Vehicle-mounted terminal data transmission method, system and device supporting edge AI preprocessing and medium

Abstract

The invention discloses a vehicle-mounted terminal data transmission method, system and device supporting edge AI preprocessing and a medium, and relates to the technical field of vehicle-mounted data transmission and edge computing, and the method comprises the following specific steps: collecting vehicle operation data, and sending a request to an edge layer to dynamically match a target edge server; vehicle operation data are preprocessed through the edge AI, and intelligent safety analysis is carried out by establishing an abnormal safety state recognition model. And uploading an intelligent safety analysis processing result to a central server to complete closed-loop data transmission. According to the invention, the load of the central server and the network bandwidth pressure are reduced through dynamic matching and edge AI preprocessing. And deep and interpretable security risk identification is realized through the knowledge graph and the graph neural network. And through a differentiated data transmission mechanism, the overall transmission efficiency is optimized while the real-time performance of key alarms is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of vehicle data transmission and edge computing technology, and in particular to a vehicle terminal data transmission method, system, device and medium that supports edge AI preprocessing. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) technology, vehicle operation data (such as location trajectory, driving behavior, and environmental perception data) generated by in-vehicle terminals is characterized by massive volume, real-time processing, and diversity. Currently, most mainstream in-vehicle terminal data transmission solutions adopt a centralized architecture where the terminal directly connects to a remote central server. Under this architecture, all raw data is uploaded directly to the central server via wireless network for storage and analysis without processing. This traditional data transmission and processing model has significant limitations: High data transmission latency is a significant issue. Directly transmitting massive amounts of raw data consumes substantial network bandwidth, increasing latency and making it difficult to guarantee the real-time requirements of critical safety data in scenarios involving high-speed vehicle movement or poor network conditions. Data overload is also a problem. Raw data often contains a large amount of redundant information unrelated to vehicle safety status; uploading all of this to the central server drastically increases its computational and storage load, creating performance bottlenecks and impacting the efficiency of overall data analysis. Furthermore, uneven data processing task allocation is another issue. Current technologies lack intelligent matching mechanisms when allocating data processing nodes to in-vehicle terminals, failing to comprehensively consider factors such as terminal location, real-time server load, and historical collaboration relationships. This results in uneven data processing task allocation, low resource utilization efficiency, and an inability to effectively support efficient collaborative management and accurate safety early warning for large-scale, distributed fleets. Summary of the Invention

[0003] In view of the aforementioned existing problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by this invention is: how to achieve real-time local preprocessing of vehicle operation data and analysis of vehicle safety status through edge AI preprocessing and multi-dimensional intelligent matching mechanism in the operation scenario of large-scale, distributed fleets, thereby reducing the load on the central server while ensuring the real-time transmission of critical safety data.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a vehicle-mounted terminal data transmission method supporting edge AI preprocessing, comprising, Vehicle operation data is collected through the fleet's onboard terminals; After collecting vehicle operation data, the fleet's onboard terminal sends a request to the edge layer to dynamically match the target edge server; The target edge server preprocesses vehicle operation data using edge AI to obtain a key dataset; Based on key datasets, establish an abnormal security status identification model for intelligent security analysis. The data results after intelligent security analysis and processing are uploaded to the central server to complete closed-loop data transmission.

[0006] As a preferred embodiment of the vehicle-mounted terminal data transmission method supporting edge AI preprocessing described in this invention, the method includes: after the fleet vehicle-mounted terminal collects vehicle operation data, it initiates a request to the edge layer to dynamically match a target edge server, including: Obtain the associated status information of the vehicle terminal and the real-time load information of each edge server; Based on the associated status information of the vehicle terminal and the real-time load information of each edge server, a set of candidate service nodes that meet the criteria is selected. Establish a priority matching model to determine the target edge service node for the vehicle terminal from the candidate service node set.

[0007] This invention employs a dynamic matching mechanism to ensure vehicle operation data flows to the most suitable edge server for processing. The matching process integrates the real-time location of the fleet with the load status of the edge servers, optimizing data transmission paths and reducing network latency. A priority matching model based on multi-dimensional priority calculation promotes a balanced distribution of data processing tasks, avoiding localized resource congestion.

