CN117636638A

CN117636638A - C-V2X-based vehicle congestion comprehensive analysis method and system

Info

Publication number: CN117636638A
Application number: CN202311599355.2A
Authority: CN
Inventors: 杨雷; 洪涛; 李金毅; 王汝文; 周正
Original assignee: Jiangsu Tianan Smart Science & Technology Co ltd
Current assignee: Jiangsu Tianan Smart Science & Technology Co ltd
Priority date: 2023-11-27
Filing date: 2023-11-27
Publication date: 2024-03-01

Abstract

The invention discloses a vehicle congestion comprehensive analysis method and system based on C-V2X, and belongs to the technical field of Internet of vehicles. According to the invention, real-time traffic data is collected and stored, real-time congestion analysis is performed, machine learning is adopted to predict a congested road section, analysis and prediction are performed on whether a road is congested or not comprehensively and three-dimensionally, and the analysis result and the prediction result are comprehensively displayed through a visual interface, so that the method has an important effect on smooth running of the road, also contributes to environmental protection, traffic management departments can perform omnibearing real-time monitoring on the road, and can compare historical data with indexes possibly causing congestion, the value of a large amount of data acquired based on a C-V2X scene is greatly mined, a basis is provided for subsequent analysis of root causes of congestion, and a great service support is provided for solving the congestion problem by using the method flexibly.

Description

C-V2X-based vehicle congestion comprehensive analysis method and system

Technical Field

The invention relates to a vehicle congestion comprehensive analysis method and system based on C-V2X, and belongs to the technical field of Internet of vehicles.

Background

With the rapid popularization of internet of vehicles and the increasing number of vehicles, urban traffic jam has become an increasingly challenging problem to be solved. The congestion of the vehicle not only causes waste of time and fuel, but also has negative effects on the environment, such as increasing exhaust emissions and air pollution. Therefore, it is important to find innovative ways to alleviate the problem of urban traffic congestion.

With the increasing development of C-V2X scenarios, internet of vehicles technology and internet of vehicles infrastructure provide us with a vast amount of data about traffic conditions, including vehicle position, speed, travel track, etc. These precious data resources provide new opportunities for solving vehicle congestion problems.

In current traffic congestion analysis, conventional techniques rely primarily on limited data sources, such as traffic cameras, road sensors, and the like. These data sources may not provide comprehensive traffic condition information, resulting in limited understanding and prediction of congestion conditions. And the data sources based on the C-V2X technology are more abundant and accurate. C-V2X (vehicle to vehicle) is an internet of vehicles communication technology that provides a wide range of data resources through communication between vehicles and infrastructure. Such data includes, but is not limited to, vehicle position, speed, travel track, ambient information, and the like. Compared to conventional data sources, data based on the C-V2X technique has the following advantages: real-time performance: the C-V2X technology can provide real-time data updating, so that the monitoring and analysis of traffic conditions are more accurate and timely. Traditional traffic visualization systems and congestion prediction methods often have delays and cannot provide timely traffic information. Coverage area: the C-V2X technology can realize wide communication among vehicles without being limited by sensors and devices. This means that the source of the data is more comprehensive, and can cover various areas and roads of the city, providing more comprehensive traffic condition information. Data precision: the data based on the C-V2X technology has higher precision, and can accurately capture the information such as the position, the speed and the like of the vehicle. Such fine data may help better analyze congestion causes revealing the mechanism by which traffic congestion forms.

Therefore, the project aims to develop a comprehensive analysis method for vehicle congestion based on C-V2X, and aims to deeply understand and solve the problem of traffic congestion.

By the method, traffic conditions can be monitored in real time, congestion hot spots can be accurately positioned, congestion causes can be deeply analyzed, real-time traffic suggestions can be provided, and urban planners can be assisted in formulating more effective traffic policies. The method is beneficial to reducing traffic jam, improving traffic efficiency, reducing carbon emission, improving travel experience of urban residents, and making an important step for promoting urban sustainable development.

The project background is based on the rapid development of the C-V2X scene, aims at coping with urban traffic jam challenges, and provides a more convenient, efficient and environment-friendly travel mode for people.

