A traffic prediction modeling and online simulation method for vehicle-road collaborative control

Abstract

The present invention belongs to the field of road monitoring and simulation technology, and in particular relates to a traffic prediction modeling and online simulation method for vehicle-road collaborative control. In view of the current technical problems of untimely data updates, low model accuracy leading to low travel efficiency, high investment costs, inability to timely monitor and predict traffic flow, low flexibility and openness of the simulation system, and large data errors, the following solution is proposed, comprising the following steps: S1: measuring traffic and establishing a model; the purpose of the present invention is to monitor and predict incoming and outgoing vehicles and road conditions by adopting a traffic prediction model and an online simulation method for vehicle-road collaborative control, timely update road information, improve the real-time nature of data monitoring, improve the accuracy of calculation results, reduce data errors, accurately describe the model, and openly analyze traffic, thereby enhancing the practicality and convenience of model application, and the simulation system is highly flexible and open.

Description

Technical Field

[0001] The present invention belongs to the technical field of road monitoring and simulation, and in particular relates to a traffic prediction modeling and online simulation method for vehicle-road collaborative control. Background Art

[0002] With the continuous development of mobile Internet technology and Internet of Things technology, the information barriers between smart parking platforms, parking guidance systems and urban parking spaces will gradually be broken down, and the rapid application of various car-related scenarios such as parking payment, refueling, car washing, shopping, etc. will continue to increase, truly building a smart urban transportation system.

[0003] Vehicle-infrastructure collaboration (V2X) is one of the foundations of smart transportation. Through technologies like 5G and C-V2X, roads, vehicles, and related traffic elements are effectively integrated, enabling real-time data exchange between vehicles, roads, people, and networks. This plays a crucial role in the development of smart transportation. Currently, there are over 40 V2X demonstration zones across China, and testing scenarios are gradually evolving from closed to open, from single to diversified, and from regional to urban and even highway scenarios. Traffic prediction and online simulation play a significant role in this. Therefore, we propose a traffic prediction modeling and online simulation method for V2X control. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of the existing technology such as untimely data updates, low model accuracy leading to low travel efficiency, high investment costs, inability to timely monitor and predict traffic flow, low flexibility and openness of the simulation system, and large data errors. A traffic prediction modeling and online simulation method for vehicle-road collaborative control is proposed.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A traffic prediction modeling and online simulation method for vehicle-road cooperative control includes the following steps:

[0007] S1: Measure traffic and build a model;

[0008] S2: Establish a traffic simulation scenario based on the actual scenario and initialize the traffic simulation scenario;

[0009] S3: Extract road, vehicle and signal information based on the plug-in;

[0010] S4: Calculate the optimal solution for vehicle speed and traffic light matching based on the real-time data model;

[0011] S5: input the obtained real-time data into the simulation software;

[0012] S6: Analyze the data generated by the simulation to verify the feasibility of the algorithm.

[0013] Preferably, in S1, traffic is predicted and a model is established. The prediction model consists of an input layer, a hidden layer and an output layer. The input features are analyzed and finally output to the LSTM layer to sample highway traffic flow photos for the final input layer of the traffic flow, and the operation process is predicted. The CNN-LSTM neural network prediction model mainly adopts convolution layer, pooling layer, full connection layer and LSTM, wherein the convolution layer corresponds to the pooling layer. After the pooling layer, all the nodes of the pooling layer are expanded into feature prediction analysis. When performing modeling and prediction, space and time must be taken as a whole, and LSTM is used as a time response model for traffic flow prediction.

