Internet of vehicles big data real-time monitoring and user service optimization platform for intelligent traffic system

Abstract

The invention belongs to the technical field of intelligent traffic Internet of Vehicles, and discloses an Internet of Vehicles big data real-time monitoring and user service optimization platform for an intelligent traffic system, and the platform comprises a preprocessing unit which carries out the real-time denoising of collected data, unifies the format, and extracts core semantic features. Realizing cross-node semantic feature encryption sharing by relying on a federated learning framework, completing semantic consistency verification by combining a traffic scene knowledge graph, and eliminating conflict abnormal data; the transmission unit identifies the data causal correlation degree based on 5G network slices, and high-correlation data is transmitted through low-delay slices, so that the data correlation is guaranteed; the data fusion unit can dynamically adjust the weight according to the traffic event type, for example, the data weight of the laser radar is improved in an accident scene, high-precision timestamps and dual-mode positioning calibration space-time migration are matched, and finally cross-modal data accurate fusion is achieved; the causal inference unit deploys a lightweight causal Markov model at a roadside edge terminal, and quickly locates traffic anomaly core inducements.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent transportation and Internet of Vehicles, and particularly relates to a real-time monitoring and user service optimization platform for intelligent transportation system based on Internet of Vehicles big data. BACKGROUND

[0002] At present, the integration of Internet of Vehicles and intelligent transportation has become an important direction of intelligent transportation development, and the road network traffic efficiency and safety are improved through the collaborative interaction of vehicles, roads, people and clouds. The surface technologies such as basic data collection and short-distance vehicle-road communication have been applied in many places. With the deep integration of Internet of Vehicles and intelligent transportation, the existing technologies have initially solved the surface problems such as data collection, simple transmission and basic service pushing. However, in the aspect of core technologies for complex traffic scenarios, the following technical problems still exist: Firstly, the semantic alignment and dynamic correlation of multi-source heterogeneous data are insufficient. The data of vehicle end, roadside, user end and cloud end belong to different dimensions, and there is obvious deviation in meaning understanding. The existing technology only realizes the standardization of data format, and does not solve the deep semantic correlation of different types of data, resulting in that the essence of traffic scene cannot be accurately reversed after data fusion, and further causing monitoring misjudgment and service recommendation disconnection. Secondly, the causality of service optimization is missing and the dynamic adaptation is insufficient. The existing service optimization relies on statistical correlation and does not consider the causal logic in the traffic scene, resulting in weak anti-interference ability of service strategy. At the same time, the service parameters are mostly static configuration, which cannot be dynamically adjusted according to sudden traffic events and sudden changes in user behavior, and it is difficult to meet the core needs of real-time response and accurate adaptation of intelligent transportation system. SUMMARY

[0003] The purpose of the present application is to provide a real-time monitoring and user service optimization platform for intelligent transportation system based on Internet of Vehicles big data, to solve the problems raised in the background technology.

[0004] In order to achieve the above purpose, the present application provides the following technical scheme: a real-time monitoring and user service optimization platform for intelligent transportation system based on Internet of Vehicles big data, which comprises: a preprocessing unit: preprocessing the collected data and extracting semantic features, and eliminating abnormal data conflicting with scene semantics; a transmission unit: identifying the causal correlation degree between the preprocessed data, setting the transmission priority, and adopting double encryption technology and dynamically adjusting the transmission rate; a data fusion unit: adjusting the data fusion weight according to the type of traffic event, correcting the spatio-temporal offset of data, and real-time monitoring the fusion effect and adjusting the fusion strategy; a causal inference unit: based on the fusion data, identifying the core inducement of abnormal events, returning the inference result to the regional cloud, calculating the causal entropy change, establishing a scenario-based causal template library and updating it regularly; Causal monitoring unit: map the causal relationship to the fusion data to the twin scene, show the causal chain propagation path and the degree of influence; predict the road network traffic and vehicle trajectory and generate execution instructions, and mine the causal relationship between traffic events and user demand; Service generation unit: establish user portrait, generate optimal travel plan for users and update recommendation strategy in real time; Optimization unit: collect feedback data, identify causal relationship and optimize parameters, and quantify the optimization effect.