[0008] As a preferred embodiment of the vehicle terminal data transmission method supporting edge AI preprocessing described in this invention, the target edge server preprocesses vehicle operation data through edge AI to obtain a key dataset, including: The target edge server normalizes the received vehicle operation data to obtain a basic valid dataset; Load the preset vehicle safety index library and selectively filter data reflecting the safe driving status of vehicles from the basic valid dataset; The data reflecting the safe driving status of vehicles will be constructed into a key dataset.

[0009] This invention utilizes edge AI to process and filter raw operational data in real time at the data source, transforming massive and heterogeneous vehicle operation data into refined key datasets. This transformation process reduces the volume and complexity of the data to be transmitted, alleviating network bandwidth pressure. Because invalid and redundant information is pre-filtered, the purity and value density of the data transmitted to the central server are improved.

[0010] As a preferred embodiment of the vehicle terminal data transmission method supporting edge AI preprocessing described in this invention, the method includes: establishing an abnormal security status identification model based on key datasets for intelligent security analysis, including: Compare the data units in the key dataset with the threshold ranges of the corresponding security dimension indicators in the security knowledge base; Based on the comparison results, the data unit is determined to be in a normal or abnormal security state. For data units that are determined to be in an abnormal safety state, associate the runtime scenario information when the data unit was generated; Generate an analysis report that includes statistics on normal states and details on abnormal states.

[0011] This invention uses a pre-defined, standardized safety knowledge base to analyze and evaluate key datasets, enabling vehicle safety status determination to be completed at the edge, freeing it from reliance on a central server's deep learning model. This localized analysis mechanism significantly shortens the response chain for safety event identification, allowing anomalies to be detected and labeled promptly. Simultaneously, unified knowledge base rules ensure the consistency of analytical conclusions.

[0012] This invention provides a vehicle-mounted terminal data transmission system that supports edge AI preprocessing.

[0013] To address the aforementioned technical problems, the present invention provides the following technical solution: a vehicle-mounted terminal data transmission system supporting edge AI preprocessing, comprising: a data acquisition module, a matching module, a preprocessing module, an analysis module, and a transmission module; The data acquisition module collects vehicle operation data through the fleet's on-board terminal. The matching module is a system in which the fleet's onboard terminal collects vehicle operation data and sends a request to the edge layer to dynamically match target edge servers. The preprocessing module is a process where the target edge server preprocesses vehicle operation data using edge AI to obtain a key dataset; The analysis module establishes an abnormal security status identification model based on key datasets to perform intelligent security analysis. The transmission module uploads the data results after intelligent security analysis and processing to the central server, completing closed-loop data transmission.

[0014] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the vehicle terminal data transmission method supporting edge AI preprocessing.

[0015] The present invention provides a computer-readable storage 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 vehicle terminal data transmission method supporting edge AI preprocessing.

[0016] The beneficial effects of this invention are as follows: This invention solves the problem of uneven data processing node allocation caused by the lack of intelligent task scheduling in existing technologies by using a mechanism where the fleet's vehicle-mounted terminal initiates requests to the edge layer and dynamically matches target edge servers. By utilizing edge AI for real-time preprocessing and key information extraction of vehicle operation data, it solves the problems of data overload and ineffective network bandwidth usage caused by uploading raw data without filtering. Furthermore, by performing intelligent security analysis of key datasets on the edge side based on a pre-set security knowledge base, it solves the problems of response latency and insufficient real-time performance caused by the reliance on a central server for security status identification in traditional architectures. Ultimately, it achieves real-time processing of vehicle operation data throughout the entire process from collection to preprocessing to analysis and reporting. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall process of a vehicle terminal data transmission method supporting edge AI preprocessing according to an embodiment of the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0020] Example 1, referring to Figure 1 As one embodiment of the present invention, this embodiment provides a vehicle terminal data transmission method supporting edge AI preprocessing, including: It should be noted that with the development of vehicle-to-everything (V2X) and intelligent driving technologies, the operational data generated by in-vehicle terminals exhibits characteristics of massive volume, heterogeneity, and strong real-time requirements. Traditional centralized processing architectures are prone to high data transmission latency and excessive load on central servers. Existing solutions lack a multi-dimensional dynamic matching mechanism in the selection of data processing nodes, and the direct back transmission of raw data results in a waste of network and computing resources. Furthermore, security analysis relies entirely on the central server, making it difficult to meet the low-latency identification requirements for driving safety.

