and system of the present invention have the following beneficial effects:

[0067] 1) This invention achieves precise integration and three-layer replication of roadside and vehicle-side data across the entire domain by constructing a layered and decoupled modular architecture, supporting beyond-line-of-sight collaborative computing. Through spatiotemporal alignment and dynamic threshold control technologies, it solves the problem of spatiotemporal misalignment of multi-source data, reduces global coordinate mapping errors, shortens intersection warning response delays, and significantly improves decision reliability.

[0068] 2) Based on standardized data interfaces and decoupled service design, it supports plug-and-play high-precision map services, reduces regional road network resource consumption, increases system concurrent access capacity, and has high scalability for digital upgrade of the entire transportation network.

[0069] 3) Through dynamic frame interpolation / frame extraction mechanisms and intelligent data filtering strategies, abnormal data interference is effectively suppressed, the data transmission packet loss rate is significantly reduced, the success rate of vehicle-road cooperative early warning information is improved, and service continuity is ensured in complex scenarios. Attached Figure Description

[0070] Figure 1 This is a flowchart illustrating a method for accurate data replication of vehicle-road cooperative systems according to an embodiment of the present invention.

[0071] Figure 2 This is a schematic diagram of the data flow in the data fusion step of an embodiment of the present invention;

[0072] Figure 3 This is a schematic diagram illustrating the calculation of vehicle distance coefficient according to an embodiment of the present invention;

[0073] Figure 4 This is a schematic diagram of the probability of observing the driving path according to an embodiment of the present invention;

[0074] Figure 5 This is a schematic diagram of an implicit Markov model according to an embodiment of the present invention;

[0075] Figure 6 This is a schematic diagram of an intersection collision algorithm according to an embodiment of the present invention;

[0076] Figure 7 This is a schematic diagram of a conflict calculation algorithm according to an embodiment of the present invention;

[0077] Figure 8 This is a schematic diagram of the system architecture for accurate replication of vehicle-road cooperative data according to an embodiment of the present invention, wherein 1 is a data extraction module, 2 is a data fusion module, 3 is a data replication module, and 4 is a data sharing module;

[0078] Figure 9 This is a schematic diagram of the architecture of a data fusion module according to an embodiment of the present invention;

[0079] Figure 10This is a schematic diagram of the specific architecture of a vehicle-road cooperative data accurate replication system in a specific embodiment of the present invention;

[0080] Figure 11 This is a schematic diagram of the data flow of a vehicle-road cooperative data accurate replication system in a specific embodiment of the present invention. Detailed Implementation

[0081] 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 for illustrative purposes only and are not intended to limit the invention.

[0082] like Figure 1 As shown, this invention proposes a method for accurate replication of vehicle-road cooperative data, comprising:

[0083] S1, Data extraction step: Obtain the original roadside data and original vehicle-end data reported by the roadside and vehicles respectively, send beyond-line-of-sight information data to the roadside and send vehicle-road cooperative early warning information to the vehicles, and perform standardization processing on the original roadside data and the original vehicle-end data. The standardization processing includes filtering out abnormal original roadside data and original vehicle-end data, and performing standardization processing on the remaining original roadside data and original vehicle-end data.

[0084] S2, the data fusion step, involves associating and integrating standardized roadside data and vehicle-side data to obtain fused perception data, and then using the perception data to perform perception calculations.

[0085] S3, the data replication step, replicates the original vehicle-side data, the original roadside data, the fused perception data, and the perception calculation results;

[0086] S4, the data sharing step, involves transmitting data internally and externally.

[0087] In the S1 data extraction step, filtering abnormal raw roadside data and raw vehicle-side data includes: checking the reporting frequency and value range of data items, and filtering data with abnormal frequency, abnormal value range of key information items, non-service vehicles, and non-service intersections. Specifically:

[0088] S1-11 checks the reporting frequency of the original roadside data and the original vehicle-side data. If the reporting frequency is lower than the reference frequency and the difference between the reporting frequency and the reference frequency is within a preset threshold, frame interpolation is performed; if it exceeds the preset threshold, it is directly filtered. Preferably, in this embodiment, frame interpolation is performed if the reporting frequency is lower than the reference frequency and within 5 frames; if it exceeds 5 frames, it is directly filtered. The reference frequency is the normal data reporting interval required by the system.

[0089] S1-12 checks the reporting frequency of raw roadside data and raw vehicle-side data. If the reporting frequency is higher than the base frequency and the difference between the reporting frequency and the base frequency is within a preset threshold, frame skipping is performed. If it exceeds the preset threshold, it is directly filtered. In this embodiment, frame skipping is preferably performed when the reporting frequency is higher than the base frequency but within 5 frames to maintain perception continuity and avoid occasional jitter affecting services. If it exceeds 5 frames, it is directly filtered to prevent outdated data from contaminating the system.

[0090] S1-13 checks whether the original roadside data and original vehicle-side data are within the preset value range. If they exceed the preset value range, they are filtered directly.

[0091] S1-14 checks whether the original roadside data and original vehicle-side data belong to vehicles and intersections in the service list. If they are not in the service list, they are filtered directly.

[0092] In the S1 data extraction step, the remaining raw roadside data and raw vehicle-side data are standardized as follows:

[0093] S1-21 Spatiotemporal Alignment: Spatiotemporal alignment includes spatial alignment and temporal alignment. Spatial alignment involves assigning a corresponding spatial partition number to each piece of raw roadside data and raw vehicle-side data. Temporal alignment involves sorting the raw roadside data and raw vehicle-side data according to their timestamps, calculating the time of the earliest unprocessed data, and discarding the earliest unprocessed data if the difference between the current time and the time of the earliest unprocessed data exceeds a first time difference threshold. Preferably, in this embodiment, the first time difference threshold is 1 second.

