Federal graph neural network-based vehicle-road cloud multi-source data dynamic fusion method and system

Abstract

The application provides a vehicle-road cloud multi-source data dynamic fusion method and system based on a federal graph neural network, and belongs to the technical field of Internet of Vehicles. The method comprises the following steps: acquiring multi-source data, including APP terminal data, vehicle-mounted terminal data and roadside terminal data; cleaning and extracting features from the multi-source data; inputting the cleaned and feature-extracted multi-source data into a trained federal graph neural network model to obtain a fusion result, wherein the federal graph neural network model adopts a human-vehicle-road dynamic correlation graph constructed based on a federal learning framework; updating the fusion result by using an event triggering and incremental learning mechanism; and ensuring the privacy security of the fusion process through three-level protection of data desensitization, model encryption and access control. The application proposes a cloud multi-source data dynamic fusion architecture driven by a federal graph neural network, and realizes the accurate, real-time and safe fusion of multi-source data through a five-layer closed-loop design of data access-feature standardization-correlation graph construction-dynamic update-privacy protection.

Description

Technical Field

[0001] This invention relates to the field of vehicle networking technology, and in particular to a method and system for dynamic fusion of multi-source data from vehicles, roads, and the cloud based on federated graph neural networks. Background Technology

[0002] After the Internet of Vehicles (IoV) entered the "vehicle-road-cloud integration" stage, data sources have expanded from single in-vehicle terminals to a three-dimensional approach encompassing "APP + in-vehicle + roadside devices" (by 2025, each vehicle was generating over 10GB of data per day, and roadside devices were generating over 100TB of data per day). Efficient fusion of multi-source data is a core prerequisite for achieving precise services and intelligent decision-making. However, existing cloud-based data fusion technologies suffer from four major bottlenecks: "association failure, poor real-time performance, privacy exposure, and weak scenario adaptability."

[0003] Low accuracy of cross-platform association: Traditional integration relies on simple matching of "timestamp + location" and ignores the association of behavioral logic (such as the strong correlation between a user's APP search for "charging pile" and the vehicle's SOC <20%). A pilot project showed that the cross-platform data association error rate was as high as 38%, which directly led to "recommendation mismatch" (such as recommending charging piles to fuel vehicles).

[0004] Dynamic update lag: The "batch offline fusion" mode (updated once per hour) cannot capture real-time changes (such as sudden failure of roadside charging piles or sudden reduction in range due to rapid vehicle acceleration). The timeliness deviation of the fusion results exceeds 5 minutes, which cannot meet the millisecond-level decision-making requirements of vehicle networking.

[0005] Privacy and security risks are prominent: In order to improve the accuracy of data fusion, raw data (such as user location and driving trajectory) is directly transmitted to the cloud. In 2024, 67% of the data breaches in the Internet of Vehicles (IoV) system were caused by the lack of anonymization processing in the multi-source data fusion process.

[0006] Weak scene adaptability: Fixed fusion rules (such as "association is made if the position error is less than 50 meters") cannot be adapted to complex scenarios (such as weak GPS signals in tunnels and multipath effects in urban canyons), and the data fusion efficiency in mountainous road sections is only 41%.

[0007] In summary, the urgent technical problem to be solved is how to break through the limitations of traditional cloud data fusion, such as "static matching, privacy protection, and rigid scenarios," and achieve accurate association, real-time updates, privacy and security, and scenario-adaptive fusion of multi-source heterogeneous data from vehicles, roads, and clouds, so as to provide a reliable data foundation for intelligent services of the Internet of Vehicles. Summary of the Invention

[0008] This invention proposes a technical solution for dynamic fusion of multi-source vehicle-road-cloud data based on federated graph neural networks, aiming to improve the accuracy of cross-end data association and solve the problems of "recommendation mismatch" and "decision failure"; reduce the latency of dynamic update of fusion results to support real-time service requirements; basically clear sensitive data, in compliance with the "Several Provisions on Data Security Management of Internet of Vehicles"; improve the fusion efficiency in complex scenarios (tunnels, mountains, urban canyons) and overcome environmental interference limitations.