[0005] Preferably, the preprocessing unit is specifically as follows: Through the roadside terminal, the data collected by the vehicle terminal sensor, roadside radar and video detector are real-time denoised, unified format, and the core semantic features of the target are synchronously extracted; A cross-node semantic feature encryption sharing mechanism is built by using a horizontal federated learning framework, and each participating node only uploads local model gradient parameters; At the same time, a semantic consistency verification module is embedded, combined with a traffic scene knowledge graph containing traffic scene semantic rules, to eliminate abnormal data conflicting with scene semantics.

[0006] Preferably, the transmission unit is specifically as follows: After the data and semantic annotation preprocessed by the preprocessing unit, a causal correlation priority scheduling mechanism is designed based on 5G network slicing technology, and the edge node identifies the causal correlation degree between the preprocessed data in real time, and the high correlation data is transmitted through low latency slicing, and the non-correlation data is transmitted in batches; A semantic label and data ontology dual encryption scheme is adopted, and only authorized nodes can parse semantic labels; At the same time, a vehicle-to-road communication delay parameter calibration module is added, and the transmission rate is dynamically adjusted to adapt to the communication needs of different traffic scenes.

[0007] Preferably, the data fusion unit is specifically as follows: After the high correlation semantic data is transmitted to the target node by the transmission unit, the data fusion unit establishes a dynamic weight fusion model, automatically adjusts the data fusion weight combined with the traffic event type, including in the accident scene, increasing the weight of laser radar data, and reducing the weight of video and millimeter wave radar data; A cross-modal data alignment module is deployed, and the spatial coordinates based on timestamp and GPS and Beidou dual-mode positioning are calibrated to correct the space-time offset of video and radar data; A data fusion quality evaluation module is integrated, which monitors the fusion effect in real time and dynamically adjusts the fusion strategy through indicators including semantic consistency and key field missing rate; Finally, cross-modal data space-time alignment is realized.

[0008] Preferably, the causal inference unit is specifically as follows: Based on the fusion data of the data fusion unit, a lightweight causal Markov model is deployed on a roadside edge terminal to compress the parameter scale and adapt to the computing power of ARMCortex-A72 architecture, so as to realize traffic anomaly causal positioning and identify the core causes of abnormal events. An iterative optimization mechanism of cloud and edge is designed, the edge terminal returns the inference results including event types and cause probabilities to the regional cloud, the cloud updates the causal network topology based on the full fusion data, calculates the causal entropy change through KL divergence, corrects the edge model parameters when the difference exceeds the preset threshold, and feeds back the optimized model parameters to the edge terminal to infer the causal relationship between traffic events and user demand. A scenario-based causal template library is established to prestore scenario templates and periodically update them based on new fusion data and event data increments.

[0009] Preferably, the causal monitoring unit is specifically as follows: Based on the causal relationship and the fusion data of the data fusion unit, a digital twin road network mapping module using BIM and GIS fusion modeling is established to map the spatio-temporal causal relationship and fusion data to the twin scene in real time, and display the causal chain propagation path and impact degree through a three-dimensional visualization interface. A macroscopic road network flow and microscopic vehicle trajectory dual-scale prediction engine is designed, the macroscopic layer uses a GraphCNN model to predict road network level flow, and the microscopic layer uses an LSTM model to predict vehicle trajectory, both of which interact in real time based on causal inference results, the macroscopic results constrain the microscopic prediction range, and the microscopic results correct the macroscopic model parameters. An integrated service output interface is integrated to directly generate executable instructions and interface with traffic control systems and subsequent user service platforms.