[0094] S1-22 Data Timeout Processing: When the difference between the time of processing data and the system time exceeds a second time difference threshold, the data being processed is discarded, or the data being processed is forcibly aligned to the latest frame of data. Preferably, in this embodiment, the second time difference threshold is also 1 second. For example, for traffic light data and vehicle navigation data, when the difference between the time of processing data and the system time exceeds 1 second, the data will be moved to the latest frame for processing. For all timeout data, the difference between the data time and the system time is calculated, and the data is submitted to the timeout discard queue.

[0095] S1-23 Vehicle Driving Area Filtering: Calculate the area to which the vehicle's real-time location belongs. If the vehicle is not within the service area, the current data will be filtered. In this embodiment, the vehicle's driving area information is calculated, and the vehicle's reported data is distributed according to the vehicle's service area information, with vehicle data outside the service area being output separately.

[0096] S1-24 Vehicle Data Transmission Status Output: Outputs the status of data packets transmitted per second, including the number of data packets transmitted and the maximum data latency.

[0097] The S1-25 intersection data transmission status perception system filters data based on factors such as intersection facility maintenance status, intersection start / stop status, intersection data transmission version, and data channel latency.

[0098] S1-26 Vehicle lifecycle management includes vehicle creation, activation, and deactivation. Creation refers to the initial registration of vehicle information within the system. The vehicle information includes the vehicle serial number (vehicle ID), vehicle model, and certificate. After activation, the vehicle obtains data interaction permissions, which include communication with the vehicle gateway. After deactivation, the vehicle loses data interaction permissions. Once the vehicle gateway establishes a connection with the vehicle, it enters the activated state. If the vehicle gateway and the vehicle experience an abnormal disconnection and re-establish a connection, the activated state of the vehicle continues. Once the vehicle is actively taken offline, it enters the deactivated state.

[0099] S1-27 Vehicle driving area management, dynamically enabling or disabling service areas, where the service area is the area where the vehicle-road cooperative data accurate replication method of the present invention is applied.

[0100] The S2 data fusion process includes perception fusion, trajectory smoothing, perception computation, decision suggestion, and static map data update. Perception fusion is divided into ID tracking and lane tracking. For the data flow in the data fusion process, please refer to [link / reference]. Figure 2 Among these processes, ID tracking, lane tracking, trajectory smoothing, perception calculation, and control commands are all based on streaming read and write data from the data replication system. The vehicle trajectory processed by ID tracking simultaneously enters lane tracking. After lane tracking determines the vehicle's lane location, it transitions to trajectory smoothing and perception calculation. Upon updating the static map data, map update commands are sent to ID tracking, lane tracking, trajectory smoothing, and perception calculation. Each service reads the high-precision map data and provides feedback on the map update results. Based on the above calculations and scenario-generated data, decision suggestions are generated, and the results are written into the data replication system, which is then distributed to the vehicle by the vehicle gateway.

[0101] The S2-1 static map data update is used to send map update commands for perception fusion, trajectory smoothing, and perception calculation. During perception fusion, trajectory smoothing, and perception calculation, it reads high-precision map data and provides feedback on the map update results. Static map data updates include importing map OD (Open DRIVE) format checks, map data format conversion, and map service data version synchronization. The map OD format check process includes: the console user inputs a new version of the high-precision map source file that conforms to the OD format standard and requests to initiate a format check. The check includes lane connectivity, traffic sign logical consistency, and geometric topological validity. If the check result is satisfactory, the data enters the conversion queue; otherwise, an error log is returned. The map data format conversion process includes: calling the map format conversion function, generating the converted map data, and pushing the conversion progress to the console in real time. The new version data is written to the map data storage, while the old version is retained and supports rollback. The map service data version synchronization process includes: the map service broadcasts an update notification to all perception computing nodes; the computing service actively requests map data; after loading the new data, the computing service sends an update confirmation to the map service; the map service verifies that all nodes are ready and sends an update confirmation command, then performs a version switch and sends feedback on the update status (success / failure) to the map service. After the map data is updated, the static data service provides functions such as importing high-precision map data in OD format of the new version, map data format conversion, map service data updates, and updating lane tracking, perception computing, and lane equalization calculation data.

[0102] S2-2 perception fusion includes ID tracking and lane tracking.

[0103] ID tracking is a service that fuses, deduplicates, and tracks dynamic data. Dynamic data comes from traffic participants in roadside MECs and from real-time raw vehicle-side data. These two data sources may have overlaps, for example, overlapping roadside MEC sensing areas or overlapping vehicle and traffic participant data. Therefore, this data needs to be fused, deduplicated, and assigned a unique serial number (ID) for global tracking. After each matching and fusion, the data publishing service packages the tracks within a partition and sends them to downstream applications. Once downstream partitions obtain the tracks within their respective regions, they can achieve ID tracking of cross-partition data during data fusion.

[0104] It should be noted that the distance between vehicles is not calculated using the distance between two points, but rather as follows: Figure 3 As shown, when calculating the distance coefficient between car B and car A, two ellipses, one large and one small, are drawn based on car A's heading. The larger ellipse is used to exclude vehicles; vehicles outside the larger ellipse do not need to be matched. The smaller ellipse is used to match vehicles; the smaller the coefficient, the higher the matching success rate.

[0105] Lane tracking identifies the lane to which dynamic data belongs. Based on data from the current and previous periods within a certain time range, the lane information of the dynamic data is determined using an HMM (Hidden Markov Model) algorithm. Lane tracking is a stateless service that supports clustered deployment and can be scaled on demand to support any number of dynamic data points. Lane tracking includes data preprocessing, historical trajectory segmentation, lane tracking algorithm model, real-time lane tracking calculation, and real-time lane query for vehicles. Details are as follows:

[0106] (1) Data preprocessing: Data preprocessing involves constructing an observation probability sequence for the driving path. The construction process of the driving path observation probability sequence is as follows: Figure 4 As shown. Figure 4 The four hollow dots in the diagram represent a vehicle's trajectory sequence. The rightmost hollow dot represents the vehicle's latest location. As shown in the diagram, the vehicle is currently on road link 1, which has three lanes: lane1, lane2, and lane3. The observation probability sequence construction process is as follows:

[0107] S11. The current road where the vehicle is located has three lanes: lane1, lane2, and lane3, forming three different driving lane paths;

[0108] S12. Calculate the lateral distances (latitude and longitude) of the vehicle's current point from the center lines of lane1, lane2, and lane3, which are d1, d, and d2, respectively;

[0109] S13. Based on the custom scoring rules, score the probability of the vehicle being located in lane 1, lane 2, and lane 3 according to the lateral distance;

[0110] S14. Repeat steps S11, S12, and S13 to obtain the score observation sequence of the driving lane sequence of the N most recent data points in ascending order of timestamp.