[0014] Preferably, in said S2, the process of vehicle-road collaborative simulation includes the following three key steps: full-time and space traffic information extraction technology, microscopic traffic flow simulation model optimization technology and multi-intersection signal light collaborative control technology, to obtain vehicle position information, obtain regional traffic flow information, process and analyze this information, and adopt collaborative control algorithm to realize collaborative control of single intersection and trunk signal, reduce vehicle travel time and average delay, each simulation step will generate simulation data, the simulation step is set at 33-50ms, after the simulation data is acquired, the data is converted, mathematical methods are used to abstractly create simulation scenes for urban road traffic systems, the simulation scenes are initialized according to the investigated traffic flow data and empirical data, and the standard expansion of the programmer is used to generate simulation data. Fill in the code and define the functional interface function. When the simulation starts, the system will automatically enter the trigger function. When a vehicle drives out of the OD matrix, the system will enter the function and enter the application through TIMESTEP to obtain vehicle information. The standard mandatory code defines the mandatory execution function, and rewrites its internal default safety distance model through the speed of the front vehicle and the speed of the vehicle. The information acquisition standard code defines the function of acquiring information. Through the mutual use of these functions, various information can be obtained during the simulation runtime. The signal light control information is obtained through the acquisition information standard code function. The information standard code is set, and the information that can be updated in real time in the simulation road network is set. The vehicle speed information function is set. By writing plug-ins, the real-time data obtained by the collection code is set to set the simulation scene data.

[0015] Preferably, in said S2, traffic flow data is extracted in real time, the freely generated traffic flow data is integrated, and the simulated road network is controlled by a specific control strategy. By setting the required parameters, a simulation saved data file is obtained at the end of the final simulation. The simulation parameters include: road network initialization time, time length of data collection, time interval for data collection, and data type to be collected. The collection time interval is set to 10-15 minutes. The collected data types include vehicle speed, vehicle ID, and vehicle type information. These data are used to evaluate traffic quality and road network initialization time.

[0016] Preferably, in the S4, trunk adaptive coordinated control is adopted, the traffic system status is continuously measured, and trunk signal control based on phase difference fuzzy adjustment is adopted. For large intersections, each phase time is distributed to 7 according to the "phase equal saturation" principle, and the import flow of large intersections is collected in real time. Each intersection calculates the signal timing according to the flow in four directions. The central control unit selects the intersection that needs to be coordinated according to the road information, and then realizes the green wave band according to the phase difference, for urban intersection signals using a four-phase control method.

[0017] Preferably, in the S5, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The modeler provides three basic operation processes, including model establishment, traffic simulation and statistical data output. In the modeler, the simulation road network and the nodes of the road network are drawn, the types of roads, the number of lanes of the roads, the speed limits of the roads, etc., the turning of the roads, the allocation of vehicle travel areas, the number of vehicles traveling in the road network is adjusted by changing the number of vehicles traveling in the area, and the information release is added to the modeler to display vehicle speed limit information and path selection information, and control the traffic lights at the intersection. The batch processor is a tool for the core model to execute batch traffic simulation, which allows Users can simulate the road network without displaying visual traffic. It supports parallel computing and can simulate multiple road networks and scenarios at the same time, reducing user downtime and accelerating model development. It is used for sensitivity and option testing. It allows users to perform simulation calculations in batch mode. The programmer provides a comprehensive application programming interface (API) for extension plug-in development. Through the use of the API interface, key simulation data can be extracted in real time, such as vehicle speed, lane, road, acceleration, and the number of lanes on the road, road speed limit, whether the node has signal control, and the operating status of a certain phase of signal control. Some path induction development can be done through the API, and different induction methods can be given according to the actual traffic flow.

[0018] Preferably, in the S5, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The emission monitor uses the API module developed by the programmer to collect exhaust emission data directly from the simulation system. During the simulation process, it can calculate the amount of exhaust emissions of all vehicles in the simulated traffic network and display it visually, and the user can define the exhaust emission level data himself; the analyzer is used to display the simulation results output by the modeler or processor, and is used to process multiple groups of data. The output evaluation indicators include: vehicle delay, vehicle travel time, vehicle parking time, road traffic flow, average number of stops, maximum vehicle queue length, and average vehicle speed. Users can customize these output evaluation indicators according to their needs. It has a special Excel wizard for filtering data and outputting statistical indicators and comparison results of the simulation at different stages. The matrix estimator is used to estimate the OD matrix at the micro level, and is fully compatible with the OD matrix, model and simulated vehicle path selection. It provides an open and visual interface, allowing users to add their own prior knowledge and experience to the system kernel of the estimator, and provides an API interface so that researchers can customize the estimation program.