[0010] Preferably, the service generation unit is specifically as follows: Based on the causal relationship between traffic events and user demand inferred by the causal inference unit and the real-time traffic state of the causal monitoring unit, a semantic and causal dual-dimensional user portrait is established, the semantic portrait labels the semantic tags of core needs, and the causal portrait records the causal association between user behavior and service experience, and the portrait data is synchronously associated with the real-time traffic state of digital twin monitoring. A causal recommendation algorithm is used to generate the optimal travel plan for users in combination with fusion data, causal inference conclusions and real-time traffic state; meanwhile, a service strategy dynamic adjustment module is deployed to update the recommendation strategy in real time to ensure service adaptability according to real-time causal factors and changes in traffic state monitored by the twin.

[0011] Preferably, the optimization unit is specifically as follows: After receiving the service scheme of the service generation unit and putting it into application, the optimization unit collects the full-link feedback data of the service scheme, establishes an optimization process of feedback collection, causal analysis, parameter optimization and service iteration, deploys a causal analysis module to analyze and identify the causal relationship of the feedback data, deploys a causal adaptive optimization algorithm to dynamically adjust the traffic monitoring model weight, data update frequency and service recommendation parameters according to the identified causal relationship, and synchronously pushes the optimized parameters to the front-end data fusion unit, the causal inference unit and the service generation unit. An optimization effect evaluation module is established to quantitatively evaluate the optimization effect by indexes including service satisfaction, monitoring accuracy and response timeliness.

[0012] The beneficial effects of the present application are as follows: 1、The present application realizes real-time denoising, unified format and extraction of core semantic features of collected data through the preprocessing unit, realizes cross-node semantic feature encryption sharing relying on a horizontal federated learning framework, completes semantic consistency verification in combination with a traffic scene knowledge graph, and eliminates conflict abnormal data; the transmission unit identifies the causal correlation degree of data based on a 5G network slice, high-correlation data is transmitted through a low-latency slice to ensure data correlation; and the data fusion unit can dynamically adjust the weight according to the type of traffic event, such as increasing the weight of laser radar data in an accident scene, and calibrating the space-time offset by matching a high-precision timestamp and a dual-mode positioning, to finally realize accurate fusion of cross-modal data.

[0013] 2、The causal inference unit of the present application deploys a lightweight causal Markov model on a roadside edge terminal to quickly locate the core inducement of traffic anomalies, and then updates the causal network topology and calculates the causal entropy change amount to correct the model parameters through a cloud-end and edge iterative optimization mechanism, and the scenario-based causal template library can also be regularly updated based on new data; the causal monitoring unit can visualize the causal chain propagation path by means of a digital twin module based on BIM and GIS fusion modeling, and can real-time interactively correct parameters by matching a double-scale engine of a GraphCNN macroscopic flow prediction and an LSTM microcosmic trajectory prediction, and the generated executable instructions are directly connected to a traffic control system, thereby greatly improving the accuracy of traffic anomaly early warning and trend prediction.

[0014] 3、The service generation unit of the present application constructs a semantic and causal dual-dimensional user portrait based on the causal relationship between traffic events and user demand, generates an optimal travel scheme by combining a causal recommendation algorithm, and the service strategy can also be dynamically adjusted according to real-time causal factors; the optimization unit collects service full-link feedback data, identifies the causal relationship in the feedback through a causal analysis module, adjusts parameters such as monitoring model weight and data update frequency by using a causal adaptive optimization algorithm, synchronously pushes them to the front-end unit, and quantitatively evaluates the effect by relying on indexes such as service satisfaction and monitoring accuracy, thereby continuously improving service adaptability and user experience. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A flow chart of a vehicle networking big data real-time monitoring and user service optimization platform for an intelligent transportation system. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0017] As shown in Figure 1 The embodiments of the present application provide a vehicle networking big data real-time monitoring and user service optimization platform for an intelligent transportation system, which comprises: The preprocessing unit is the starting link of platform data flow, and through a roadside terminal, the preprocessing unit performs real-time denoising, unified format, and synchronous extraction of core semantic features such as target position, speed, and type on data collected by vehicle-end sensors, roadside radars, and video detectors; The roadside terminal integrates an edge computing chip and a lightweight semantic processing module, and can process 16 heterogeneous data in parallel; The edge computing chip integrated in the roadside terminal is an ARM architecture chip with a computing power of not less than 2TOPS and a power consumption of less than 15W; the 16 heterogeneous data specifically include 4 vehicle-end laser radar data, 4 roadside millimeter wave radar data, 4 roadside high-definition video data, 2 vehicle networking V2X communication data, and 2 weather sensor data, so as to ensure that the data cover the core monitoring dimensions of the traffic scene; A cross-node semantic feature encryption sharing mechanism is built by using a horizontal federated learning framework, each participating node only uploads local model gradient parameters without leaking original data; at the same time, a semantic consistency verification module is embedded, combined with a traffic scene knowledge graph containing typical traffic scene semantic rules, to eliminate abnormal data conflicting with the scene semantics, improve the data efficiency and semantic feature extraction accuracy, and provide qualified semantic data for the transmission unit.