[0009] In a first aspect, embodiments of the present invention provide a method for dynamic fusion of multi-source data from vehicle-road-cloud based on federated graph neural networks, including:

[0010] Acquire multi-source data, including: APP data, vehicle data, and roadside data;

[0011] The multi-source data is cleaned and its features are extracted.

[0012] The multi-source data after cleaning and feature extraction is input into a trained federated graph neural network model to obtain the fusion result. The federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on a federated learning framework.

[0013] An event-triggered, incremental learning mechanism is used to update the fusion results;

[0014] By employing a three-tiered protection system of data anonymization, model encryption, and access control, privacy and security are ensured throughout the integration process.

[0015] In a preferred embodiment, the step of acquiring multi-source data, including APP data, vehicle-mounted data, and roadside data, involves:

[0016] App-side data includes at least one of the following: user actions and device status;

[0017] Vehicle-mounted data includes at least one of the following: vehicle operating conditions and driving behavior;

[0018] Roadside data includes at least one of the following: infrastructure status and road network environment.

[0019] The built-in protocol conversion engine parses the multi-source data according to the protocol of each multi-source data and unifies the data format to JSON-LD.

[0020] In a preferred embodiment, the steps of cleaning and feature extraction of the multi-source data include:

[0021] Outliers were identified and removed using the 3σ criterion and the isolated forest algorithm.

[0022] When the multi-source data is unstructured, it is transformed into a 128-dimensional semantic vector using the BERT model;

[0023] When the multi-source data is time-series data, 18-dimensional statistical features are extracted using a sliding window.

[0024] When the multi-source data is spatial data, it is converted into planar coordinates through Gaussian projection and then mesh encoded;

[0025] All features are mapped to a unified dimensional space and normalized using min-max, with values ​​ranging from [0,1].

[0026] In a preferred embodiment, the step of inputting the cleaned and feature-extracted multi-source data into a trained federated graph neural network model to obtain a fusion result, wherein the federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on a federated learning framework, includes the following steps:

[0027] The node definitions of the dynamic human-vehicle-road relationship graph include:

[0028] Human nodes: Federated ID, semantic features, preference tags;

[0029] Vehicle node: Vehicle ID, operating condition characteristics, location characteristics;

[0030] Road node: Roadside equipment ID, facility characteristics, environmental characteristics;

[0031] The edge relationships defined in the dynamic human-vehicle-road relationship graph include:

[0032] Strongly correlated edges: When a user's app searches for a charging station and the vehicle's SOC is less than 20%, the weight is set to 0.8-1.0; when the distance between the vehicle and the roadside equipment is less than 500 meters, the weight is set to 0.7-0.9.

[0033] Weakly correlated edges: The weight for matching user historical preferences with roadside facility type is set to 0.3-0.5, and the weight for matching vehicle driving direction with roadside equipment location is set to 0.2-0.4.

[0034] In a preferred embodiment, the step of inputting the cleaned and feature-extracted multi-source data into a trained federated graph neural network model to obtain a fusion result, wherein the federated graph neural network model employs a dynamic human-vehicle-road association graph constructed based on a federated learning framework, includes:

[0035] The model parameters uploaded from the APP, vehicle, and roadside terminals are obtained. The edge parameters are aggregated using the federated averaging algorithm to generate a global correlation model and obtain a trained federated graph neural network model. The model parameters of the APP, vehicle, and roadside terminals are obtained by embedding the training nodes of the graph neural network locally.

[0036] The fusion result is obtained by calculating the association probability between nodes using a pre-trained federated graph neural network model through a graph attention mechanism.

[0037] In a preferred embodiment, the step of updating the fusion result using an event-triggered incremental learning mechanism includes:

[0038] When the triggering conditions for an event are met, the affected subgraph is located using a difference graph.