[0019] Preferably, in said S6, the module analyzes the log file generated by the simulation to evaluate the feasibility of the control algorithm, and conducts two experiments by injecting the following model of this article and the traditional safety distance control model, with a warm-up time of 3-6 minutes, data collection of 54-57 minutes, and a data collection interval of 5-8 minutes. Vehicle delay, parking time and driving time are used as evaluation indicators for comparison. In different simulation time periods, the simulation data based on the vehicle following model of this article is better than the simulation data of the traditional minimum safety distance following model.

[0020] The beneficial effects of the present invention are:

[0021] 1. Use traffic prediction models to monitor regional vehicle traffic and flow measurement, and predict road conditions at the same time. This eliminates the need for manual monitoring, reduces labor costs, and alleviates traffic congestion. Timely monitoring improves the real-time nature of data and greatly improves travel efficiency.

[0022] 2. The online simulation method of vehicle-road cooperative control is adopted. The flexibility and softness of the model mechanism, the description of the changing rules and interaction relationships of the basic elements in the system are closely consistent with the actual operation process of the system, and the impact of various road and traffic conditions is accurately and flexibly reflected. The true model's description of the changing rules and interaction relationships of the basic elements in the system is closely consistent with the actual operation process of the system, effectively improving the accuracy of the calculation results and the real-time nature of the monitoring data.

[0023] The purpose of the present invention is to monitor and predict the information of vehicles and road conditions, update road information in a timely manner, improve the real-time nature of data monitoring, improve the accuracy of calculation results, reduce data errors, accurately and flexibly describe the model, and analyze traffic in an open manner, thereby enhancing the practicality and convenience of model application. The simulation system is highly flexible and open. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of a traffic prediction modeling and online simulation method for vehicle-road collaborative control proposed in the present invention. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0026] Example 1

[0027] Reference Figure 1 A traffic prediction modeling and online simulation method for vehicle-road cooperative control includes the following steps:

[0028] S1: Measure traffic and build a model;

[0029] S2: Establish a traffic simulation scenario based on the actual scenario and initialize the traffic simulation scenario;

[0030] S3: Extract road, vehicle and signal information based on the plug-in;

[0031] S4: Calculate the optimal solution for vehicle speed and traffic light matching based on the real-time data model;

[0032] S5: input the obtained real-time data into the simulation software;

[0033] S6: Analyze the data generated by the simulation to verify the feasibility of the algorithm.

[0034] In this embodiment, traffic is predicted and a model is established. The prediction model consists of an input layer, a hidden layer, and an output layer. The input features are analyzed and finally output to the LSTM layer for the final input layer of the traffic flow to sample highway traffic flow photos and predict the operation process. The CNN-LSTM neural network prediction model mainly adopts convolutional layer, pooling layer, fully connected layer and LSTM, wherein the convolutional layer corresponds to the pooling layer. After the pooling layer, all the nodes of the pooling layer are expanded into feature prediction analysis. When performing modeling and prediction, space and time must be taken as a whole, and LSTM is used as a time response model for traffic flow prediction.

[0035] In this embodiment, the process of vehicle-road collaborative simulation includes the following three key steps: full-time and space traffic information extraction technology, microscopic traffic flow simulation model optimization technology and multi-intersection signal light collaborative control technology, to obtain vehicle position information, obtain regional traffic flow information, process and analyze this information, and use collaborative control algorithms to achieve collaborative control of single intersection and trunk signals to reduce vehicle travel time and average delay. Each simulation step will generate simulation data, and the simulation step is set at 33ms. After the simulation data is obtained, the data is converted, and mathematical methods are used to abstract the urban road traffic system to create a simulation scene. The simulation scene is initialized according to the investigated traffic flow data and empirical data, and the standard expansion code provided by the programmer is used. , define the functional interface function. When the simulation starts, the system will automatically enter the trigger function. When a vehicle leaves the OD matrix, the system will enter the function and enter the application through TIMESTEP to obtain vehicle information. The standard mandatory code defines the mandatory execution function. The internal default safety distance model is rewritten by the speed of the preceding vehicle and the speed of the vehicle. The information acquisition standard code defines the function for acquiring information. Through the mutual use of these functions, various information can be obtained during the simulation run. The signal light control information is obtained through the acquisition information standard code function. The information standard code is set. The information that can be updated in real time in the simulation road network is set. The vehicle speed information function is set. By writing plug-ins, the real-time data obtained by the collection code is set to set the simulation scene data.