[0018] The transmission unit receives the data and semantic annotation preprocessed by the preprocessing unit, and designs a cause-effect association priority scheduling mechanism based on 5G network slicing technology, so as to identify the cause-effect association degree between the preprocessed data in real time by an edge node, such as the association relationship between accident data and surrounding road section traffic data, bundle transmission of high-association data through a low-latency slice, and batch transmission of non-association data. 5G network slice is divided into low latency slice and ordinary slice, the end-to-end latency of low latency slice is ≤15ms, the uplink bandwidth is ≥20Mbps, and it is used for transmitting high correlation data; the causal correlation degree is calculated by mutual information value, when the mutual information value of two data modules is ≥0.6, it is determined as high correlation data, and when the mutual information value is <0.6, it is classified as non-correlation data, and ordinary slice is used for batch transmission; A semantic label and data ontology double encryption scheme is adopted, only authorized nodes can parse the semantic label, and the semantic information generated in the preprocessing stage is avoided to be leaked; meanwhile, a vehicle-road communication time delay parameter calibration module is added, and the transmission rate is dynamically adjusted to adapt to the communication demand of different traffic scenes.

[0019] Among them, the data fusion unit is specifically as follows: After the transmission unit transmits the high correlation semantic data to the target node, the data fusion unit establishes a dynamic weight fusion model, and automatically adjusts the data fusion weight in combination with the traffic event type (such as accident, congestion, and severe weather) fed back by the causal inference unit, including that in the accident scene, the weight of laser radar data is increased from 30% to 65%, and at the same time, the weights of video and millimeter wave radar data are reduced; A cross-modal data alignment module is deployed, and based on high-precision timestamps with an error ≤1ms and spatial coordinates of GPS and Beidou dual-mode positioning, the time and space offset of video and radar data is corrected; a data fusion quality evaluation module is integrated, through indexes including a cosine similarity of semantic consistency threshold 0.85 and a key field missing rate ≤0.5%, the fusion effect is monitored in real time and the fusion strategy is dynamically adjusted; finally, the time and space alignment of cross-modal data is realized.

[0020] The semantic consistency cosine similarity calculation object is the semantic feature vector of the data source and the corresponding scene rule vector in the traffic scene knowledge graph, for example, in the accident scene, the similarity of the obstacle position-velocity feature vector extracted by the laser radar data and the obstacle feature rule vector in the knowledge graph in the accident scene needs to be ≥0.85; the key fields include vehicle ID, timestamp, spatial coordinates, and event type, to ensure that the core information is not missing.

[0021] The dynamic weight fusion formula of multi-source data is: ; In the formula: represents the final output result after multi-source semantic data fusion; represents the total number of multi-source data participating in fusion, for example, when laser radar data at the vehicle end, roadside millimeter wave radar data, and roadside video semantic data are fused, =3; represents the dynamic fusion weight of the th data source, which is a correction coefficient about event type a function of the traffic event type, the weight value ranges from [0, 1], and the sum of all data source weights satisfies = 1. represents a traffic event type correction coefficient, different event types correspond to different values, for example, in the accident scenario = 0.65, the weight of the laser radar data source (0.65) = 0.65; in the congestion scenario = 0.4, the weight of the video semantic data source (0.4) = 0.4; in the normal scenario = 0.3, the weights of various data sources tend to be balanced.