[0111] (2) Historical trajectory segmentation: For real-time reported vehicle trajectory data, the trajectory data is segmented once by taking the trajectory data of the most recent n seconds as a group. In this embodiment, n is preferably 3.

[0112] (3) Establish the lane tracking algorithm model. The Hidden Markov Model (HMM) is a probabilistic model about time series. See the formula for the Hidden Markov Model. Figure 5 .

[0113] The formula for the probability distribution of the implicit Markov hypothesis is as follows:

[0114] ,

[0115] In the input parameters, i represents the vehicle's historical lane information, and Pi represents the probability of the vehicle appearing at a certain moment. The output parameters represent the probability Pi of vehicle i being in each lane.

[0116] (4) Real-time lane tracking calculation: The dynamic data processed by the lane tracking service is dynamic data with continuously changing location (information actively reported by vehicles and traffic participants in the roadside MEC). Based on the historical information of the dynamic data, lane tracking is optimized (using Hidden Markov Model (HMM) calculation).

[0117] The computational process for Hidden Markov Models is as follows:

[0118] S21. Retrieve dynamic data with historical information, which has been processed by the ID tracking service. The historical data consists of 10 records (if the dynamic data has just been generated and there are fewer than 10 records, pass in all the historical data corresponding to the dynamic data).

[0119] S22. Obtain the multiple lanes closest to all historical points of dynamic data through the lane query service;

[0120] S23. Obtain the specified lane information through lane information service, calculate the distance from dynamic data to the specified lane, and convert it into observation probability;

[0121] S24. Obtain the lane change probability through the lane migration service. If left or right turn information exists, adjust the lane change probability again;

[0122] S25. By calculating historical and current dynamic data, the lane with the highest probability is selected as the current lane.

[0123] (5) Real-time vehicle lane query provides the ability to query vehicles in the search lane, the lane in which the vehicle is located, and the vehicles behind and in front of the vehicle in the lane, enabling mutual query of vehicle and lane information.

[0124] S2-3 Perceptual computing provides computational capabilities for processing dynamic real-time data (data generated after perception fusion). Perceptual computing provides vehicle-road cooperative early warning information. In this embodiment, vehicle-road cooperative early warning information includes intersection collision warnings, pedestrian and non-motorized vehicle collision and blind spot warnings, and left-turn assistance. The perceptual computing service is a clustered, stateless service that can be expanded on demand to support any number of dynamic data. Through roadside-reported traffic light data and traffic participants, accurately replicated perceptual computing generates early warning information, providing early warning information to connected vehicles, which then display the warnings.

[0125] Specifically, intersection collisions are calculated using Time-to-Collision (TTC) technology. This calculates the collision risk between a vehicle entering the intersection and vehicles in intersecting lanes within the intersection. The algorithm process is divided into scenario condition determination, conflict calculation, calculation of warning time, and warning generation. For example... Figure 6 As shown, the detailed steps for intersection collision warning are as follows:

[0126] S31. Determine if there is a vehicle D meters ahead of the current vehicle Hv. If there is, proceed to the next step; otherwise, end.

[0127] S32. Calculate the collision and find the two closest vehicles (Rv), their speeds (Vf) and distances (S_d). Collision calculation identifies the locations where collisions occur, providing a data basis for subsequent early warning events.

[0128] S33. Calculate the warning time t.

[0129] S34. Determine whether the warning time t is greater than the level 1 warning time threshold T1. If yes, the warning ends without warning; otherwise, proceed to the next step.

[0130] S35. Determine whether the warning time t is greater than the level 2 warning time threshold T2. If so, trigger a level 1 warning and end the warning; otherwise, trigger a level 2 warning and end the warning.

[0131] In conflict calculation, conflict points are defined as conflicts at intersections and at Y-junctions. Conflicts at intersections refer to conflicts between the approach lanes of an intersection, while conflicts at Y-junctions refer to conflicts between the approach lanes of a Y-junction. For example... Figure 7 As shown, the specific process for conflict calculation in S32 is as follows:

[0132] S321. Determine whether HV is within the intersection range D (preferably 100m in this embodiment). If not, end; if yes, proceed to step S322.

[0133] S322. Determine if the number of downstream intersection directions N of HV is a preset value? If not, end; if yes, continue. In this embodiment, the preset value is preferably 3 or 4.

[0134] S323. Define the HV entrance direction ln1 and number each direction of the intersection clockwise.

[0135] S324. Determine the exit direction OutX based on the HV lane and turn signals or a known path.

[0136] S325. Call the conflict table of N, which indicates which directions of RV conflict with HV. In this embodiment, a conflict table is shown in Table 1 below.

[0137] Table 1 Conflict Table

[0138]

[0139] S326. Select RVs with a speed greater than 3km / h and calculate the time tfi = Sfi / Vfi for each RV to reach the conflict point.

[0140] S327. Determine if the HV speed Vs is greater than 3km / h. If not, proceed to step S328. If yes, calculate the time ts = Ss / Vs from HV to the conflict point and proceed to step S329.

[0141] S328. Determine if the distance between HV and the conflict point is greater than dm. If not, end the process. If yes, determine if the conflict point time tfi of RV is less than the time difference threshold t_Threshold. If yes, issue an intersection warning and end the process. Otherwise, end the process directly. In this embodiment, the preferred time difference threshold t_Threshold is 5 seconds, and the distance between HV and the conflict point dm is 10 meters.