[0036] In this embodiment, traffic flow data is extracted in real time, the freely generated traffic flow data is integrated, and the simulated road network is controlled by a specific control strategy. By setting the required parameters, a data file saved by the simulation is obtained at the end of the simulation. The simulation parameters include: road network initialization time, data collection time length, data collection time interval and data type to be collected. The collection time interval is set to 10 minutes. The collected data types include vehicle speed, vehicle ID, and vehicle type information. These data are used to evaluate traffic quality and road network initialization time.

[0037] In this embodiment, trunk adaptive coordinated control is adopted, the traffic system status is continuously measured, and trunk signal control based on phase difference fuzzy adjustment is adopted. For large intersections, the time of each phase is allocated according to the "phase equal saturation" principle, and the import flow of large intersections is collected in real time. Each intersection calculates the signal timing according to the flow in four directions. The central control unit selects the intersection that needs coordinated control based on the road information, and then realizes the green wave band based on the phase difference, which is aimed at urban intersection signals using a four-phase control method.

[0038] In this embodiment, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The modeler provides three basic operation processes, including model establishment, traffic simulation and statistical data output. In the modeler, the simulation road network and the nodes of the road network are drawn, the road type, the number of lanes on the road, the speed limit of the road, etc., the road turning, the allocation of vehicle travel areas, and the number of vehicles traveling in the road network are adjusted by changing the number of vehicles traveling in the area. The modeler adds an information release button to display vehicle speed limit information and path selection information, and control the traffic lights at the intersection. The batch processor is a tool for the core model to execute batch traffic simulation, which allows users to It simulates the road network without displaying visual traffic, supports parallel computing, and simulates multiple road networks and scenarios at the same time, reducing user downtime and accelerating model development. It is used for sensitivity and option testing, and allows users to perform simulation calculations in batch mode. The programmer provides a comprehensive application programming interface (API) for extension plug-in development. Through the use of the API interface, key simulation data can be extracted in real time, such as vehicle speed, lane, road, acceleration, and the number of lanes on the road, road speed limit, whether the node has signal control, and the operating status of a certain phase of signal control. Some path induction development can be done through the API, and different induction methods can be given according to the actual traffic flow.

[0039] In this embodiment, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The emission monitor uses the API module developed by the programmer to directly collect exhaust emission data from the simulation system. During the simulation process, it can calculate the amount of exhaust emissions of all vehicles in the simulated traffic network and display it visually, and users can define the exhaust emission level data themselves; the analyzer is used to display the simulation results output by the modeler or processor, and is used to process multiple groups of data. The output evaluation indicators include: vehicle delay, vehicle travel time, vehicle parking time, road traffic flow, average number of stops, maximum vehicle queue length, and average vehicle speed. Users can customize these output evaluation indicators according to their needs. It has a special Excel wizard for filtering data and outputting statistical indicators and comparison results of the simulation at different stages. The matrix estimator is used to estimate the OD matrix at the micro level. It is fully compatible with the OD matrix, model and simulated vehicle path selection, provides an open and visual interface, and allows users to add their own prior knowledge and experience to the system kernel of the estimator. It provides an API interface so that researchers can customize the estimation program.