[0021] Therefore, addressing the issues of high data transmission latency, data overload, and uneven data processing task allocation in existing vehicle terminal data transmission solutions, the following steps (S1-S5) are implemented: First, vehicle operation data is collected. Second, a request is sent to the edge layer to dynamically match the target edge server, resolving the uneven data processing task allocation problem in existing methods. Third, edge AI preprocesses the vehicle operation data and performs intelligent security analysis using a pre-set security knowledge base. Fourth, the intelligent security analysis results are uploaded to the central server, completing closed-loop data transmission and resolving the high data transmission latency and data overload issues of existing solutions.

[0022] S1: Collect vehicle operation data through the fleet's onboard terminal; S2: After the fleet's onboard terminal collects vehicle operation data, it sends a request to the edge layer to dynamically match the target edge server; S3: The target edge server preprocesses vehicle operation data through edge AI to obtain key datasets; S4: Based on key datasets, establish an abnormal security status identification model for intelligent security analysis; S5: Uploads the data results after intelligent security analysis and processing to the central server to complete closed-loop data transmission.

[0023] Example 2, an embodiment of the present invention, provides a method for transmitting data from an in-vehicle terminal that supports edge AI preprocessing, based on the previous embodiment, including: In step S1, vehicle operation data is collected through the fleet's onboard terminal, including the following steps: The vehicle operation data collected by the fleet's onboard terminals includes location trajectory data, driving behavior data, and environmental perception data; Location trajectory data: sourced from the vehicle positioning module, used to reflect the vehicle's driving path and real-time location.

[0024] Driving behavior data: derived from vehicle control sensors, used to record driver operating characteristics.

[0025] Environmental perception data: originates from sensing devices such as vehicle cameras and millimeter-wave radar, and is used to capture external environmental information such as road conditions, obstacles, and weather around the vehicle.

[0026] In step S2, after the fleet's onboard terminal collects vehicle operation data, it sends a request to the edge layer to dynamically match the target edge server, including the following steps A1-A3: A1: Obtain the associated status information of the vehicle terminal and the real-time load information of each edge server.

[0027] After collecting operational data on-board, a matching request is sent to the edge server cluster. Then, the location of each vehicle in each fleet and the total amount of operational data are obtained, along with the remaining data storage capacity of each edge server in the edge server cluster.

[0028] The acquisition of the location of each vehicle in each fleet and the total amount of operational data includes the real-time collection and preprocessing of the location of each vehicle in each fleet by the vehicle positioning module, and the total amount of operational data by the vehicle terminal collecting data through multiple interfaces and calculating the data volume of a single vehicle for a time period.

[0029] Obtaining the remaining data storage capacity of each edge server in the edge server cluster involves retrieving the total storage capacity and real-time used storage data of each edge server through the edge server cluster management platform, and then calculating the remaining data storage capacity.

[0030] A2: Based on the associated status information of the vehicle terminal and the real-time load information of each edge server, select a set of candidate service nodes that meet the conditions.

[0031] Extract the total operational data of each vehicle in each fleet, sum them to obtain the total operational data of each fleet, and simultaneously obtain the remaining data storage capacity of each edge server from the logs. Edge servers with remaining data storage capacity greater than the total data to be transmitted by each fleet are selected as candidate edge servers for that fleet.

[0032] A3: Establish a priority matching model to determine the target edge service node for the vehicle terminal from the candidate service node set.

[0033] In this embodiment of the application, the specific steps for determining the target edge service node for the vehicle terminal in step A3 are as follows: The real-time location of each vehicle in each fleet and the corresponding location of each candidate edge server are obtained. The distance between each vehicle in each fleet and each candidate edge server is then calculated by averaging these distances, and this average is denoted as the distance between each fleet and each candidate edge server. .

[0034] in, Indicates the team number. Indicates the candidate edge server number. and All are positive integers.