[0142] S329. Determine whether the absolute value of the difference between the conflict point time tfi and ts is less than the time difference threshold t_Threshold. If so, issue an intersection warning and terminate the process; otherwise, terminate the process directly.

[0143] Where D represents the intersection range, t_Threshold is the time difference threshold, Vs is the speed of the HV vehicle (m / s), Vfi is the speed of multiple RV vehicles (m / s), i is the vehicle number (i=1, 2, ...), Sfi is the distance from multiple RV vehicles to the conflict point (m), and Ss is the distance from the HV vehicle to the conflict point (m).

[0144] The specific process for calculating the warning time t in step S33 is as follows:

[0145] S331. If the acceleration, constant speed, or deceleration of the distant vehicle is greater than the negative threshold (Af1≥-A_set, Af2≥-A_set);

[0146] ;

[0147] S332. If the deceleration of the remote vehicle is less than the negative threshold (Af1 <- A_set, Af2 <- A_set);

[0148] ;

[0149] ;

[0150] When the main vehicle speed is S333.HV (Vs≠0), the warning time is calculated based on the minimum safe distance:

[0151] ;

[0152] Where d0 is the minimum safe distance when stationary, in meters; in this embodiment, D0 is preferably 10m. A_set is the deceleration threshold; in this embodiment, it is preferably +2.5m / s². 2 Sd1 is the actual distance between the first vehicle in front of the RV and the HV, in meters. Sd2 is the actual distance between the second vehicle in front of the RV and the HV, in meters. Vs is the speed of the HV, in meters per second. Vf1 is the speed of the first vehicle in front of the RV, in meters per second. Vf2 is the speed of the second vehicle in front of the RV, in meters per second. Af1 is the deceleration of the first vehicle in front, in meters per second. 2 Af2 is the deceleration of the second-to-last vehicle, in m / s². 2 .

[0153] In blind spot collision warning scenarios, Uu link communication uses traffic light data reported from the roadside and traffic participants to generate blind spot collision warning information through precise perception calculations, which is then sent to connected vehicles for warning display.

[0154] The core algorithms for pedestrian / non-motorized vehicle collision warning and right-turn blind spot warning are based on a minimum safe distance model, which prioritizes pedestrian / non-motorized vehicle passage while ensuring safety. Warnings and decisions should ensure the vehicle stops before the stop line to avoid collisions with pedestrians / non-motorized vehicles.

[0155] Based on the minimum safe distance model of vehicle braking dynamics:

[0156] ;

[0157] Under vehicle braking and stopping conditions, the vehicle warning and decision distance is simplified as follows:

[0158] ;

[0159] The following procedure is used to calculate whether a collision between a vehicle and a pedestrian / non-motorized vehicle is likely, and to output a warning result:

[0160] 1. Based on intersection and map data, determine whether there are pedestrians crossing the intersection;

[0161] 2. Based on the pedestrian's crossing position, determine whether there are vehicles within the range D (range threshold) of the pedestrian's oncoming direction;

[0162] 3. Determine the distance sd between the vehicle and the person and the warning thresholds s1 (Level 1 warning threshold) and s2 (Level 2 warning threshold), and generate the warning result.

[0163] 1) If sd is greater than s1, no warning is issued;

[0164] 2) If sd is less than s1 and greater than s2, a Level 1 warning and deceleration decision suggestion are given;

[0165] 3) If sd is less than s2, a decision recommendation of Level 2 warning and emergency braking is given.

[0166] Wherein, D is the functional range, preferably 300m in this embodiment, and D depends on the communication distance and functional requirements. T is the avoidance reaction time, preferably 2s in this embodiment. The reaction time includes the driver's reaction time and the vehicle's braking reaction time. d0 is the minimum safe distance when stationary, preferably 10m in this embodiment, considering a vehicle length of 5m and a distance between vehicles of 5m. A is the deceleration threshold, preferably +2.5m / s² in this embodiment. 2 Emergency braking prediction is considered and is configurable. T is the time it takes for the vehicle to travel from its current speed to S0. In this embodiment, 3 seconds is preferred. Considering the change in deceleration decision suggestions with speed and the rationality of the decision issuance time, it is configurable.

[0167] Left-turn assist is a special type of collision warning system used at intersections. Based on the vehicle's driving intention, it determines if the vehicle is turning left and a collision with another vehicle is possible, then issues a warning message to the vehicle, prompting it to slow down. Therefore, left-turn assist adds vehicle intention judgment to the basic intersection collision warning system. When a collision occurs at an intersection and the intention matches that of left-turn assist, a left-turn assist warning is issued directly. Determination of vehicle driving intention:

[0168] Ignoring U-turns, the vehicle's intention to turn left is determined based on the following conditions.

[0169] 1. The vehicle is in the left-turn lane.

[0170] 2. The vehicle turns on its left turn signal when it is in the left turn lane.

[0171] 3. The vehicle's heading angle changes to the left in the left-turn lane.

[0172] If one or more of the above conditions are met, it can be considered an intention to turn left.

[0173] S2-4 Decision Recommendations are generated based on data from perception fusion, trajectory smoothing, and perception computation. In this embodiment, lane change and speed recommendation commands are generated based on events, warning information, traffic lights, and vehicle-reported data. Cooperative control sends decision recommendation information to vehicles. This information provides safe decision-making suggestions for vehicles based on various scenarios and events on the cloud control platform. The decision recommendation information mainly includes lane change and speed recommendations. The decision recommendation information is generated by combining event and scenario warning information, vehicle-reported data, and traffic light data obtained from the data replication. It also obtains warning information generated by events or various scenarios from the data replication, makes single-event lane change decisions based on the warning information and vehicle-reported information in the data replication, and calculates the recommended speed by combining traffic light data with vehicle-reported data in the data replication. After the decision recommendation information is generated, it is written into the data replication and then distributed to designated vehicles by the vehicle grid.