[0039] Only the node embedding and association probabilities are recalculated for this subgraph;

[0040] The updated association results are cached via a sliding window.

[0041] In a preferred embodiment, the step of ensuring privacy and security during the fusion process through three levels of protection—data anonymization, model encryption, and access control—is as follows:

[0042] The data anonymization includes: applying grid-based differential privacy to location data, retaining only 10m×10m grid information; and generating a federated ID for identity information using SHA-256 with a random salt value.

[0043] The model encryption includes: model parameters uploaded from the APP, vehicle, and roadside terminals are encrypted using the national cryptographic algorithm SM4, and homomorphic encryption is used when the data is aggregated in the cloud;

[0044] The access control includes: distributing the fusion results according to the principle of least privilege.

[0045] In a second aspect, embodiments of the present invention provide a dynamic fusion system for multi-source vehicle-road-cloud data based on federated graph neural networks, configured to implement any of the methods described in the first aspect, the system comprising:

[0046] A multi-source data acquisition module is used to acquire multi-source data, including: APP terminal data, vehicle terminal data, and roadside terminal data;

[0047] The cleaning and feature extraction module is used to clean and extract features from the multi-source data.

[0048] The inference module is used to input the cleaned and feature-extracted multi-source data into the trained federated graph neural network model to obtain the fusion result. The federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on the federated learning framework.

[0049] The update module is used to update the fusion results using an event-triggered incremental learning mechanism;

[0050] The privacy protection module is used to ensure privacy and security during the integration process through three levels of protection: data anonymization, model encryption, and access control.

[0051] Thirdly, embodiments of the present invention provide an electronic device, including:

[0052] One or more processors;

[0053] Memory, used to store one or more programs;

[0054] When the one or more programs are executed by the one or more processors, the one or more processors implement any of the methods described in the first aspect.

[0055] Fourthly, embodiments of the present invention provide a computer-readable medium storing a computer program that, when executed by a processor, implements the steps of any of the methods described in the first aspect.

[0056] This invention achieves breakthrough improvements in four dimensions—association accuracy, real-time performance, privacy and security, and scene adaptability—through dynamic fusion technology driven by federated GNNs. Specific improvements are detailed below:

[0057] Significantly improved correlation accuracy: Improves the accuracy of cross-end data correlation, reduces the occurrence of "recommendation mismatch" (such as pushing charging piles to fuel vehicles), and improves the matching efficiency of intelligent services;

[0058] Real-time performance meets automotive-grade requirements: Reduces dynamic update latency and supports millisecond-level response for "vehicle rapid acceleration → sudden reduction in range → real-time connection with roadside charging piles", improving the lead time for emergency avoidance decisions in high-speed scenarios;

[0059] Privacy and security compliance implementation: reducing the rate of sensitive data leakage, passing the Level 3 certification of the National Cybersecurity Classified Protection, reducing the rate of privacy complaints, and fully complying with the requirements of the Personal Information Protection Law;

[0060] Enhanced adaptability to complex scenarios: Improves the fusion efficiency in scenarios such as tunnels (weak GPS), mountainous areas (sparse roadside equipment), and urban canyons (multipath effect), and enhances data availability under severe weather conditions (heavy rain, heavy fog);

[0061] Significant commercial value: Improves the conversion efficiency of charging services based on precise fusion data, increases the average monthly consumption of vehicle-to-everything (V2X) services per user; and enhances the utilization rate of roadside resources (such as charging piles) to support the revenue growth of partners. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of a method for dynamic fusion of vehicle-road-cloud multi-source data based on federated graph neural networks, provided in an embodiment of the present invention.

[0063] Figure 2 This is a flowchart illustrating one possible implementation of step S2 provided in an embodiment of the present invention.

[0064] Figure 3 This is a flowchart illustrating one possible implementation of step S4 provided in an embodiment of the present invention.

[0065] Figure 4 This is a schematic diagram of a dynamic fusion system for vehicle-road-cloud multi-source data based on federated graph neural networks, provided as an embodiment of the present invention.