[0022] represents the semantic data of the i-th data source after preprocessing (denoising, semantic labeling, and space-time calibration), for example may be an obstacle position-velocity semantic feature vector collected by a laser radar, may be a vehicle number-lane occupancy semantic feature vector extracted from a video image.

[0023] The causal inference unit is specifically as follows: Based on the fusion data of the data fusion unit, a lightweight causal Markov model is deployed on a roadside edge terminal, the parameter size is compressed to 500KB, thereby adapting to the computing power of ARM Cortex-A72 architecture, realizing traffic anomaly causal positioning, and quickly identifying the core causes of abnormal events, such as accidents causing congestion and signal failure causing traffic confusion. The parameter size of the lightweight causal Markov model before compression is 2MB, and after compression, it is reduced to 500KB, and the time consumption of a single traffic anomaly causal inference on the ARM Cortex-A72 architecture is ≤80ms, and the accuracy is ≥92%; the scenario-based causal template library update trigger condition is that the amount of new fusion data ≥10GB and the number of new traffic events ≥50, if the threshold is not reached in the current month, it is postponed to the next month, to ensure that the template library update is based on sufficient effective data; An edge iteration optimization mechanism is designed, the edge terminal returns the inference results including the event type and the cause probability to the regional cloud, the cloud updates the causal network topology based on the full amount of fusion data, calculates the causal entropy change amount through KL divergence, corrects the edge model parameters when the difference exceeds 0.1, and feeds back the optimized model parameters to the edge terminal to infer the causal relationship between traffic events and user demand; ​A scenario-based causal template library is established, and typical scenario templates such as heavy rain and holidays are pre-stored. The library is updated based on the incremental data and event data every month to improve the efficiency of scenario matching and inference, and to improve the accuracy of causal relationship identification and the efficiency of typical scenario causal inference.

[0024] The causal entropy calculation formula is: ; In the formula: represents the conditional causal entropy, which is used to quantify the uncertainty of the result variable under the premise of knowing the cause variable , the smaller the value, the stronger the causal relationship between and ; when the difference between the causal inference result fed back by the edge terminal and the calculated based on the full data in the cloud exceeds 0.1, the edge model parameter correction is triggered; represents the cause variable set of the traffic event, which includes all possible factors that can induce traffic anomalies, such as signal light failure ( 1 ), vehicle accident ( 2 ), bad weather ( 3 ), temporary traffic control ( 4 ) and the like, each corresponds to a specific inducement type, and each inducement type is mutually exclusive; represents the result variable set of the traffic event, which includes all possible traffic state results, such as road congestion ( 1 ), vehicle slow-down ( 2 ), traffic interruption ( 3 ), normal traffic ( 4 ) and the like, each corresponds to a specific traffic state, and each state type is mutually exclusive; represents the joint probability of the cause variable and the result variable occurring at the same time, which is calculated based on the historical traffic event data accumulated in the cloud, such as the past 12 months of accident-congestion association records, signal light failure-traffic interruption association records, for example, P( 2, 1 ) represents the probability of vehicle accident and road congestion occurring at the same time; represents the conditional probability of the result variable occurring under the premise of the cause variable occurring, and the calculation formula is , wherein is the cause variable The marginal probability of an isolated occurrence, for example This represents the probability of traffic congestion occurring on a given road segment, assuming a vehicle accident has occurred.