[0035] Obtain the total amount of historical data for each vehicle fleet stored on each candidate edge server from the logs, denoted as . .

[0036] The expression for the priority of each fleet among each candidate edge server is: R=0.5× -0.5× , Where R represents the team With candidate edge servers Priority among them.

[0037] It should be noted that obtaining the real-time location of each vehicle in each fleet and the location of each candidate edge server includes collecting real-time location data of each vehicle in each fleet through the vehicle positioning module and retrieving the deployment location information of each candidate edge server from the edge server cluster management database.

[0038] The candidate edge server with the highest priority is selected as the best edge server for the corresponding fleet.

[0039] It should be noted that the edge server cluster used in this invention is deployed in multiple regions, located between the vehicle terminal and the central server, and is used to preprocess data to reduce the load on the central server. The edge server cluster supports dynamic expansion. When the number of vehicles in a fleet in a certain region exceeds a preset threshold, a backup edge node is automatically activated and added to the cluster. The threshold for the number of vehicles in a fleet is set by professionals according to requirements.

[0040] In an optional implementation, step A3, which determines the target edge service node for the vehicle-mounted terminal, can also employ a fixed geographical area division method. During the network planning phase, the geographic space is pre-divided into several fixed service areas based on administrative regions or the coverage area of ​​cellular network base stations. One or more default edge servers are assigned to each area. When the vehicle-mounted terminal generates data and issues a request, it is mapped to the corresponding preset geographic area using the terminal's real-time GPS positioning information.

[0041] In another optional implementation, step A3, which determines the target edge service node for the vehicle-mounted terminal, can also employ a matching method based on server load status. The edge management node periodically collects the core performance indicators of each server within the cluster and uses these indicators to calculate the real-time load coefficient of each server. When a connection request is received from the vehicle-mounted terminal, the system sorts all available servers in ascending order based on this load coefficient, prioritizing the server with the lowest current load.

[0042] It should be noted that this invention dynamically matches the optimal edge service node for each vehicle terminal by constructing a priority matching model that integrates physical proximity and historical cooperation. This model makes the distribution of data processing tasks more balanced within the edge cluster, avoiding overload or idleness of local nodes. Since the matching process uses the network distance between the terminal and the server as a key factor, the data upload path is shortened, reducing transmission latency. Prioritizing server nodes with historical cooperation makes data correlation analysis and contextual understanding more coherent.

[0043] In step S3, the target edge server preprocesses the vehicle operation data using edge AI to obtain a key dataset, including the following steps B1-B3: B1: The target edge server normalizes the received vehicle operation data to obtain a basic valid dataset.

[0044] The vehicle terminal transmits operational data to the target edge server. The target edge server first normalizes the operational data transmitted by the vehicle terminal through edge AI to obtain a basic valid dataset.

[0045] B2: Load the preset vehicle safety index library and selectively filter data reflecting the safe driving status of vehicles from the basic valid dataset.

[0046] B3: Construct a key dataset from data reflecting the safe driving status of vehicles.

[0047] In this embodiment of the application, the specific steps of selectively filtering data reflecting the safe driving status of vehicles from the basic valid dataset in step B2 are as follows: Establish a mapping relationship between centralized data units of basic valid datasets and safety dimension indicators in the vehicle safety indicator library; The edge AI of the target edge server calls a pre-defined vehicle safety indicator library, which includes various safety indicators and their corresponding basic judgment criteria. Based on the basic valid vehicle dataset, each data point in the basic valid dataset is first matched with each safety indicator in the vehicle safety indicator library. If each data point in the basic valid dataset is associated with only a single safety indicator, a one-to-one mapping relationship is established; if each data point in the basic valid dataset is associated with multiple safety indicators, a one-to-many mapping relationship is established.

[0048] It should be noted that establishing a one-to-one mapping relationship involves adding a classification label to each data point in the basic valid dataset, such as vehicle speed data, braking pressure data, and obstacle distance data. Simultaneously, a unique identifier is assigned to each safety indicator in the vehicle safety indicator library. Precise matching of data labels and indicator identifiers ensures that each data point is mapped to only one safety indicator, avoiding ambiguity in filtering caused by one data point corresponding to multiple indicators.