[0066] Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0067] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0068] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0069] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0070] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0071] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0072] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information all comply with relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution follows relevant national laws and regulations (e.g., the "Information Security Technology - Personal Information Security Specification"). For example: appropriate measures are taken for personal information access control; restrictions are imposed on the display of personal information; the purpose of using personal information does not exceed the scope of direct or reasonable association; and explicit identity targeting is eliminated when using personal information to avoid precisely locating a specific individual.

[0073] The core of this invention lies in:

[0074] We propose a cloud-based multi-source data dynamic fusion architecture driven by Federated Graph Neural Networks (FNN). Through a five-layer closed-loop design of "data access - feature standardization - association graph construction - dynamic update - privacy protection", we achieve accurate, real-time and secure fusion of multi-source data.

[0075] Federated Graph Neural Network (FedGNN) is a technical framework that combines federated learning with graph neural networks. Graph neural networks (GNNs) are deep learning techniques specifically designed for processing graph-structured data, which consists of nodes and edges. GNNs can effectively learn the relationships and features between nodes. Federated learning is a distributed machine learning framework whose core idea is "the data doesn't move, the model moves," meaning that multiple participants can collaboratively train a model while protecting data privacy and not revealing the original data.

[0076] Figure 1 This is a flowchart illustrating a method for dynamic fusion of vehicle-road-cloud multi-source data based on a federated graph neural network, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0077] Step S1: Obtain multi-source data, including: APP terminal data, vehicle terminal data, and roadside terminal data;

[0078] Step S2: Clean and extract features from the multi-source data to eliminate heterogeneity interference;

[0079] Step S3: Input the cleaned and feature-extracted multi-source data into the trained federated graph neural network model to obtain the fusion result. The federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on the federated learning framework.

[0080] Step S4: Update the fusion result using an event-triggered incremental learning mechanism;

[0081] Step S5 ensures privacy and security during the fusion process through three levels of protection: data anonymization, model encryption, and access control.

[0082] As shown in the steps above, this invention is the first to combine federated learning with GNN to construct a "people-vehicle-road" association graph without transmitting the original data, thus resolving the contradiction between "data silos" and "privacy leaks" and enabling accurate cross-end data matching. This invention also embeds a three-level protection system of "data anonymization, model encryption, and access control" into the entire process to achieve "usable but invisible" data collaboration.

[0083] In some embodiments, step S1, acquiring multi-source data, includes: APP terminal data, vehicle terminal data, and roadside terminal data.

[0084] The data from the APP includes at least one of the following: user behavior (search keywords, click records) and device status (APP version, mobile phone location), and is transmitted via HTTPS protocol;

[0085] The vehicle-mounted data includes at least one of the following: vehicle operating conditions (remaining fuel / SOC, vehicle speed, tire pressure) and driving behavior (frequency of rapid acceleration, braking force), and is transmitted via CAN bus + 5G RedCap protocol.

[0086] Roadside data includes at least one of the following: infrastructure status (number of available charging piles, gas station queue length) and road network environment (congestion index, weather, construction warning), which is forwarded through edge nodes (MQTT protocol).

[0087] It employs a built-in protocol conversion engine (capable of parsing 6 mainstream protocols: HTTPS / CAN / MQTT / DSRC / C-V2X / 5G) to parse the multi-source data according to the protocol of each multi-source data source and unify the data format to JSON-LD (supporting semantic description) to reduce access latency.

[0088] In some embodiments, such as Figure 2 As shown, step S2, the step of cleaning and feature extraction of the multi-source data, includes:

[0089] Step S21: The 3σ criterion and the isolated forest algorithm are used to identify and remove outliers (such as vehicle SOC jump values ​​and congestion information falsely reported by roadside equipment) to improve the filtering accuracy.