[0040] In this embodiment, the log files generated by the simulation are analyzed by the module to evaluate the feasibility of the control algorithm. Two experiments are conducted by injecting the following model of this article and the traditional safety distance control model, with a warm-up of 3 minutes, data collection of 54 minutes, and a data collection interval of 5 minutes. Vehicle delay, parking time and driving time are used as evaluation indicators for comparison. In different simulation time periods, the simulation data based on the vehicle following model of this article is better than the simulation data of the traditional minimum safety distance following model.

[0041] Example 2

[0042] Reference Figure 1 A traffic prediction modeling and online simulation method for vehicle-road cooperative control includes the following steps:

[0043] S1: Measure traffic and build a model;

[0044] S2: Establish a traffic simulation scenario based on the actual scenario and initialize the traffic simulation scenario;

[0045] S3: Extract road, vehicle and signal information based on the plug-in;

[0046] S4: Calculate the optimal solution for vehicle speed and traffic light matching based on the real-time data model;

[0047] S5: input the obtained real-time data into the simulation software;

[0048] S6: Analyze the data generated by the simulation to verify the feasibility of the algorithm.

[0049] In this embodiment, traffic is predicted and a model is established. The prediction model consists of an input layer, a hidden layer, and an output layer. The input features are analyzed and finally output to the LSTM layer for the final input layer of the traffic flow to sample highway traffic flow photos and predict the operation process. The CNN-LSTM neural network prediction model mainly adopts convolutional layer, pooling layer, fully connected layer and LSTM, wherein the convolutional layer corresponds to the pooling layer. After the pooling layer, all the nodes of the pooling layer are expanded into feature prediction analysis. When performing modeling and prediction, space and time must be taken as a whole, and LSTM is used as a time response model for traffic flow prediction.

[0050] In this embodiment, the process of vehicle-road collaborative simulation includes the following three key steps: full-time and space traffic information extraction technology, micro-traffic flow simulation model optimization technology and multi-intersection signal light collaborative control technology, to obtain vehicle position information, adopt collaborative control algorithm to realize collaborative control of single intersection and trunk signal, reduce vehicle travel time and average delay, initialize the simulation scene according to the investigated traffic flow data and empirical data, use the standard expansion code provided by the programmer to define the functional interface function, when the simulation starts, the system will automatically enter this trigger function, when a vehicle leaves the OD matrix, the system will enter the function, enter the application through TIMESTEP to obtain vehicle information, standard mandatory code defines mandatory execution function, through these functions can be used to obtain various information during simulation operation, obtain signal light control information through the standard code function of obtaining information, set information that can be updated in real time in the simulated road network, set vehicle speed information function, write plug-ins, collect real-time data obtained by the code, and set the simulation scene data.

[0051] In this embodiment, traffic flow data is extracted in real time, the freely generated traffic flow data is integrated, and the simulated road network is controlled by a specific control strategy. By setting the required parameters, a data file saved by the simulation is obtained at the end of the simulation. The simulation parameters include: road network initialization time, time length of data collection, time interval for data collection, and data type to be collected. The collection time interval is set to 12 minutes. The collected data types include vehicle speed, vehicle ID, and vehicle type information. These data are used to evaluate traffic quality and road network initialization time.

[0052] In this embodiment, trunk adaptive coordinated control is adopted, the traffic system status is continuously measured, and trunk signal control based on phase difference fuzzy adjustment is adopted. For large intersections, the time of each phase is allocated according to the "phase equal saturation" principle, and the import flow of large intersections is collected in real time. Each intersection calculates the signal timing according to the flow in four directions. The central control unit selects the intersection that needs coordinated control based on the road information, and then realizes the green wave band based on the phase difference, which is aimed at urban intersection signals using a four-phase control method.