[0049] Establishing a one-to-many mapping relationship involves adding a classification label to each data point in the base dataset, such as braking distance data. Simultaneously, a unique identifier is assigned to each safety indicator in the vehicle safety indicator library. Through multi-dimensional matching of data labels and indicator identifiers, it is ensured that each data point can be associated with all relevant safety indicators. This achieves a mapping between one data point and multiple safety indicators, avoiding the omission of safety-related dimensions due to single-dimensional matching.

[0050] Calculate the correlation strength between the data unit and the mapped security dimension indicators; Based on the two established mapping relationships, for each piece of data in the basic valid dataset, the basic judgment standard range of the security index corresponding to each piece of data is extracted.

[0051] Each data point is compared with the baseline judgment standard range for each security indicator, and the correlation coefficient between the two is calculated using the Pearson correlation coefficient calculation method.

[0052] Let the sequence of observations of the data unit to be evaluated over n consecutive sampling times be... For the security indicators associated with this data unit, let the basic judgment standard range be... Where L is the lower limit of the interval and U is the upper limit of the interval. Based on this standard interval, an ideal reference value sequence corresponding to the time window is generated. Each of them The value can be set to the median of this range based on the characteristics of the indicator. , representing the expected value sequence of the indicator under ideal conditions.

[0053] The Pearson correlation coefficient between the observed sequence X and the ideal reference sequence Y is calculated using the following expression: , Where r is the Pearson correlation coefficient, which ranges from [−1,1] and its absolute value is used to measure the linear correlation strength between the changing trend of the data unit and the ideal expected trend of the security indicator. n is the number of sampling points within the time window. Let X be the average of the observed sequence. The average value of the ideal reference sequence Y.

[0054] It should be noted that vehicle operation data is highly dynamic and subject to numerous interferences. If the threshold is too low, a large amount of low-correlation data will be retained, increasing redundancy and the risk of false alarms. If the threshold is too high, it may lead to over-screening and missed detections of genuine risks. Setting the correlation threshold to 0.6, above the lower limit of moderate correlation, effectively filters out weakly correlated data caused by random fluctuations or noise, retaining data with a clear trend consistent with safety indicators, thus balancing the rigor and reliability of the screening process.

[0055] The correlation coefficient is compared with the correlation threshold. If the correlation coefficient is greater than or equal to the correlation threshold, the data is considered to be correlated with the corresponding security indicator and is retained. If the correlation coefficient is less than the correlation threshold, the data is considered uncorrelated and is removed.

[0056] In an optional implementation, step B2, which involves selectively filtering data reflecting the safe driving status of vehicles from the basic valid dataset, can also employ an independent threshold comparison method. A pre-defined list of safety thresholds is used, with fixed upper and lower limits set for the core parameters. When data filtering is required, the system iterates through each data entry in the basic dataset, checking if its name appears in the pre-defined list. If the data entry is in the list, its value is further compared with the corresponding fixed threshold. If the value exceeds the threshold range, the entire data entry is retained for analysis.

[0057] In another optional implementation, step B2, which involves selectively filtering data reflecting the safe driving status of vehicles from the basic valid dataset, can also employ a pattern matching filtering method based on fixed event rules. A series of discrete safety event rule templates are defined in the system. Each template corresponds to a combination of data pattern conditions. During filtering, the system compares the time-series data stream in the basic dataset with these predefined fixed event templates. When a segment of data stream matches the condition pattern described by a template, the data points related to that event are extracted as valid output.

[0058] It should be noted that this invention utilizes a structured security knowledge base to perform targeted identification on cleaned basic data. This transforms massive heterogeneous information flows into a key dataset highly focused on security dimensions, clearly defining the analysis objectives. Because the selection criteria are based on a pre-defined indicator system, irrelevant or low-relevance data is effectively filtered out, optimizing the limited computing and storage resources at the edge. The output key dataset directly corresponds to clearly defined security analysis dimensions.

[0059] In step S4, based on the key dataset, an abnormal security status identification model is established for intelligent security analysis, including the following steps C1-C3: C1: Construct a dynamic security knowledge graph based on key datasets.