[0174] Single-event lane-change decision-making primarily targets lateral lane-change recommendations. The data sources for this decision mainly include vehicle-reported data (location, speed, heading angle) from data replication, events (including abnormal vehicles and road construction events, with event details including location, type, and impact range), and warning information (collision warnings, blind spot warnings, etc.). Single-event lane-change decision-making queries connected vehicles affected by the event based on event information, determines whether they are in the same lane based on a high-precision map, and generates lane-change avoidance decision-making information.

[0175] The speed guidance decision-making system primarily provides longitudinal speed recommendations. The decision data sources mainly include vehicle-reported data (position, speed, heading angle) and intersection traffic light data. Based on a high-precision map, this module calculates suggested speed ranges for intersections where vehicles can pass through on a green light. If a vehicle cannot pass through an intersection, it prompts the vehicle to stop before the stop line using invalid speed suggestions.

[0176] S2-5 Trajectory Smoothing: This involves correcting vehicle position information displayed in a digital twin, including steering angle correction, road departure correction, and red light violation correction. Trajectory smoothing achieves smoothness through four methods: vehicle position information correction, steering angle correction, red light violation correction, and road departure correction. Vehicle position information correction combines historical vehicle data and dynamics. Steering angle correction uses high-precision map lane heading data to correct abnormal steering angles. Red light violation correction uses high-precision map and traffic light data to correct red light violations. Road departure correction uses high-precision map lane information to correct lane departures.

[0177] This invention also proposes a vehicle-road cooperative data accurate replication system, which executes the vehicle-road cooperative data accurate replication method described above. This vehicle-road cooperative data accurate replication system includes a data extraction module 1, a data fusion module 2, a data replication module 3, and a data sharing module 4. (See also...) Figure 8 The data extraction module is connected to the data fusion module and the data replication module, the data replication module is connected to the data fusion module, and the data sharing module is connected to the data extraction module, the data fusion module, and the data replication module.

[0178] The data extraction module can acquire raw roadside data and raw vehicle-side data reported by roadside personnel and vehicles respectively. It can also send beyond-line-of-sight information data to the roadside and vehicle-road cooperative early warning information to vehicles. The data extraction module performs standardization processing on the raw roadside data and raw vehicle-side data. The standardization processing includes filtering out abnormal raw roadside data and raw vehicle-side data, and standardizing the remaining raw roadside data and raw vehicle-side data.

[0179] The data extraction module includes a roadside gateway cluster, a vehicle gateway cluster, and a data standardization processing unit, which provides vehicle and road data for accurate replication and also has the ability to distribute data.

[0180] A roadside gateway cluster consists of multiple roadside gateways networked together to work collaboratively. Each roadside gateway communicates with both the roadside mobile edge computing cloud (MEC) and roadside units (RSUs). The roadside gateway is a stateless service that supports clustered deployment and can expand its service nodes as needed to support any number of roadside MECs / RSUs. The roadside gateway acquires raw roadside data reported by the roadside mobile edge computing cloud and roadside units, and distributes beyond-line-of-sight information data to both the roadside mobile edge computing cloud and the roadside units. This raw roadside data includes traffic participant, traffic event, and traffic light data reported by the roadside mobile edge computing cloud, as well as basic safety messages, vehicle information reports, and traffic light phase and timing information reported by the roadside units. Beyond-line-of-sight information data includes dynamic roadside information and traffic light phase and timing information. Traffic light phase and timing information includes the intersection where the traffic light is located, its precise location, sequence number, direction angle, the intersection where the traffic light phase is located, the traffic light phase attribute, the sequence number of the traffic light phase, the sequence number of the next intersection corresponding to the traffic light phase, and the road segment information of the approach lane of the next intersection corresponding to the traffic light phase.

[0181] In most scenarios requiring sustained connectivity and efficient communication, HTTP-like communication protocols often fail to achieve optimal performance. This invention addresses vehicle-to-infrastructure (V2I) data extraction by employing a long-lived connection approach, which helps optimize performance in scenarios involving prolonged connection maintenance, frequent data interaction, and waiting for data notifications. For example, the data extracted by the data extraction module is diverse. To unify the data extraction process, a unified communication protocol, termed the vrc-generic protocol, is defined. The unified communication protocol defined in this invention is based on long-lived connections and reliable data transmission protocols for underlying data communication. For roadside data extraction applications, TCP is used as the underlying connection method. For vehicle-side data extraction applications, WebSockets (also based on TCP) is used as the underlying connection method. A roadside / vehicle terminal establishes a long-lived connection with the corresponding Gateway, and all data transmissions occur over this connection. TCP / WebSockets encapsulates a long-lived, full-duplex communication link, upon which a general application layer protocol is built.

[0182] A vehicle gateway cluster is a network of multiple vehicle gateways working collaboratively, communicating with vehicles. The vehicle gateway is a stateless service that supports clustered deployment and can be expanded on demand to support any number of vehicles. The vehicle gateway acquires raw vehicle-side data reported by vehicles and sends vehicle-to-infrastructure (V2I) warning information to them. The vehicle gateway establishes a long-term connection with vehicles via wireless communication networks such as 4G / 5G to exchange data. It obtains raw vehicle-side data, such as real-time location, speed, and heading angle, and sends out calculated Uu-V2X warning information. Depending on its computing resources, each service node of the vehicle gateway can support reliable access for TCP links from several (tens or even thousands) of vehicles and manage and maintain these TCP links. Raw vehicle-side data includes vehicle information data and vehicle navigation data. Vehicle information data should be uploaded in real-time at fixed intervals, while vehicle navigation data is uploaded when the vehicle starts navigation or when the navigation path changes. The vehicle information data includes vehicle location, speed, heading angle, driving module status, headlight status, transmission timestamp, longitudinal acceleration, lateral acceleration, steering wheel angle, and accuracy. Vehicle navigation information includes navigation route name, origin, destination, waypoints, total road length, road segment information, road connection type, number of lanes, road segment length, road name, traffic restriction information, road level, road segment range, gradient information, traffic camera information, road segment type, latitude and longitude, slope angle, slope length, elevation difference, and speed limit information. Based on the data reported by vehicles and roadside infrastructure, the vehicle gateway generates corresponding warning / information service data and sends it to the vehicles, enabling them to perform information prompts or warning actions. This type of data is named vehicle-road cooperative warning information, including green wave speed guidance information, traffic event information (including speed limit information and road hazard conditions), hazard warning information (including red light violation warnings, left turn assistance, intersection collision warnings, and vulnerable road user collision warnings), roadside sign information, and traffic light prompts. When a vehicle's OBU (On-Board Unit) certificate is abnormal, the corresponding abnormal information needs to be transmitted.