[0053] In this embodiment, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The modeler provides three basic operation processes, including model establishment, traffic simulation and statistical data output. In the modeler, the simulation road network and the nodes of the road network are drawn, the road type, the number of lanes on the road, the speed limit of the road, etc., the road turning, the allocation of vehicle travel areas, and the number of vehicles traveling in the road network are adjusted by changing the number of vehicles traveling in the area. The modeler adds an information release button to display vehicle speed limit information and path selection information, and control the traffic lights at the intersection. The batch processor is a tool for the core model to execute batch traffic simulation, which allows users to It simulates the road network without displaying visual traffic, supports parallel computing, and simulates multiple road networks and scenarios at the same time, reducing user downtime and accelerating model development. It is used for sensitivity and option testing, and allows users to perform simulation calculations in batch mode. The programmer provides a comprehensive application programming interface (API) for extension plug-in development. Through the use of the API interface, key simulation data can be extracted in real time, such as vehicle speed, lane, road, acceleration, and the number of lanes on the road, road speed limit, whether the node has signal control, and the operating status of a certain phase of signal control. Some path induction development can be done through the API, and different induction methods can be given according to the actual traffic flow.

[0054] In this embodiment, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator, and users can define the exhaust emission level data themselves; the analyzer is used to display the simulation results output by the modeler or processor, and is used to process multiple groups of data. The output evaluation indicators include: vehicle delay, vehicle travel time, vehicle parking time, road traffic flow, average number of stops, maximum vehicle queue length, and average vehicle speed. It has a special Excel wizard for filtering data and outputting statistical indicators and comparison results of the simulation at different stages. The matrix estimator is used to estimate the OD matrix at the micro level, and is fully compatible with the OD matrix, model and simulated vehicle path selection. It provides an open and visual interface, allowing users to add their own prior knowledge and experience to the system kernel of the estimator, and provides an API interface so that researchers can customize the estimation program.

[0055] In this embodiment, the log files generated by the simulation are analyzed by the module to evaluate the feasibility of the control algorithm. Two experiments are conducted by injecting the following model of this article and the traditional safety distance control model, with a warm-up of 4 minutes, data collection of 55 minutes, and a data collection interval of 7 minutes. Vehicle delay, parking time, and driving time are used as evaluation indicators for comparison. In different simulation time periods, the simulation data based on the vehicle following model of this article is better than the simulation data of the traditional minimum safety distance following model.

[0056] Example 3

[0057] Reference Figure 1 A traffic prediction modeling and online simulation method for vehicle-road cooperative control includes the following steps:

[0058] S1: Measure traffic and build a model;

[0059] S2: Establish a traffic simulation scenario based on the actual scenario and initialize the traffic simulation scenario;

[0060] S3: Extract road, vehicle and signal information based on the plug-in;

[0061] S4: Calculate the optimal solution for vehicle speed and traffic light matching based on the real-time data model;

[0062] S5: input the obtained real-time data into the simulation software;

[0063] S6: Analyze the data generated by the simulation to verify the feasibility of the algorithm.

[0064] In this embodiment, traffic is predicted and a model is established. The prediction model consists of an input layer, a hidden layer, and an output layer. The input features are analyzed and finally output to the LSTM layer for the final input layer of the traffic flow to sample highway traffic flow photos and predict the operation process. The CNN-LSTM neural network prediction model mainly adopts convolutional layer, pooling layer, fully connected layer and LSTM, wherein the convolutional layer corresponds to the pooling layer. After the pooling layer, all the nodes of the pooling layer are expanded into feature prediction analysis. When performing modeling and prediction, space and time must be taken as a whole, and LSTM is used as a time response model for traffic flow prediction.

[0065] In this embodiment, the process of vehicle-road collaborative simulation includes the following three key steps: full-time and space traffic information extraction technology, micro-traffic flow simulation model optimization technology and multi-intersection signal light collaborative control technology, to obtain vehicle position information, obtain regional traffic flow information, process and analyze this information, and adopt collaborative control algorithm to realize collaborative control of single intersection and trunk signal to reduce vehicle travel time and average delay. Each simulation step will generate simulation data once, and the simulation step is set at 45ms. After the simulation data is obtained, the data is converted, and mathematical methods are used to create a simulation scene for the abstraction of the urban road traffic system. The standard mandatory code defines the mandatory execution function, and the internal default safety distance model is rewritten by the speed of the preceding vehicle and the speed of the vehicle. The standard code defines the function of obtaining information. Through the mutual use of these functions, various information during the simulation runtime can be obtained. The real-time data obtained by the collection code is set to the simulation scene data.