Disclosure of Invention

In order to solve the problems of inaccurate congestion analysis, difficult collection of a large amount of original data in prediction and visual display and the like existing at present, the invention provides a vehicle congestion comprehensive analysis method and system based on C-V2X, and the technical scheme is as follows:

the first object of the invention is to provide a vehicle congestion comprehensive analysis method based on C-V2X, comprising the following steps:

step 1: and collecting vehicle and traffic related data in real time through the V2X equipment, and uploading the data to the data storage unit.

Step 2: and calculating a congestion index based on the acquired data, and analyzing the congestion condition.

Step 3: based on the collected data, a machine learning model is utilized to predict future congestion conditions.

Step 4: and displaying the real-time congestion situation and the predicted future congestion situation through a visualization unit.

Optionally, the analyzing the congestion condition in the step 2 includes: setting a threshold value of the congestion index; the congestion index data are collected and monitored in real time; and comparing the real-time index data with a preset threshold value, and judging whether the road is congested.

Optionally, the congestion index includes: average speed of vehicle, density of traffic flow, length of line of intersection, and dead time of vehicle;

optionally, the step 3 includes:

step 31: collecting and preparing data, including: target variables and feature variables; wherein the target variable is data representing the degree of traffic congestion; the characteristic variables include: traffic flow, weather conditions, time;

step 32: carrying out pearson correlation analysis on the target variable and each characteristic variable, extracting characteristic variables or characteristic variable combinations with correlation, and forming a data set;

step 33: dividing the data set into a training set and a testing set, and training and testing a machine learning model;

step 34: and predicting the future congestion condition by using the trained machine learning model, and outputting a prediction result.

Optionally, the machine learning model is a recurrent neural network RNN.

A second object of the present invention is to provide a C-V2X-based vehicle congestion integrated analysis system, comprising: the traffic data analysis unit is used for analyzing traffic data, and the traffic data analysis unit is used for analyzing traffic data;

the traffic data acquisition unit is used for acquiring data related to vehicles and traffic in real time and storing the data into the data storage unit, and the traffic data analysis unit is used for analyzing congestion indexes based on the acquired traffic data; the congestion prediction unit predicts future congestion conditions based on a machine learning model; the visualization unit is used for displaying real-time monitoring of the crossing, congestion index conditions and data charts.

Optionally, the traffic data acquisition unit includes a V2X device, and the V2X device includes: MEC, RSU, camera, radar, learning machine.

Optionally, the congestion index analysis includes: vehicle average speed analysis, traffic density and road capacity analysis, intersection queuing length analysis and vehicle dead time analysis; the analysis process comprises the following steps: setting a threshold value of the congestion index; the congestion index data are collected and monitored in real time; and comparing the real-time index data with a preset threshold value, and judging whether the road is congested.

Optionally, the congestion prediction unit predicts the future congestion condition by adopting a machine learning model.

Optionally, the visualization unit includes: the system comprises a real-time monitoring module, a congestion analysis display module, a congestion prediction display module and a historical congestion data display module.

The invention has the beneficial effects that:

according to the invention, through real-time traffic data collection and storage, vehicle average speed analysis, traffic flow density and road capacity analysis, intersection queuing length analysis and vehicle dead time analysis, machine learning predicts a congested road section, a visual interface is established, whether the road is congested or not is comprehensively and stereoscopically analyzed and predicted, and analysis results and prediction results are comprehensively displayed through the visual interface, so that the method has an important effect on smooth running of the road and also greatly contributes to environmental protection.

By means of the analysis and prediction results obtained by the method, traffic management departments can conduct omnibearing real-time monitoring on roads, historical data comparison can be conducted on indexes possibly causing congestion, the value of a large amount of data obtained based on a C-V2X scene is greatly mined, a basis is provided for subsequent analysis of root causes of congestion, and a great business support is provided for solving the congestion problem by using the method flexibly.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of the C-V2X based vehicle congestion analysis-by-synthesis method of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings.

Embodiment one:

the embodiment provides a vehicle congestion comprehensive analysis method based on C-V2X, which comprises the following steps:

V2X devices are installed in areas such as roads, intersections, and around cities, such as: MEC, RSU, camera, radar, learning machine, etc., which can collect data related to vehicles and traffic.