[0183] The data standardization processing unit performs data standardization processing on the raw roadside data and raw vehicle-side data. As mentioned above, the standardization processing includes filtering out abnormal raw roadside data and raw vehicle-side data, and standardizing the remaining raw roadside data and raw vehicle-side data.

[0184] The data fusion module associates and integrates standardized roadside and vehicle-side data to obtain fused perception data. This perception data is then used for perception calculations, the results of which include vehicle-road cooperative early warning information. Data fusion is based on high-precision maps to achieve temporal and spatial synchronization of multi-source data. It deduplicates multi-source data and establishes a dynamic mapping relationship between multi-source data and static high-precision maps, enabling unified management of multi-source data on the same dimension.

[0185] like Figure 9 As shown, the data fusion module comprises five units: a perception fusion unit, a trajectory smoothing unit, a perception computing unit, a decision suggestion unit, and a static map data unit. The perception fusion unit is further divided into two main modules: ID tracking service and lane tracking service. These modules respectively implement perception fusion, trajectory smoothing, perception computing, decision suggestion, and static map data updates in the data fusion process.

[0186] The data replication module includes a raw data replication layer, a fused data replication layer, and a calculated data replication layer. The raw data replication layer replicates the original vehicle-side data and the original roadside data. The fused data replication layer replicates the fused perception data. The calculated data replication layer replicates the perception calculation results. The data replication module continuously receives raw data uploaded from the vehicle gateway and the roadside gateway. The data fusion module cyclically subscribes to the replicated data and writes the calculated fused perception data into the data replication module. The data replication module cyclically writes the data to the data storage.

[0187] The original perception replication layer will stop updating the replicated data when the following conditions are met:

[0188] 1) The replica layer identifies a traffic participant that has left the perception area of ​​the current location.

[0189] 2) The replication layer identifies a traffic participant object as being lost to tracking. If the traffic participant object cannot be tracked again within a specified time (e.g., 1 second), it is considered to be lost to tracking. At this time, updates for the traffic participant object are stopped. All data of the traffic participant object will become historical data and will be gradually discarded in a first-in-first-out (FIFO) manner until the last frame of data is discarded, at which point the traffic participant object will completely disappear from the data replication layer.

[0190] The data sharing module is responsible for data transmission between the data extraction, data fusion, and data replication modules, as well as data transmission from these modules to external systems. Acting as a data access proxy, data sharing supports all internal and external data flow for accurate replication, ensuring excellent scalability and data interoperability for the entire accurate replication service. The newly added external API provides a loosely coupled, pluggable access proxy for data sharing, meeting the ever-expanding business needs. Data sharing includes data queuing, data subscription, and data routing. Data queuing management is implemented for data forwarded from data extraction and replication, and subsequent data fusion-related computing services consume data from the data replication layer through data subscription. As the central hub of accurate replication, data sharing handles data requests from data fusion and collaborative sensing (excluding data extraction) to the replication layer through data queuing and subscription mechanisms, ensuring accurate data flow within the chain. For interactions between the external API and internal data subscriptions, data sharing provides data routing management, integrating collaborative sensing and data replication information to achieve fast and reliable service data delivery and transmission control. For data storage and other service needs, data sharing can also provide support for data subscription output.

[0191] In one specific embodiment, the vehicle-road cooperative data precision replication system first obtains raw data from the vehicle gateway, roadside gateway, and vehicle information service management through the data extraction module, and performs preprocessing to obtain uplink data. In the data replication module, the raw data replication layer replicates the uplink data, which is then fused by the data fusion module. After fusion, the data is transmitted back to the data replication module, where it is replicated by the fused data replication layer. The replicated data is then transmitted again to the data fusion module for perception calculation. The data after perception calculation is then transmitted back to the data replication module, where it is replicated by the calculated data replication layer. All data transmissions are shared in real time through the data sharing module. See [link to relevant documentation]. Figure 10 The specific data flow in this embodiment can be found in the following reference. Figure 11 .

[0192] The vehicle-road cooperative data accurate replication method and system of the present invention have the following beneficial effects:

[0193] 1) This invention achieves precise integration and three-layer replication of roadside and vehicle-side data across the entire domain by constructing a layered and decoupled modular architecture, supporting beyond-line-of-sight collaborative computing. Through spatiotemporal alignment and dynamic threshold control technologies, it solves the problem of spatiotemporal misalignment of multi-source data, reduces global coordinate mapping errors, shortens intersection warning response delays, and significantly improves decision reliability.

[0194] 2) Based on standardized data interfaces and decoupled service design, it supports plug-and-play high-precision map services, reduces regional road network resource consumption, increases system concurrent access capacity, and has high scalability for digital upgrade of the entire transportation network.

[0195] 3) Through dynamic frame interpolation / frame extraction mechanisms and intelligent data filtering strategies, abnormal data interference is effectively suppressed, the data transmission packet loss rate is significantly reduced, the success rate of vehicle-road cooperative early warning information is improved, and service continuity is ensured in complex scenarios.

[0196] The embodiments described above are merely further illustrations of the present invention and are not intended to limit the present invention in any other way. The present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding modifications and changes based on the present invention, but all such modifications and changes should fall within the protection scope of the present invention.