[0090] Step S22: When the multi-source data is unstructured data (such as a user searching for "cheap gas stations nearby"), it is converted into a 128-dimensional semantic vector through the BERT model;

[0091] Step S23: When the multi-source data is time-series data (such as the speed sequence of a vehicle within 10 minutes), 18-dimensional statistical features such as mean, variance, and kurtosis are extracted through a sliding window.

[0092] Step S24: When the multi-source data is spatial data (such as latitude and longitude), it is converted into planar coordinates through Gaussian projection and then grid-encoded (10m×10m granularity, balancing accuracy and privacy).

[0093] Step S25: Map all features to a unified dimensional space (256 dimensions) and eliminate dimensional differences through min-max standardization (value range [0,1]).

[0094] As shown in the steps above, this invention designs a unique feature extraction method for heterogeneous data such as APP semantics, vehicle time series, and roadside space, and adapts to complex scenarios through dynamic thresholds (such as relaxing the location matching error in mountainous areas).

[0095] In some embodiments, step S3 involves inputting the cleaned and feature-extracted multi-source data into a trained federated graph neural network model to obtain a fusion result. The federated graph neural network model employs a dynamic human-vehicle-road association graph constructed based on a federated learning framework.

[0096] The node definitions of the dynamic human-vehicle-road relationship graph include:

[0097] Human nodes: Federated ID (anonymized user identifier), semantic features (APP search / click vector), preference tags (such as "discount sensitive");

[0098] Vehicle node: Vehicle ID (encrypted VIN code), operating condition characteristics (SOC / fuel quantity sequence), location characteristics (grid coordinates);

[0099] Road node: Roadside equipment ID, facility characteristics (charging pile power / queue number), environmental characteristics (congestion index / weather);

[0100] The edge relationships defined in the dynamic human-vehicle-road relationship graph include:

[0101] Strongly correlated edges: When a user searches for "charging pile" in the app and the vehicle's SOC is less than 20% (weight 0.8-1.0), and the distance between the vehicle and the roadside equipment is less than 500 meters (weight 0.7-0.9).

[0102] Weakly correlated edges: user historical preferences match the type of roadside facilities (e.g., users frequently visit PetroChina gas stations, weight 0.3-0.5), and vehicle driving direction matches the location of roadside equipment (weight 0.2-0.4).

[0103] In some embodiments, step S3 involves inputting the cleaned and feature-extracted multi-source data into a trained federated graph neural network model to obtain a fusion result. The federated graph neural network model employs a dynamic human-vehicle-road association graph constructed based on a federated learning framework. The steps include:

[0104] The model parameters uploaded from the APP, vehicle, and roadside terminals are obtained. The edge parameters are aggregated using the FedAvg algorithm to generate a global correlation model and obtain a trained federated graph neural network model. The model parameters of the APP, vehicle, and roadside terminals are obtained by embedding nodes trained locally using graph neural network (GNN) (without transmitting the original data).

[0105] The fusion result is obtained by calculating the association probability between nodes using a pre-trained federated graph neural network model through the graph attention mechanism (GAT).

[0106] In some embodiments, such as Figure 3 As shown, step S4, which uses an event-triggered incremental learning mechanism to update the fusion result, includes:

[0107] Step S41: When the triggering condition for the event is met, locate the affected subgraph using the difference graph (average number of affected nodes < 50).

[0108] Step S42: Recalculate node embeddings and association probabilities only for this subgraph to avoid retraining the entire graph;

[0109] Step S43: After the update, the associated results are cached through a sliding window (keeping the most recent 100 results) to support fast backtracking and verification.

[0110] As shown in the steps above, this invention uses difference graphs + subgraph retraining to achieve millisecond-level updates of the fusion results, solving the lag problem of traditional batch fusion.

[0111] The triggering conditions include:

[0112] Data changes triggering events: vehicle SOC decreases, roadside charging pile status changes from "idle" to "busy", user app initiates new search;

[0113] Time-cycle trigger: A micro-update is performed on the correlation graph every 100ms (only the affected nodes are adjusted).