Real-time traffic data, such as vehicle speed, location, traffic density, etc., from the V2X devices is collected and stored. By collecting real-time traffic data, the traffic flow and traffic jam condition on the road can be monitored, traffic management departments can be helped to better know the traffic condition, and storing a large amount of traffic data is helpful for historical traffic pattern analysis, so that more effective traffic plans are formulated, including road reconstruction and improvement of public traffic systems, and a foundation is provided for subsequent data analysis.

The congestion index includes: the average speed of the vehicle, the density of the traffic flow, the queuing length of the crossing and the dead time of the vehicle are analyzed as follows:

first, vehicle average speed analysis focuses on the average speed of a vehicle in congestion, typically expressed in miles per hour or kilometers per hour, which provides an impact of congestion on vehicle travel speed, and can be used to determine the degree of congestion and the smoothness of traffic by the following methods:

(1) Setting a speed threshold: for average speed analysis, it is first necessary to determine an appropriate average speed threshold. This is typically a numerical value associated with a particular road, city prescription, or historical traffic data. For example, the average speed threshold may be set to a reasonable speed limit for a certain road, such as an average speed of 10km/h for a certain vehicle per minute. And specifies a congestion ratio threshold, such as 80%, which generally refers to the ratio of the current number of low speed vehicles to the number of all vehicles.

(2) Real-time speed monitoring: the current lane is monitored by the vehicle speed data in the time series database.

(3) Threshold comparison: for each vehicle, the system will compare its current average speed over the minute to a predetermined average speed threshold. If the average speed of the vehicle is below the average speed threshold and the vehicle below the average speed threshold exceeds the congestion ratio threshold, the system will mark the lane as likely to experience congestion, the congestion indicator being the average speed of the vehicle.

(4) Persistence of comparison results: the compared business results are persisted to disk and may be stored using a database such as mysql.

Second, traffic density and road capacity analysis are important indicators in congestion analysis, which provide important information about the operating efficiency and congestion status of a road. Traffic density refers to the ratio of the number of vehicles on a certain section of road to the length of the road, typically expressed in vehicles/minute, and traffic density analysis is used to measure the degree of congestion of vehicles on the road. A higher traffic density generally indicates that the vehicles on the road are very congested, possibly resulting in congestion, and such analysis helps traffic authorities to know when and where the road is most congested to take appropriate action; road capacity refers to the maximum number of vehicles that a section of road can accommodate under ideal conditions, typically expressed in vehicles per minute, and is an important reference point for determining the potential operating capacity of the road, which may lead to congestion if the actual traffic flow is higher than the road capacity, and this analysis helps to assess whether the road infrastructure is adequate to meet traffic demands and when it is necessary to build up or improve the road. Comprehensively considering traffic density and road capacity analysis can help determine whether a road is operating normally, approaching saturation or has been saturated, and if the traffic density approaches or exceeds the road capacity, this may be a primary indicator of congestion, indicating that the road is operating at full load, this information is important for traffic planning and congestion management, and the traffic density and road capacity analysis flow is as follows:

(1) Setting a vehicle density threshold: for vehicle density analysis, it is first necessary to determine an appropriate vehicle density threshold. This is typically a numerical value associated with a particular road, city prescription, or historical traffic data. For example, the vehicle density threshold may be set to a reasonable vehicle density limit for a certain road, such as vehicle/minute. And defines a flow/capacity ratio threshold, such as 80%, which generally refers to the current ratio of road traffic to road capacity.

(2) Real-time vehicle density monitoring: and monitoring the current lane through the vehicle density data in the time sequence database and the recorded road GIS data.

(3) Threshold comparison: for each vehicle, the system compares its current vehicle density in minutes to a predetermined vehicle density threshold. If the vehicle density of the vehicle is above the vehicle density threshold and the vehicle above the vehicle density threshold exceeds the flow/capacity ratio threshold, the system will mark the lane as likely to experience congestion, the congestion indicator being the vehicle density.