Claims

1. A method for accurate replication of vehicle-road cooperative data, characterized in that, include: The data extraction step involves acquiring the original roadside data and original vehicle-side data reported by the roadside and vehicles respectively, sending beyond-line-of-sight information data to the roadside and sending vehicle-road cooperative early warning information to the vehicles, and performing standardization processing on the original roadside data and the original vehicle-side data. The standardization processing includes filtering out abnormal original roadside data and original vehicle-side data, and performing standardization processing on the remaining original roadside data and original vehicle-side data. The data fusion step involves associating and integrating standardized roadside data and vehicle-side data to obtain fused perception data, and then using the perception data to perform perception calculations. The data replication process involves replicating the original vehicle-side data, the original roadside data, the fused perception data, and the perception calculation results. Data sharing steps involve transmitting data internally and externally; The data fusion steps include perception fusion, trajectory smoothing, perception computing, decision suggestion, and static map data update. Static map data updates are used to send map update instructions for perception fusion, trajectory smoothing, and perception calculation. During perception fusion, trajectory smoothing, and perception calculation, high-precision map data is read and map update results are fed back. After the map data is updated, the static map data update provides services such as importing new version map OD format high-precision map data for inspection, map data format conversion, and map service data version synchronization. The trajectory smoothing is to correct the vehicle position information that needs to be displayed in digital twins, including steering angle correction, road departure target correction, and red light violation correction. The perception fusion includes ID tracking and lane tracking. ID tracking fuses and deduplicates raw roadside data from the roadside mobile edge computing cloud and raw vehicle-side data from the vehicle, assigning each data a unique serial number. Lane tracking identifies the lane to which the dynamic data belongs. The perception computing provides vehicle-road cooperative early warning information, which includes intersection collision warning, pedestrian and non-motorized vehicle collision and blind spot warning, and left turn assistance service. The decision recommendations are generated based on data from perception fusion, trajectory smoothing, and perception computing. The decision recommendations include lane change and speed suggestion instructions for the vehicle. The perception calculation includes the following steps for intersection collision warning: S31. Determine if there is a vehicle within a preset range in front of the current vehicle HV. If so, proceed to the next step; otherwise, end. S32. Find the two closest vehicles (RVs) to the current vehicle, along with their speeds and distances, through conflict calculation. The conflict calculation includes the following steps: S321. Determine whether the current vehicle is within the preset range. If not, end; if yes, proceed to step S322. S322. Determine whether the current vehicle's downstream intersection direction number N is a preset value. If not, end; otherwise, continue. S323. Define the current vehicle entry direction ln1 and number each direction at the intersection clockwise; S324. Determine the exit direction based on the current vehicle lane and turn signal or a known path; S325. Call N's conflict table, which indicates which directions the two closest vehicles to the current vehicle are in conflict with the current vehicle; S326. Filter the two RVs with a speed greater than 3km / h that are closest to the current vehicle, and calculate the time tfi = Sfi / Vfi for each RV to reach the conflict point, where Vfi is the speed of the multiple RVs in m / s, Sfi is the distance of the multiple RVs to the conflict point in m, and i is the vehicle number. S327. Determine whether the current vehicle speed Vs is greater than 3km / h. If not, proceed to step S328. If yes, calculate the time ts = Ss / Vs for the current vehicle to reach the conflict point and proceed to step S329, where Ss is the distance from HV to the conflict point in meters and Vs is the speed of HV in meters per second. S328. Determine whether the distance between the current vehicle and the conflict point is greater than the preset distance dm. If not, end. If yes, determine whether the conflict point time tfi of the RV is less than the time difference threshold t_Threshold. If yes, issue an intersection warning and end. Otherwise, end directly. Here, t_Threshold is the time difference threshold in seconds. Step S329. Determine whether the absolute value of the difference between the conflict point time tfi and ts is less than the time difference threshold t_Threshold. If so, issue an intersection warning and end the process; otherwise, end the process directly. S33. Calculate the warning time; S34. Determine if the warning time exceeds the first-level warning time threshold. If yes, the warning ends; otherwise, proceed to the next step. S35. Determine whether the warning time is greater than the second-level warning time threshold. If so, trigger the first-level warning and end the warning; otherwise, trigger the second-level warning and end the warning.

2. The method for accurate replication of vehicle-road cooperative data according to claim 1, characterized in that, In the data extraction step, the filtering of abnormal raw roadside data and raw vehicle-side data includes: Check the reporting frequency of the original roadside data and the original vehicle-side data. If the reporting frequency is lower than the base frequency and the difference between the reporting frequency and the base frequency is within the preset threshold, perform frame interpolation. If it exceeds the preset threshold, filter directly. Check the reporting frequency of raw roadside data and raw vehicle-side data. If the reporting frequency is higher than the base frequency and the difference between the reporting frequency and the base frequency is within the preset threshold, perform frame extraction. If it exceeds the preset threshold, filter directly. Check whether the original roadside data and original vehicle-side data are within the preset value range. If they exceed the preset value range, filter them directly. Check whether the original roadside data and original vehicle data belong to vehicles and intersections in the service list. If they are not in the service list, filter them directly.

3. The method for accurate replication of vehicle-road cooperative data according to claim 1, characterized in that, In the data extraction step, the standardization processing of the remaining original roadside data and original vehicle-side data includes: Spatiotemporal alignment includes spatial alignment and temporal alignment. Spatial alignment includes assigning a corresponding spatial partition number to each original roadside data and original vehicle-end data. Temporal alignment includes sorting the original roadside data and original vehicle-end data according to the timestamp, calculating the time of the earliest unprocessed data, and discarding the earliest unprocessed data if the difference between the current time and the time of the earliest unprocessed data exceeds a first time difference threshold. Data timeout handling: When the difference between the time of data processing and the system time exceeds the second time difference threshold, the data being processed is discarded or the data being processed is forcibly aligned to the latest frame data. Vehicle driving area filtering: Calculate the area to which the vehicle's real-time location belongs; if the vehicle is not within the service area, the current data will be filtered. Vehicle data transmission status output: Outputs the status of data packets transmitted per second, including the number of data packets transmitted and the maximum data latency; Intersection data transmission status awareness, filtering data based on intersection facility maintenance status, intersection start / stop status, intersection data transmission version, and data channel latency; Vehicle lifecycle management includes vehicle creation, activation, and deactivation. Creation refers to the initial registration of vehicle information within the system. The vehicle information includes the vehicle serial number, vehicle model, and certificate. After activation, the vehicle obtains data interaction permissions, which include communication with the vehicle gateway. After deactivation, the vehicle loses data interaction permissions. Once the vehicle gateway establishes a connection with the vehicle, it enters the activated state. If the vehicle gateway and the vehicle experience an abnormal disconnection and re-establish a connection, the activated state of the vehicle can be maintained. When a vehicle is actively taken offline, it enters the deactivated state. Vehicle driving area management, dynamically enabling or disabling service areas, wherein the service area is the area where the vehicle-road cooperative data accurate replication method is applied.