[0060] A dynamic safety knowledge graph is constructed, where nodes represent different safety entities (such as vehicles, drivers, obstacles ahead, and slippery road surfaces), safety indicators (such as longitudinal acceleration, steering wheel angle, and following distance), or safety states (such as fatigued driving, potential collision, and stable cruise). The edges of the graph represent the relationships between entities, the logical / statistical associations between indicators, or the transition probabilities between states.

[0061] It should be noted that the edges in the graph, acting as directed or undirected links connecting different nodes, have the core function of defining and quantifying the multidimensional interaction mechanisms between security elements. Specifically, these edges first encode the inherent physical or logical relationships between entities, such as subordination, spatial location, and operational behavior. Building upon this, they further characterize the deep coupling relationships between different security indicators. This coupling includes both deterministic logical rules based on domain knowledge and statistical dependency and correlation patterns mined from historical data.

[0062] When a key dataset is received from a specific vehicle, the system first transforms the key dataset into a series of time-stamped fact triples (e.g., Vehicle A - instantaneous speed - 120 km / h, Vehicle A - detected - stationary vehicle B ahead). These fact triples are dynamically injected into the dynamic safety knowledge graph corresponding to that vehicle, serving as new, temporary nodes and edges.

[0063] C2: Establish an abnormal security status identification model through a graph neural network, and input a dynamic security knowledge graph to identify abnormal security statuses; C3: The abnormal security status identification model outputs an abnormal security status identification and analysis report.

[0064] The abnormal security status identification model targets state nodes marked as abnormal and their associated fact nodes. Using graph path tracing technology, it backtracks through a dynamic knowledge graph to find the most critical series of data facts and their relationships that led to the abnormal status judgment. The analysis results are automatically integrated to generate an analysis report containing statistics on normal states and details of abnormal states.

[0065] In this embodiment of the application, the specific steps for establishing the abnormal security state identification model in step S4 are as follows: A graph neural network (GNN) is used to establish an abnormal security status identification model, which is then input into a dynamic security knowledge graph for abnormal security status identification. The abnormal security status identification model uses multi-layer GNN computation. The information of a single data point in the dynamic security knowledge graph will interact and fuse with other related nodes along the edges of the graph in multiple rounds. Finally, it outputs a real-time confidence score for each node representing a security status in the graph. All abnormal status nodes with confidence scores exceeding the threshold and their associated fact nodes are marked as abnormal.

[0066] In an optional implementation, the abnormal security state identification model established in step S4 can also employ symbolic reasoning matching based on expert system rules. This transforms the facts and relationships represented by nodes and edges in the dynamic security knowledge graph into several sets of symbolic logic propositions. An expert rule base with predefined formal rules is invoked. The symbolic propositions are matched against the conditional parts in the rule base one by one. If all the preconditions of a rule are satisfied by the current facts, the rule is triggered to mark a certain security state referred to by its conclusion as abnormal.

[0067] In another optional implementation, the abnormal security state identification model established in step S4 can also employ heterogeneous graph neural network anomaly detection based on meta-paths and attention mechanisms. Several meta-paths are predefined. For each target node in the graph, its multi-hop neighbors are selected along different meta-paths, and weights are aggregated using an attention mechanism. Node representations with multiple types of semantic information are learned through multi-layer stacked aggregation operations to classify abnormal states.

[0068] It should be noted that this invention uses graph neural networks to perform end-to-end reasoning learning on a dynamic security knowledge graph, enabling the abnormal security state identification model to directly utilize the relational structure and semantic information in the graph. Because graph neural networks possess powerful neighborhood information aggregation and transmission capabilities, deep feature interaction and collaborative perception are possible between security entities, indicators, and state nodes, allowing the system to automatically identify complex security risks caused by the nonlinear coupling of multiple indicators.

[0069] In step S5, the data results after intelligent security analysis and processing are uploaded to the central server to complete the closed-loop data transmission, including the following steps D1-D3: D1: The target edge server uses a phased aggregation upload mechanism to send the abnormal security status identification and analysis report to the central server.

[0070] D2: Receive data reception confirmation feedback from the central server.

[0071] D3: Based on the data reception confirmation feedback, record the completion status of this data transmission task locally.