Third, the intersection queue length analysis focuses on the length of vehicles that are queued at or at an intersection. It generally involves observing the actual distance of vehicle queuing to determine traffic conditions at a particular intersection, which can help traffic authorities to learn which intersections are congested, and more focus on intersections and specific locations of intersections to determine the number and length of vehicles queued, which is important for traffic control and improvement in planning a particular intersection, and to provide important information for determining congested hotspots, the intersection queuing length analysis flow is as follows:

(1) Setting an intersection queuing length threshold value: to perform the intersection queuing length analysis, an appropriate intersection queuing length threshold is first determined. This is typically a numerical value associated with a particular road, city prescription, or historical traffic data. For example, the intersection queue length threshold may be set to a reasonable intersection queue length limit for a road, such as 10m.

(2) Real-time queuing length monitoring: and monitoring the current intersection through the stored intersection queuing length information in the time sequence database.

(3) Threshold comparison: for each intersection, the system compares its current intersection queue length to a predetermined intersection queue length threshold value. If the vehicle's crossing queue length is above the crossing queue length threshold, the system will mark the lane as likely to encounter congestion, the congestion indicator being the crossing queue length.

Fourth, vehicle dead time analysis is concerned with the dead time of the vehicle on the road, i.e., the time the vehicle stops due to waiting when it is congested. This may help determine the intensity and duration of traffic congestion, as well as the impact on traffic flow. More attention is paid to the influence of congestion, which tells us how long the waiting time of the vehicle in congestion is, which is helpful for measuring the severity of traffic congestion and influence on drivers, and the vehicle dead time analysis flow is as follows:

(1) Setting a vehicle dead time threshold: in order to perform a vehicle dead time analysis, it is first necessary to determine an appropriate vehicle dead time threshold. This is typically a numerical value associated with a particular road, city prescription, or historical traffic data. For example, the vehicle dead time threshold may be set to a reasonable time limit for a certain road, such as 2 minutes. And specifies a congestion ratio threshold, such as 80%, which generally refers to the proportion of the number of vehicles currently above the dead time threshold to all of the number of vehicles.

(2) Real-time vehicle dead time monitoring: the current lane is monitored by the vehicle speed data in the time series database.

(3) Threshold comparison: for each vehicle, the system compares its vehicle dead time to a predetermined average speed threshold. If the dead time of the vehicle is above the vehicle dead time threshold and the vehicle above the vehicle dead time threshold exceeds the congestion ratio threshold, the system will mark the lane as likely to experience congestion, the congestion indicator being vehicle dead time.

By using machine learning to predict the congestion road section, the manager or the public can predict which roads are likely to be congested earlier, and measures such as route recommendation and traffic signal adjustment can be taken in advance for traffic departments to relieve congestion.

The present embodiment employs RNN (recurrent neural network) for congestion prediction.

Embodiment two:

the embodiment provides a vehicle congestion comprehensive analysis system based on C-V2X, which comprises: the traffic data analysis unit is used for analyzing traffic data, and the traffic data analysis unit is used for analyzing traffic data; the traffic data acquisition unit is used for acquiring vehicles and traffic related data in real time and storing the data into the data storage unit, and the traffic data analysis unit is used for analyzing indexes such as vehicle speed, traffic flow density, intersection queuing length, vehicle dead time and the like based on the acquired traffic data; the congestion prediction unit predicts the future congestion condition based on a machine learning model; the visualization unit is used for displaying intersection real-time monitoring, congestion index conditions, data line diagrams and the like. The following describes the parts of the system of this embodiment in detail:

1. traffic data acquisition unit

The infrastructure for collecting data includes V2X devices such as MECs, RSUs, cameras, radars, learning machines, etc. installed in areas around roads, intersections, cities, etc. that can collect real-time data related to vehicles and traffic.

After the V2X devices collect data, the data are transmitted to a central server or a cloud platform through wireless communication or a wired network, so that the real-time performance of the data is ensured, and the data can be rapidly analyzed.

Data broadcast to the cloud platform by the RSU is integrated with higher confidence data that may be collected by other sources, such as GPS data: many vehicles are equipped with GPS devices that can provide real-time location data of the vehicle that can be used to track the movement and speed of the vehicle on the road; sensor data: traffic flow sensors may detect the number and speed of vehicles, which may help determine the density of traffic on a road. Information from different data sources is integrated together to create a comprehensive real-time traffic data set. This may include information on vehicle position, speed, road conditions, signal light status, etc.