4. The method for accurate replication of vehicle-road cooperative data according to claim 1, characterized in that, The process of identifying the lane to which the dynamic data belongs for lane tracking includes the following steps: Data preprocessing involves constructing an observation probability sequence for the driving path, and the process of constructing the observation probability sequence for the driving path. Historical trajectory segmentation: For real-time reported vehicle trajectory data, the trajectory data is segmented by taking the trajectory data of the most recent n seconds as a group each time. Based on historical information from dynamic data, a hidden Markov model is used to optimize lane tracking. The dynamic data refers to continuously changing dynamic data.

5. The method for accurate replication of vehicle-road cooperative data according to claim 1, characterized in that, Step S33, calculating the warning time includes the following steps: If the acceleration, constant speed, and deceleration of the two closest vehicles to the current vehicle are greater than the negative threshold, then the following calculations are performed: If the deceleration of the two vehicles closest to the current vehicle is less than the negative threshold, then the following calculation is performed: ； Warning time ; Where d0 is the minimum safe distance when stationary, in meters, A_set is the deceleration threshold, and S d1 The distance between the first vehicle in front of the RV and the HV is in meters (m) and s. d2 Vs is the actual distance between the second vehicle before the RV and the HV, in meters; Vs is the speed of the HV, in meters per second (m / s); Vs ≠ 0. f1 V represents the speed of the first vehicle in the RV, measured in m / s. f2 A is the speed of the second vehicle in the RV, in m / s. f1 The deceleration of the first vehicle is expressed in m / s². 2 A f2 The deceleration of the second-to-first vehicle is expressed in m / s². 2 .

6. A vehicle-road cooperative data accurate replication system, executing the vehicle-road cooperative data accurate replication method as described in any one of claims 1-5, characterized in that, include: The system includes a data extraction module, a data fusion module, a data replication module, and a data sharing module. The data extraction module is connected to the data fusion module and the data replication module, respectively. The data replication module is connected to the data fusion module, respectively. The data sharing module is connected to the data extraction module, the data fusion module, and the data replication module, respectively. The data extraction module acquires the original roadside data and original vehicle-side data reported by the roadside and vehicles respectively. The data extraction module sends beyond-line-of-sight information data to the roadside and vehicle-road cooperative early warning information to the vehicles. The data extraction module performs standardization processing on the original roadside data and the original vehicle-side data. The standardization processing includes filtering out abnormal original roadside data and original vehicle-side data, and performing standardization processing on the remaining original roadside data and original vehicle-side data. The data fusion module associates and integrates standardized roadside data and vehicle-side data to obtain fused perception data, and uses the perception data to perform perception calculations. The data replication module replicates the original vehicle-side data, the original roadside data, the fused perception data, and the perception calculation results. The data sharing module is responsible for data transmission between the data extraction module, the data fusion module, and the data replication module, as well as the external data transmission of the data extraction module, the data fusion module, and the data replication module.

7. The vehicle-road cooperative data accurate replication system according to claim 6, characterized in that, The data extraction module includes a roadside gateway cluster, a vehicle gateway cluster, and a data standardization processing unit. The roadside gateway cluster consists of multiple roadside gateways forming a collaborative group through networking. The roadside gateways communicate with the roadside mobile edge computing cloud and roadside units respectively. The roadside gateways acquire raw roadside data reported by the roadside mobile edge computing cloud and roadside units, and send beyond-line-of-sight information data to the roadside mobile edge computing cloud and roadside units. The vehicle gateway cluster consists of multiple vehicle gateways forming a collaborative group through networking. The vehicle gateways communicate with vehicles, acquire raw vehicle-side data reported by vehicles, and send vehicle-road cooperative early warning information to vehicles. The data standardization processing unit performs data standardization processing on the raw roadside data and raw vehicle-side data. The original roadside data includes traffic participant, traffic event, and traffic light data reported by the roadside mobile edge computing cloud, as well as basic safety messages, vehicle information reports, and traffic light phase and timing information reported by the roadside units; the beyond-line-of-sight information data includes dynamic roadside information and traffic light phase and timing information; the vehicle-side data includes vehicle information data and vehicle navigation data, with the vehicle information data uploaded in real time according to a fixed cycle, and the vehicle navigation data uploaded when the vehicle starts navigation or when the navigation path changes; the vehicle-road cooperative early warning information includes green wave speed guidance information, traffic event information, hazard warning information, roadside sign information, and traffic light prompt information. The data replication module includes an original data replication layer, a fused data replication layer, and a calculated data replication layer. The original data replication layer replicates the original vehicle-side data and the original roadside data. The fused data replication layer replicates the fused perception data. The calculated data replication layer replicates the perception calculation results. In the data replication module, the original perception replication layer will stop updating the data when the following situations occur: the replication layer identifies that a traffic participant has left the perception area of ​​the current location, or the replication layer identifies that a traffic participant has been lost in tracking. The step of stopping the updating of the replicated data includes: stopping the updating of the traffic participant object, all data of the traffic participant object will become historical data and will be gradually discarded in a first-in-first-out manner until the last frame of data is discarded, after which the traffic participant object will completely disappear from the data replication layer.

Images