[0025] The causal monitoring unit is specifically as follows: Based on the causal relationship and the fused data of the data fusion unit, a digital twin road network mapping module using BIM and GIS fusion modeling is established. The spatiotemporal causal relationship (such as the speed of road congestion transmission and the scope of event impact) and the fused data are mapped to the twin scene in real time. The causal chain propagation path and degree of impact are displayed through a three-dimensional visualization interface. The digital twin road network is constructed by fusing BIM models and GIS maps. The twin scene is updated every 2 seconds based on the fused data to ensure synchronization with the actual road network status. The GraphCNN model is trained with 300 iterations and a batch size of 64 to predict road network traffic in the next 30-120 minutes. The LSTM model is trained with 200 iterations and a batch size of 32 to predict vehicle trajectories in the next 10-60 seconds. Both models use the Adam optimizer to improve prediction stability and accuracy. A dual-scale prediction engine for macro-level road network traffic and micro-level vehicle trajectory is designed. The macro-level layer uses a GraphCNN model (inputting traffic data fused at 15-minute intervals) to predict road network-level traffic, while the micro-level layer uses an LSTM model (inputting location data at 1-second intervals) to predict vehicle trajectories. The two interact in real time based on causal inference results. The macro-level results constrain the micro-level prediction range, and the micro-level results correct the macro-level model parameters. It integrates service-oriented output interfaces to directly generate executable commands such as signal timing adjustment and charging pile scheduling, and connects to traffic control systems and subsequent user service platforms.

[0026] Formula for comprehensive evaluation of two-scale prediction error: ; ; ; In the formula: This represents the overall error of the two-scale traffic prediction, used to assess prediction accuracy. A smaller value indicates a better match between the prediction and actual traffic conditions. When the value is ≤0.05, the prediction result is considered to meet the accuracy requirements for traffic control and user services; , The weighting coefficients for macroscopic forecast error and microscopic forecast error are respectively, satisfying the following conditions: + =1, adjusted according to the actual application scenario, such as the control scenario of urban main roads. = 0.6 (focus on macroscopic flow), under the scenario of park internal traffic = 0.7 (focus on microscopic trajectory), under the default scenario = 0.5 represents the macroscopic road network flow prediction error, used to quantify the prediction deviation of the macroscopic layer in the double-scale prediction engine, and the unit is consistent with the flow unit; represents the number of road segments of macroscopic road network flow prediction, for example, 5 key monitoring road segments of a main road are used for flow prediction, = 5, each road segment corresponds to a flow prediction value and an actual value; represents the prediction flow value of the th road segment, that is, the traffic flow of the road segment in the next 15 minutes calculated by the GraphCNN model, and the unit is vehicle / 15 minutes; represents the actual flow value of the th road segment, that is, the real traffic flow of the road segment in the corresponding 15 minutes collected by the roadside flow detector, and the unit is consistent with , used to compare and calculate the error with the prediction value; represents the microscopic vehicle trajectory prediction error, used to quantify the prediction deviation of the microscopic layer in the double-scale prediction engine, and the unit is meter; represents the time step number of microscopic vehicle trajectory prediction, for example, the trajectory of a vehicle in the next 30 seconds is predicted, = 30, each time step corresponds to a trajectory coordinate prediction value and an actual value; , respectively represent the predicted horizontal coordinate and the predicted vertical coordinate of the vehicle at the tth time step, that is, the planar coordinates of the vehicle at the time step calculated by the LSTM model, and the unit is meter, and the coordinate system is consistent with the space coordinates of the digital twin road network; , respectively represent the actual horizontal coordinate and the actual vertical coordinate of the vehicle at the tth time step, that is, the real planar coordinates of the vehicle at the time step collected by the vehicle GPS / Beidou positioning module, and the unit is consistent with , , used to compare and calculate the trajectory deviation with the predicted coordinates.