[0072] In this embodiment of the application, the specific steps for completing closed-loop data transmission in step S5 are as follows: The central server serves as the core of data transmission and processing, including uploading data from multiple regional edge servers, global in-depth analysis beyond edge processing capabilities, scheduling edge node computing power and processing priorities, data storage, edge AI preprocessing, and sending updates to the edge servers, forming a collaborative processing mode with the edge servers.

[0073] A phased aggregation upload mechanism is adopted. The target edge server does not immediately upload the complete report after each analysis. Instead, it first pushes the anomaly details, which have the highest real-time requirements, to the central server. Simultaneously, the system temporarily caches the complete analysis report, including detailed evidence chains and normal security status statistics, on the edge side. During network downtime, a background synchronization session is initiated to incrementally upload the cached aggregated report and complete evidence data to the central server in batches. Upon receiving the data, the central server merges the incremental data with the existing records.

[0074] After receiving data, the central server sends a confirmation signal to the target edge server confirming successful data reception. Upon receiving the confirmation signal, the target edge server records the data transmission completion information, thus completing the data transmission.

[0075] In an optional implementation, the closed-loop data transmission in step S5 can also employ a transmission protocol based on priority queues and asynchronous acknowledgments. Before sending data, the target edge server assigns different transmission priorities to different security events according to the urgency level in the analysis report and places the data packets into the corresponding priority transmission queues. During transmission, the system strictly retrieves data packets from the queues in priority order.

[0076] In another optional implementation, the closed-loop data transmission in step S5 can also be completed using a blockchain-based evidence upload method. Before uploading the data, the target edge server generates a digital digest for the analysis report to be uploaded and submits the digest to a lightweight blockchain network for evidence storage. The server then sends the complete analysis report data package along with the blockchain transaction receipt to the central server.

[0077] It should be noted that this invention employs a phased aggregation upload mechanism to decompose the data upload process into two stages: real-time push and background synchronization. This achieves differentiated transmission of critical security information and complete analytical data. Because detailed abnormal status information is reported first, the central server can respond quickly and issue early warnings for high-risk events, meeting the core requirement of low latency for driving safety. Simultaneously, storing the complete report temporarily at the edge avoids continuous large data streams from consuming network bandwidth.

[0078] In summary, this invention addresses the network congestion and excessive central load issues caused by full data backhaul through edge AI preprocessing and security knowledge base filtering. By constructing a dynamic security knowledge graph and utilizing graph neural networks for deep relational reasoning, it achieves the identification and status assessment of complex driving safety risks. A phased aggregation upload mechanism ensures real-time reporting of critical alarms while optimizing the transmission efficiency of complete analysis results.

[0079] Example 3 is an embodiment of the present invention, which provides a vehicle terminal data transmission system supporting edge AI preprocessing, including: a collection module, a matching module, a preprocessing module, an analysis module, and a transmission module; The data acquisition module collects vehicle operation data through the fleet's on-board terminal. The matching module is a system in which the fleet's onboard terminal collects vehicle operation data and sends a request to the edge layer to dynamically match target edge servers. The preprocessing module is a process where the target edge server preprocesses vehicle operation data using edge AI to obtain a key dataset; The analysis module establishes an abnormal security status identification model based on key datasets to perform intelligent security analysis. The transmission module uploads the data results after intelligent security analysis and processing to the central server, completing closed-loop data transmission.

[0080] This embodiment also provides an electronic device applicable to a vehicle-mounted terminal data transmission method supporting edge AI preprocessing, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the vehicle-mounted terminal data transmission method supporting edge AI preprocessing as proposed in the above embodiment.

[0081] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a vehicle terminal data transmission method supporting edge AI preprocessing as proposed in the above embodiments.

[0082] The storage medium proposed in this embodiment and the method for implementing a vehicle terminal data transmission method supporting edge AI preprocessing proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0083] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data transmission method for an in-vehicle terminal supporting edge AI preprocessing, characterized in that: include, Vehicle operation data is collected through the fleet's onboard terminals; After collecting vehicle operation data, the fleet's onboard terminal sends a request to the edge layer to dynamically match the target edge server; The target edge server preprocesses vehicle operation data using edge AI to obtain a key dataset; Based on key datasets, establish an abnormal security status identification model for intelligent security analysis. The data results after intelligent security analysis and processing are uploaded to the central server to complete closed-loop data transmission.