Typically, these data are large volumes of time series data that are stored in a time series database such as TDengine for later analysis and querying.

2. Traffic data analysis unit

The traffic data analysis unit is mainly used for analyzing the average speed of the vehicle, the density of the traffic flow and the capacity of the road, the queuing length of the crossing and the dead time of the vehicle.

The vehicle average speed analysis flow is as follows: (1) setting a speed threshold: for average speed analysis, it is first necessary to determine an appropriate average speed threshold. This is typically a numerical value associated with a particular road, city prescription, or historical traffic data. For example, you can set the average speed threshold to a reasonable speed limit for a certain road, such as an average speed of 10km/h for a certain vehicle per minute. And specifies a congestion ratio threshold, such as 80%, which generally refers to the ratio of the current number of low speed vehicles to the number of all vehicles.

The flow density and road capacity analysis flow is as follows:

The analysis flow of the crossing queuing length is as follows:

The vehicle dead time analysis flow is as follows:

3. Congestion prediction unit

The congestion prediction unit predicts future congestion conditions based on the collected or analyzed traffic parameters by using a machine learning model, and comprises the following steps:

1. feature extraction: correlation analysis is performed on features that may affect congestion:

(1) Performing data preparation

First, data is collected and prepared, including: target variables and characteristic variables, which often depend on the data of the traffic data acquisition unit.

Wherein the target variable is data representing the degree of traffic congestion;

the characteristic variables include:

traffic flow: in addition to conventional traffic flow data, it can be further subdivided into traffic flows of individual road segments, traffic flows of different lanes, etc. This enables the influence of traffic flow on the degree of congestion to be captured more finely.

Weather conditions: in addition to common weather factors such as rainfall, temperature, etc., other weather related variables such as solar duration, wind speed, visibility, etc. may also be considered. These factors may have an impact on traffic congestion.

Time: in addition to specific hour and date information, finer granularity time variations, such as early and late peak hours, weekdays and non-weekdays, etc., may be considered to better distinguish traffic congestion conditions for different time periods.

Road attribute: road properties such as road class, road width, number of turns, intersection density, etc. are also important factors to consider. These attributes can help to understand in depth the congestion situation of different road segments.

(2) Spearman correlation analysis

The following are examples of using Spearman correlation analysis to verify the relationship between weather and congestion:

data preparation: the collected data is arranged into a form of a table, and each sample is ensured to have corresponding weather and congestion values.

The Spearman correlation coefficient calculated using Python language may use the Spearman function in the scipy. First, it is ensured that the scipy library has been installed. If not, the following commands can be used for installation: (pip install scipy) then importing the required modules and functions: (from scipy. Stats import spin) two variables are prepared that possess relevant data, such as weather and congestion. (weather= [25,20,30,22,28] traffic= [5,8,3,7,6 ]), spearman correlation coefficients were calculated using the Spearman function: this will return a tuple containing the correlation coefficient and the p value (corridation, p_value=spin (weather)).

Interpretation results: from the calculated Spearman correlation coefficient, the strength and direction of the relationship between weather and congestion can be explained. If the correlation coefficient is close to 1 or-1, it indicates that there is a strong correlation between weather and congestion. If the correlation coefficient is close to 0, no obvious correlation exists between the correlation coefficient and the correlation coefficient, correlation of all the features or feature combinations is calculated, and the features or feature combinations with the correlation are reserved for subsequent model training.

(3) Pierson correlation analysis

The following are examples of using pearson correlation analysis to verify the relationship between weather and congestion:

for the prepared data, a pearson correlation analysis was performed using the numpy library of python, assuming that the data contains two arrays, one representing the data of the degree of congestion (referred to as "congestion"), and the other representing weather parameters such as rainfall (referred to as "rain fall").

First, a bin (import numpy as np) is imported, then an average value of congestion and rainfall is calculated (mean_congestion=np.mean (congestion) mean_rain=np.mean (rain)), then a difference value between each data point is calculated (displacement_congestion=congestion-mean_ congestion deviation _rain=rain_rain=rain_rain), and a correlation coefficient is calculated (correlation=np.sum (displacement_congestion)

The value of displacement_rainfall)/(np.sqrt (np.sum (displacement_diagnostic) 2) np.sum (displacement_rainfall) 2) will be between-1 and 1, indicating a linear relationship between them.