[0027] The service generation unit is specifically as follows: ​The causal relationship between the traffic event inferred by the causal inference unit and the user demand is correlated with the real-time traffic state of the causal monitoring unit to establish a semantic and causal dual-dimensional user portrait. The semantic portrait labels the semantic tags of the core demands of the user, such as commuting, medical treatment, and tourism. The causal portrait records the causal correlation between the user behavior and service experience, such as the correlation between route selection and congestion avoidance effect. The portrait data is synchronously correlated with the real-time traffic state monitored by the digital twin monitoring; The semantic user portrait includes a basic dimension, a demand dimension (demand tags such as commuting / medical treatment / tourism / shopping, travel time preference, and destination type), and a preference dimension (preferred route type and whether to avoid construction sections). The causal portrait records the correlation data between the user's selection of the recommended route and the travel time reduction rate and the congestion avoidance success rate, such as the probability of selecting the recommended route A corresponding to a 15% travel time reduction rate being 80%. The causal recommendation algorithm calculates the causal correlation strength between route adjustment and travel efficiency improvement (if the correlation strength is greater than or equal to 0.7, the route is recommended), dynamically adjusts the route recommendation weight in combination with the real-time traffic state, and ensures that the scheme adapts to the user demand. The causal recommendation algorithm is used to generate optimal travel schemes such as route planning, parking lot recommendation, and public transportation transfer suggestion for users in combination with fused data, causal inference conclusions, and real-time traffic states. At the same time, a service strategy dynamic adjustment module is deployed to update the recommendation strategy in real time to ensure service adaptability according to real-time causal factors such as sudden congestion and temporary regulation and changes in the traffic state monitored by the twin monitoring.

[0028] The optimization unit specifically includes the following: After receiving the service scheme of the service generation unit and putting it into application, the optimization unit collects full-link feedback data of the service scheme, including user ratings of the service, core causes of poor service effect recorded through semantic questionnaires, and data on the execution effect of the regulation and control instructions of the causal monitoring unit and the feedback data on the inference accuracy of the causal inference unit. Based on these multi-dimensional feedback, the platform algorithm and service strategy are dynamically iterated to continuously improve the system adaptability. The optimization effect evaluation index calculation method is as follows: service satisfaction = (4-5 evaluation times / total evaluation times) x 100%; monitoring accuracy = (number of consistent monitoring results and actual traffic states / total monitoring times) x 100%; response timeliness = average time consumption from data collection to service scheme generation. The clear calculation method ensures that the optimization effect is quantifiable and comparable. An optimization process of feedback collection, causal analysis, parameter optimization and service iteration is established, a causal analysis module is deployed to analyze and identify the causal relationship between service delay and user dissatisfaction, inference bias and monitoring error, etc.; a causal adaptive optimization algorithm is deployed to dynamically adjust the traffic monitoring model weight, data update frequency and service recommendation parameters according to the identified causal relationship, and the optimized parameters are synchronized and pushed to the front-end data fusion unit, causal inference unit and service generation unit. An optimization effect evaluation module is established to quantitatively evaluate the optimization effect by indexes including service satisfaction, monitoring accuracy, response timeliness, etc., to ensure the correct iteration direction and improve the service satisfaction and monitoring accuracy after optimization.