2. The vehicle terminal data transmission method supporting edge AI preprocessing as described in claim 1, characterized in that: After collecting vehicle operation data, the fleet's onboard terminal sends a request to the edge layer to dynamically match target edge servers, including: Obtain the associated status information of the vehicle terminal and the real-time load information of each edge server; Based on the associated status information of the vehicle terminal and the real-time load information of each edge server, a set of candidate service nodes that meet the criteria is selected. Establish a priority matching model to determine the target edge service node for the vehicle terminal from the candidate service node set.

3. The vehicle terminal data transmission method supporting edge AI preprocessing as described in claim 2, characterized in that: The target edge server preprocesses vehicle operation data using edge AI to obtain a key dataset, including: The target edge server normalizes the received vehicle operation data to obtain a basic valid dataset; Load the preset vehicle safety index library and selectively filter data reflecting the safe driving status of vehicles from the basic valid dataset; The data reflecting the safe driving status of vehicles will be constructed into a key dataset.

4. The vehicle terminal data transmission method supporting edge AI preprocessing as described in claim 3, characterized in that: The step of establishing an abnormal security status identification model based on key datasets for intelligent security analysis includes: Construct a dynamic security knowledge graph based on key datasets; An abnormal security status identification model is established by using a graph neural network, and the abnormal security status is identified by inputting a dynamic security knowledge graph. The abnormal security status identification model outputs an abnormal security status identification and analysis report.

5. The vehicle terminal data transmission method supporting edge AI preprocessing as described in claim 4, characterized in that: The process of uploading the data results after intelligent security analysis and processing to the central server to complete closed-loop data transmission includes: The target edge server uses a phased aggregation upload mechanism to send the abnormal security status identification and analysis report to the central server; Receive confirmation feedback for data reception from the central server; Based on the data reception confirmation feedback, record the completion status of this data transmission task locally.

6. The vehicle terminal data transmission method supporting edge AI preprocessing as described in claim 5, characterized in that: The establishment of a priority matching model to determine the target edge service node for the vehicle terminal from the candidate service node set includes: Calculate the overall distance metric between each edge server and the vehicle terminal that issued the request; Acquire historical data volume measurement of the group to which the vehicle terminals that issued the requests are located and which are processed by each edge server; The comprehensive distance metric and data volume metric are normalized and weighted and fused to generate the matching priority for each edge server; A unique target edge server is assigned to the vehicle terminal based on the matching priority.

7. The vehicle terminal data transmission method supporting edge AI preprocessing as described in claim 6, characterized in that: The loading of the preset vehicle safety index library involves selectively filtering data reflecting the safe driving status of vehicles from the basic valid dataset, including: Establish a mapping relationship between centralized data units of basic valid datasets and safety dimension indicators in the vehicle safety indicator library; Calculate the correlation strength between the data unit and the mapped security dimension indicators; Based on a preset correlation strength threshold, data units that meet the correlation criteria are identified and retained to form the key dataset.

8. A vehicle-mounted terminal data transmission system supporting edge AI preprocessing, employing the vehicle-mounted terminal data transmission method supporting edge AI preprocessing as described in any one of claims 1-7, characterized in that, include: The system comprises an acquisition module, a matching module, a preprocessing module, an analysis module, and a transmission module. The data acquisition module collects vehicle operation data through the fleet's on-board terminal. The matching module is a system in which the fleet's onboard terminal collects vehicle operation data and then sends a request to the edge layer to dynamically match target edge servers. The preprocessing module is a process where the target edge server preprocesses vehicle operation data using edge AI to obtain a key dataset; The analysis module establishes an abnormal security status identification model based on key datasets to perform intelligent security analysis. The transmission module uploads the data results after intelligent security analysis and processing to the central server, completing closed-loop data transmission.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the vehicle terminal data transmission method supporting edge AI preprocessing as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the vehicle terminal data transmission method supporting edge AI preprocessing as described in any one of claims 1 to 7.

Images