[0114] The update latency control is as follows: update time triggered by a single event < 50ms, update time for periodic updates < 30ms, and total dynamic update latency < 100ms.

[0115] In some embodiments, step S5, through a three-tiered protection system of data anonymization, model encryption, and access control, ensures privacy and security during the fusion process.

[0116] The data anonymization includes: applying grid-based differential privacy (adding Gaussian noise with ε=0.3) to location data, retaining only 10m×10m grid information, and not exposing specific latitude and longitude; and using federated IDs generated by adding random salt values ​​to SHA-256 for identity information, making it impossible to deduce user accounts or vehicle VIN codes.

[0117] The model encryption includes: the model parameters uploaded from the edge (APP, vehicle, roadside) are encrypted using the national cryptographic algorithm SM4, and homomorphic encryption is used when the cloud aggregates the data (supporting parameter summation in encrypted form).

[0118] The access control includes: the fusion results are distributed according to the principle of least privilege (e.g., the navigation service only obtains roadside congestion correlation results and does not access user behavior data).

[0119] Based on the same inventive concept, such as Figure 4 As shown, this embodiment of the invention also provides a dynamic fusion system for vehicle-road-cloud multi-source data based on federated graph neural networks, configured to implement any of the methods described in the above embodiments. The system includes:

[0120] A multi-source data acquisition module is used to acquire multi-source data, including: APP terminal data, vehicle terminal data, and roadside terminal data;

[0121] The cleaning and feature extraction module is used to clean and extract features from the multi-source data.

[0122] The inference module is used to input the cleaned and feature-extracted multi-source data into the trained federated graph neural network model to obtain the fusion result. The federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on the federated learning framework.

[0123] The update module is used to update the fusion results using an event-triggered incremental learning mechanism;

[0124] The privacy protection module is used to ensure privacy and security during the integration process through three levels of protection: data anonymization, model encryption, and access control.

[0125] Based on the same inventive concept, embodiments of the present invention also provide an electronic device. Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Figure 5 As shown, an embodiment of the present invention provides an electronic device including: one or more processors 101, a memory 102, and one or more I / O interfaces 103. The memory 102 stores one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement any of the methods described in the above embodiments; the one or more I / O interfaces 103 are connected between the processor and the memory, configured to enable information interaction between the processor and the memory.

[0126] The processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read / write interface) 103 is connected between the processor 101 and the memory 102, and can realize information interaction between the processor 101 and the memory 102, including but not limited to a data bus (Bus).

[0127] In some embodiments, the processor 101, memory 102, and I / O interface 103 are interconnected via bus 104, and thus connected to other components of the computing device.

[0128] In some embodiments, the one or more processors 101 include a field-programmable gate array.

[0129] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable medium. This computer-readable medium stores a computer program, wherein, when executed by a processor, the program implements the steps of any of the methods described in the above embodiments. The computer-readable storage medium may be a volatile or non-volatile computer-readable storage medium.

[0130] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0131] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0132] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0133] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0134] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0135] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0136] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0137] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0138] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0139] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A method for dynamic fusion of multi-source vehicle-road-cloud data based on federated graph neural networks, characterized in that, include: Acquire multi-source data, including: APP data, vehicle data, and roadside data; The multi-source data is cleaned and its features are extracted. The multi-source data after cleaning and feature extraction is input into a trained federated graph neural network model to obtain the fusion result. The federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on a federated learning framework. An event-triggered, incremental learning mechanism is used to update the fusion results; By employing a three-tiered protection system of data anonymization, model encryption, and access control, privacy and security are ensured throughout the integration process.

2. The method according to claim 1, characterized in that, The step of acquiring multi-source data, including APP data, vehicle data, and roadside data, involves: App-side data includes at least one of the following: user actions and device status; Vehicle-mounted data includes at least one of the following: vehicle operating conditions and driving behavior; Roadside data includes at least one of the following: infrastructure status and road network environment. The built-in protocol conversion engine parses the multi-source data according to the protocol of each multi-source data and unifies the data format to JSON-LD.