[0066] In this embodiment, traffic flow data is extracted in real time, and the freely generated traffic flow data is integrated. The simulated road network is controlled using a specific control strategy. By setting the required parameters, a data file saved by the simulation is obtained at the end of the simulation. The collected data types include vehicle speed, vehicle ID, and vehicle type information. These data are used to evaluate traffic quality and road network initialization time.

[0067] In this embodiment, trunk adaptive coordinated control is adopted, the traffic system status is continuously measured, and trunk signal control based on phase difference fuzzy adjustment is adopted. For large intersections, the time of each phase is allocated according to the "phase equal saturation" principle, and the import flow of large intersections is collected in real time. Each intersection calculates the signal timing according to the flow in four directions. The central control unit selects the intersection that needs coordinated control based on the road information, and then realizes the green wave band based on the phase difference, which is aimed at urban intersection signals using a four-phase control method.

[0068] In this embodiment, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The modeler provides three basic operating processes, including model establishment, traffic simulation and statistical data output. In the modeler, the simulation road network and the nodes of the road network are drawn, the types of roads are drawn, and the traffic lights at the intersections are controlled. The batch processor is a tool for the core model to execute batch traffic simulation. It allows users to simulate the road network without displaying visual traffic, supports parallel computing, and simulates multiple road networks and scenarios at the same time, reducing user downtime, accelerating model development speed, and extracting key simulation data in real time, such as vehicle speed, lane, road, acceleration, and the number of lanes, road speed limit, whether the node has signal control, and the operating status of a certain phase of signal control. Some path induction development is done through the API, and different induction methods are given according to the actual traffic flow.

[0069] In this embodiment, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The emission monitor uses the API module developed by the programmer to collect exhaust emission data directly from the simulation system. During the simulation process, it can calculate the amount of exhaust emissions of all vehicles in the simulated traffic network and display it visually, and the user can define the exhaust emission level data himself; it is used to filter data and output statistical indicators and comparison results of the simulation at different stages. The matrix estimator is used to estimate the OD matrix at the micro level, and is fully compatible with the OD matrix, model and path selection of simulated vehicles. It provides an open and visual interface, allowing users to add their own prior knowledge and experience to the system kernel of the estimator, and provides an API interface so that researchers can customize the estimation program.

[0070] In this embodiment, the log files generated by the simulation are analyzed by the module to evaluate the feasibility of the control algorithm. Two experiments are conducted by injecting the following model of this article and the traditional safety distance control model, with a warm-up of 6 minutes, data collection of 57 minutes, and a data collection interval of 8 minutes. Vehicle delay, parking time, and driving time are used as evaluation indicators for comparison. In different simulation time periods, the simulation data based on the vehicle following model of this article is better than the simulation data of the traditional minimum safety distance following model.

[0071] Comparative Example 1

[0072] The difference from Example 1 is that S1: measures traffic and establishes a model; predicts traffic and establishes a model. The prediction model consists of an input layer, a hidden layer and an output layer. The input features are analyzed and finally output to the LSTM to perform the final input layer on the traffic flow. The highway traffic flow photos are sampled and the operation process is predicted. The CNN-LSTM neural network prediction model mainly adopts convolution layer, pooling layer, full connection layer and LSTM, wherein the convolution layer corresponds to the pooling layer. After the pooling layer, all the nodes of the pooling layer are expanded into feature prediction analysis. When performing modeling and prediction, space and time must be taken as a whole, and LSTM is used as a time response model for traffic flow prediction.