[0029] It should be noted that, in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0030] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives and variations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent transportation system-oriented vehicle networking big data real-time monitoring and user service optimization platform, characterized in that: The platform comprises: a preprocessing unit: preprocessing the collected data and extracting semantic features, and eliminating abnormal data conflicting with the scene semantics; a transmission unit: identifying the causal correlation degree between the preprocessed data, setting the transmission priority, using double encryption technology and dynamically adjusting the transmission rate; a data fusion unit: adjusting the data fusion weight according to the traffic event type, correcting the spatio-temporal offset of the data, monitoring the fusion effect in real time and adjusting the fusion strategy; a causal inference unit: based on the fusion data, identifying the core cause of the abnormal event, returning the inference result to the regional cloud, calculating the causal entropy change, establishing a scenario-based causal template library and updating it regularly; a causal monitoring unit: mapping the causal relationship and the fusion data to the twin scene, showing the causal chain propagation path and the influence degree; predicting the road network flow and vehicle trajectory and generating execution instructions, and mining the causal relationship between the traffic event and the user demand; a service generation unit: establishing a user portrait, generating an optimal travel plan for the user and updating the recommendation strategy in real time; an optimization unit: collecting feedback data, identifying causal relationships and optimizing parameters, and quantitatively evaluating the optimization effect. 2.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 1, wherein: The preprocessing unit is specifically as follows: Through the roadside terminal, the data collected by the vehicle terminal sensor, roadside radar and video detector are real-time denoised, unified in format, and the core semantic features of the target are synchronously extracted; A cross-node semantic feature encryption sharing mechanism is built using a federated learning framework, and each participating node only uploads local model gradient parameters; at the same time, a semantic consistency verification module is embedded, combined with a traffic scene knowledge graph containing traffic scene semantic rules, to eliminate abnormal data conflicting with the scene semantics. 3.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 2, wherein: The transmission unit is specifically as follows: Based on the network slicing technology, the causal correlation priority scheduling mechanism is designed to accept the preprocessed data and semantic annotation, and the edge node identifies the causal correlation degree between the data in real time, bundles and transmits the high-correlation data through a low-latency slice, and transmits the non-correlation data in batches; A dual encryption scheme of semantic label and data ontology is adopted, only authorized nodes can parse the semantic label; combined with the communication delay parameter, the transmission rate is dynamically adjusted to adapt to the communication needs of different traffic scenes. 4.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 3, wherein: The data fusion unit is specifically as follows: High-correlation semantic data from the transmission unit is received, a dynamic weight fusion model is established, and the data fusion weight is automatically adjusted according to the traffic event type; A cross-modal data alignment module is deployed, the time and space offset is corrected based on the timestamp and spatial coordinates; a data fusion quality evaluation module is integrated, the fusion effect is monitored in real time and the fusion strategy is dynamically adjusted through indicators including semantic consistency and key field missing rate, realizing the spatio-temporal alignment of cross-modal data. 5.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 4, wherein: The causal inference unit is specifically as follows: Based on the fusion data, a lightweight causal model is deployed on the roadside edge terminal to realize traffic anomaly causal positioning and identify the core cause of the abnormal event; An iterative optimization mechanism between the cloud and the edge is designed, the edge terminal returns the inference result to the regional cloud, the cloud updates the causal network topology combined with the full amount of fusion data, calculates the causal entropy change and corrects the edge model parameters, and feeds back the optimized model parameters to the edge terminal; A scenario-based causality template library is established, and scenario templates are pre-stored and periodically updated based on newly added data increments. 6.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 5, wherein: The causality monitoring unit is specifically as follows: Based on the causality and fusion data, a digital twin road network mapping module is established to map the spatio-temporal causality and fusion data to the twin scene in real time, and to show the causality chain propagation path and impact degree; A macroscopic road network flow and microscopic vehicle trajectory dual-scale prediction engine is designed, which interacts in real time based on the causality inference results; An integrated service output interface is designed to generate executable instructions and interface with the traffic control system and user service platform. 7.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 6, wherein: The service generation unit is specifically as follows: Based on the causality of traffic events and user demand and real-time traffic state, a semantic and causal dual-dimensional user portrait is established, and the portrait data is synchronously associated with the real-time traffic state of digital twin monitoring; A causal recommendation algorithm is used to generate the optimal travel plan in combination with fusion data, causality inference conclusion and real-time traffic state; meanwhile, a service strategy dynamic adjustment module is deployed to update the recommendation strategy in real time according to real-time causal factors and traffic state changes. 8.The intelligent transportation system oriented vehicle-to-everything big data real-time monitoring and user service optimization platform of claim 7, wherein: The optimization unit is specifically as follows: Collect full-link feedback data of service schemes, establish an optimization process of feedback collection, causal analysis, parameter optimization and service iteration, and deploy a causal analysis module to identify the causal relationship of feedback data; A causal adaptive optimization algorithm is used to dynamically adjust the traffic monitoring model weight, data update frequency and service recommendation parameters according to the causal relationship, and the optimized parameters are synchronously pushed to the front-end data fusion unit, causality inference unit and service generation unit. An optimization effect evaluation module is established to quantitatively evaluate the optimization effect through indicators including service satisfaction, monitoring accuracy and response timeliness.

Images