3. The method according to claim 1, characterized in that, The steps of cleaning and feature extraction of the multi-source data include: Outliers were identified and removed using the 3σ criterion and the isolated forest algorithm. When the multi-source data is unstructured, it is transformed into a 128-dimensional semantic vector using the BERT model; When the multi-source data is time-series data, 18-dimensional statistical features are extracted using a sliding window. When the multi-source data is spatial data, it is converted into planar coordinates through Gaussian projection and then mesh encoded; All features are mapped to a unified dimensional space and normalized using min-max, with values ​​ranging from [0,1].

4. The method according to claim 3, characterized in that, The step of inputting the cleaned and feature-extracted multi-source data into a trained federated graph neural network model to obtain the fusion result, wherein the federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on a federated learning framework, is as follows: The node definitions of the dynamic human-vehicle-road relationship graph include: Human nodes: Federated ID, semantic features, preference tags; Vehicle node: Vehicle ID, operating condition characteristics, location characteristics; Road node: Roadside equipment ID, facility characteristics, environmental characteristics; The edge relationships defined in the dynamic human-vehicle-road relationship graph include: Strongly correlated edges: When a user's app searches for a charging station and the vehicle's SOC is less than 20%, the weight is set to 0.8-1.0; when the distance between the vehicle and the roadside equipment is less than 500 meters, the weight is set to 0.7-0.

9. Weakly correlated edges: The weight for matching user historical preferences with roadside facility type is set to 0.3-0.5, and the weight for matching vehicle driving direction with roadside equipment location is set to 0.2-0.

4.

5. The method according to claim 4, characterized in that, The step of inputting the cleaned and feature-extracted multi-source data into a trained federated graph neural network model to obtain a fusion result, wherein the federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on a federated learning framework, includes the following steps: The model parameters uploaded from the APP, vehicle, and roadside terminals are obtained. The edge parameters are aggregated using the federated averaging algorithm to generate a global correlation model and obtain a trained federated graph neural network model. The model parameters of the APP, vehicle, and roadside terminals are obtained by embedding the training nodes of the graph neural network locally. The fusion result is obtained by calculating the association probability between nodes using a pre-trained federated graph neural network model through a graph attention mechanism.

6. The method according to claim 4, characterized in that, The step of updating the fusion result using an event-triggered incremental learning mechanism includes: When the triggering conditions for an event are met, the affected subgraph is located using a difference graph. Only the node embedding and association probabilities are recalculated for this subgraph; The updated association results are cached via a sliding window.

7. The method according to claim 4, characterized in that, The three-tiered protection system—data anonymization, model encryption, and access control—ensures privacy and security during the fusion process. The data anonymization includes: applying grid-based differential privacy to location data, retaining only 10m×10m grid information; and generating a federated ID for identity information using SHA-256 with a random salt value. The model encryption includes: model parameters uploaded from the APP, vehicle, and roadside terminals are encrypted using the national cryptographic algorithm SM4, and homomorphic encryption is used when the data is aggregated in the cloud; The access control includes: distributing the fusion results according to the principle of least privilege.

8. A dynamic fusion system for multi-source vehicle-road-cloud data based on federated graph neural networks, characterized in that, The system, configured to implement the method as described in any one of claims 1 to 7, comprises: A multi-source data acquisition module is used to acquire multi-source data, including: APP terminal data, vehicle terminal data, and roadside terminal data; The cleaning and feature extraction module is used to clean and extract features from the multi-source data. The inference module is used to input the cleaned and feature-extracted multi-source data into the trained federated graph neural network model to obtain the fusion result. The federated graph neural network model adopts a dynamic human-vehicle-road association graph constructed based on the federated learning framework. The update module is used to update the fusion results using an event-triggered incremental learning mechanism; The privacy protection module is used to ensure privacy and security during the integration process through three levels of protection: data anonymization, model encryption, and access control.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 7.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images