If "correlation" is a positive number, indicating their positive correlation, and if it is a negative number, indicating their negative correlation, 0 indicates no linear correlation, the correlation for all features or feature combinations is calculated, and the features or feature combinations with correlation are retained for subsequent model training.

(4) Iterative iteration

And for the extracted features or feature combinations with correlation, the features with low correlation rank can be considered for removal for repeated verification in the subsequent model training process.

2. Data set partitioning

The raw data is divided into training and testing sets, typically 70% for training and 30% for testing.

3. Selecting an appropriate machine learning model

Because of the characteristics of large quantity, short time sequence and the like of the original data, RNN (cyclic neural network) and LSTM (long and short time memory network) are initially selected, the prediction results of the two learning models are not greatly different in the actual test process, but the RNN shows faster execution efficiency and smaller memory occupation, and finally the RNN is selected as a subsequent training algorithm.

4. Model training using RNN

(1) Model construction

Selecting an activation function: using the ReLU activation function, the input data avoids using a negative number in order to avoid gradients being small, taking into account the properties of the ReLU.

Input sequence processing: consider how input sequence data is processed. You can use a sliding window or sequence lag data to capture a pattern of time series. In addition, it may be considered to package consecutive time-step data into an input sequence to better exploit time dependencies.

Regularization and batch normalization: adding regularization techniques, such as L2 regularization or Dropout, helps to prevent overfitting. Batch normalization can improve training stability and accelerate convergence.

Determining the layer number of the model: the method has the advantages that the number of layers is determined to be dependent on original data, a cross experiment is needed, and for a simple time series task, a single RNN layer can solve the problem, the cross verification is carried out by adopting 1-2 RNN layers, and 2 RNNs have more advantages for time scales of multiple levels.

Super-parameter adjustment: defining the number of layers, the number of hidden units, the length of an input sequence, the learning rate, the batch processing size and the training iteration times. The best model configuration is found by trying different combinations of hyper-parameters.

Determining a visual mode: visual interpretation was performed using the SHAP values and the Attention mechanism. SHAP values can be used to interpret the prediction results of the RNN model, particularly when the RNN is used for classification or regression tasks of the sequence data. SHAP values can tell you how much each time step or feature contributes to the model output, providing a global interpretation. The Attention mechanism can visualize Attention weights to explain the prediction of the model, the Attention weights can display the Attention degree of the model on different time steps or sequence elements, and the Attention weights are helpful for understanding the decision basis of the model.

(2) Training and tuning

The model is trained using the training set data. Historical data is input into the model at each time step, congestion conditions are targeted, and model parameters are updated through the back propagation and optimizer.

The verification set data is used to fine tune the super parameters, such as learning rate and batch size, to achieve optimal performance.

5. Model evaluation

And evaluating whether the model prediction model is successful according to the MSE, MAE and accuracy values. Mean square error (MSE-Mean Squared Error), which is a common evaluation index of regression problems, and is calculated by the following method: for each sample, the difference between the model's predicted value and the actual label is calculated, then the squares of these differences are summed, and finally the average is taken, the smaller the MSE, the closer the model's predicted value and the actual value are represented. A smaller MSE generally indicates that the model performs better on regression tasks; mean absolute error (MAE-Mean Absolute Error) mean absolute error was also used for regression problem assessment, calculation method: for each sample, the absolute differences between the model's predicted and actual labels are calculated, then summed, and finally averaged, and MAE measures the average absolute deviation between the model's predicted and actual values. Unlike MSE, MAE is not affected by outliers and is therefore more robust to outliers; accuracy (Accuracy) Accuracy is typically used for the assessment of classification problems, accuracy represents the proportion of the model that is correctly classified. However, for unbalanced classification problems, accuracy may not be a suitable indicator as it may overestimate model performance.

6. Deployment and monitoring: the model is deployed to the production environment, predicted in real time, and performance is monitored periodically to ensure effectiveness.

(1) Preparing a model: prior to deployment, the model was ensured to be trained and evaluated and perform well on the test data.

(2) Preparing environment: selecting a proper deployment environment, such as a cloud server, a containerized environment or edge equipment, deploying needed software, libraries and dependent items, and configuring proper hardware resources.

(3) Integrated into the production system: the model is integrated into the production system for real-time prediction. This may involve creating an API, micro-service, or other means to provide predictive functionality for the model.

(4) Monitoring and maintaining: tracking the accuracy, response time, etc. of the models to ensure that they meet expectations in production; checking the quality of the input data to ensure that it is consistent with the data distribution during model training; and monitoring whether the model drifts, namely whether the performance of the model gradually decreases.

4. And a visualization unit:

(1) And the real-time monitoring module is used for: the real-time monitoring is carried out by taking the intersection as a unit, and the track of each object at the intersection, the basic data provided by the C-V2X scene and the real-time picture of the camera are required to be seen.

(2) And a congestion analysis display module: showing the relative index condition of the congestion analysis, whether the intersection is congested, and which index has a problem.

(3) Congestion prediction presentation module: intersection prediction conditions, such as whether congestion, feature correlation, and weight ratio, are expected for several days in the future.

(4) Historical congestion data display module: and comparing the historical data of the congestion analysis index with the current data and displaying the historical data by using a line graph.

The visualization unit also provides a data export interface supporting the export of analysis reports for a single intersection or multiple intersections, wherein the analysis report content comprises congestion analysis results, prediction results, historical data comparison results and related index data in the analysis process.

The visualization unit is realized based on front-end and back-end technologies.

The back-end personnel uses the back-end server to develop interfaces, corresponding interfaces are developed for corresponding requirements, all displays in the current requirements can be obtained in the database, corresponding data can be obtained from the database in a sql statement writing mode and provided for the page portal to be obtained, and different portals are used by different modules.

And the front-end personnel develop the page according to the requirement, and the data are acquired and displayed from the back-end server in a http sending mode.

5. Data storage unit

The data storage unit is a database deployed on the central server or the cloud platform, and comprises the time sequence database, mysql and the like, and is used for storing service data for analysis and display.

Some steps in the embodiments of the present invention may be implemented by using software, and the corresponding software program may be stored in a readable storage medium, such as an optical disc or a hard disk.

The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims

1. A comprehensive analysis method for vehicle congestion based on C-V2X, the method comprising:

step 1: the method comprises the steps of collecting vehicle and traffic related data in real time through V2X equipment, and uploading the data to a data storage unit;

step 2: calculating a congestion index based on the acquired data, and analyzing the congestion condition;

step 3: based on the collected data, predicting future congestion by using a machine learning model;

2. The method according to claim 1, wherein the analyzing the congestion condition in the step 2 includes: setting a threshold value of the congestion index; the congestion index data are collected and monitored in real time; and comparing the real-time index data with a preset threshold value, and judging whether the road is congested.

3. The method of claim 1, wherein the congestion indicator comprises: average speed of vehicle, density of traffic, length of line of intersection, and dead time of vehicle.

4. The method according to claim 1, wherein the step 3 comprises:

5. The method of claim 1, wherein the machine learning model is a recurrent neural network RNN.

6. A C-V2X based vehicle congestion integrated analysis system, the system comprising: the traffic data analysis unit is used for analyzing traffic data, and the traffic data analysis unit is used for analyzing traffic data;

7. The system of claim 6, wherein the traffic data acquisition unit comprises a V2X device, the V2X device comprising: MEC, RSU, camera, radar, learning machine.

8. The system of claim 6, wherein the congestion index analysis comprises: vehicle average speed analysis, traffic density and road capacity analysis, intersection queuing length analysis and vehicle dead time analysis; the analysis process comprises the following steps: setting a threshold value of the congestion index; the congestion index data are collected and monitored in real time; and comparing the real-time index data with a preset threshold value, and judging whether the road is congested.

9. The system of claim 6, wherein the congestion prediction unit predicts future congestion conditions using a machine learning model.

10. The system of claim 6, wherein the visualization unit comprises: the system comprises a real-time monitoring module, a congestion analysis display module, a congestion prediction display module and a historical congestion data display module.