[0073] Comparative Example 2

[0074] The difference from Example 2 is that S2: establishes a traffic simulation scenario based on the actual scenario, initializes the traffic simulation scenario, and uses the vehicle-road collaborative simulation process, including the following three key steps: full-time and space traffic information extraction technology, micro traffic flow simulation model optimization technology and multi-intersection signal light collaborative control technology to obtain vehicle position information, and adopts collaborative control algorithms to achieve collaborative control of single intersection and trunk signals, reducing vehicle travel time and average delay, and initializes the simulation scenario based on the investigated traffic flow data and empirical data. The functional interface function is defined using the standard expansion code of the programmer. When the simulation starts, the system will automatically enter this trigger function. When a vehicle exits the OD matrix, the system will enter the function and enter the application through TIMESTEP to obtain vehicle information. The standard mandatory code defines the mandatory execution function. By using these functions interchangeably, you can obtain various information during simulation runtime, obtain signal light control information through obtaining information standard code functions, set information that can be updated in real time in the simulation road network, set vehicle speed information functions, and write plug-ins to collect real-time data obtained by the code. Set the simulation scene data, extract traffic flow data in real time, integrate freely generated traffic flow data, and use specific control strategies to control the simulation road network. By setting the required parameters, you can get the simulation saved data file at the end of the simulation. The simulation parameters include: road network initialization time, data collection time length, data collection time interval, and data type that needs to be collected. The collection time interval is set to 12 minutes. The collected data types include vehicle speed, vehicle ID, and vehicle type information. These data are used to evaluate traffic quality. Road network initialization time.

[0075] Comparative Example 3

[0076] The difference from Example 3 is that, S6: Analyze the data generated by the simulation to verify the feasibility of the algorithm, analyze the log files generated by the simulation through the module to evaluate the feasibility of the control algorithm, and conduct two experiments by injecting the following model of this article and the traditional safety distance control model, with a warm-up of 6 minutes, data collection of 57 minutes, and a data collection interval of 8 minutes. Vehicle delay, parking time and driving time are used as evaluation indicators for comparison. In different simulation time periods, the simulation data based on the vehicle following model of this article is better than the simulation data of the traditional minimum safety distance following model.

[0077] Experimental example

[0078] The traffic prediction modeling and online simulation methods for vehicle-road cooperative control in Examples 1, 2, and 3 were tested, and the following results were obtained:

[0079] Example 1 Example 2 Example 3 Existing methods Accuracy 97％ 78％ 66％ 46％ Real-time rate 98.5％ 67％ 58％ 76％

[0080] Compared with the existing traffic prediction modeling and online simulation methods for vehicle-road collaborative control, the traffic prediction modeling and online simulation methods for vehicle-road collaborative control in Examples 1, 2 and 3 have significantly improved accuracy and real-time rate, and Example 1 is the best example.

[0081] Test report

[0082] The purpose of the present invention is to address the shortcomings of the existing technology, such as untimely data updates, low model accuracy leading to low travel efficiency, high investment costs, inability to monitor and predict traffic flow in a timely manner, low flexibility and openness of the simulation system, and large data errors. A traffic prediction modeling and online simulation method for vehicle-road collaborative control is proposed. By adopting a traffic prediction model and an online simulation method for vehicle-road collaborative control, the information of incoming and outgoing vehicles and road conditions is monitored and predicted, the road information is updated in a timely manner, the real-time nature of data monitoring is improved, the accuracy of the calculation results is improved, and data errors are reduced. The model description is accurate and flexible, and traffic is analyzed openly, thereby enhancing the practicality and convenience of model application. The simulation system is highly flexible and open.

[0083] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A traffic prediction modeling and online simulation method for vehicle-road collaborative control, characterized by: The following steps are involved: S1: predict traffic and build models; In said S1, traffic is predicted and a model is established. The prediction model consists of an input layer, a hidden layer and an output layer. The input features are analyzed and finally output to the LSTM to perform the final input layer of the traffic flow. The highway traffic flow photos are sampled and the operation process is predicted. The CNN-LSTM neural network prediction model mainly adopts the convolution layer, the pooling layer, the fully connected layer and the LSTM. The convolution layer corresponds to the pooling layer. After the pooling layer, all the nodes of the pooling layer are expanded into the prediction analysis of the features. When modeling and predicting, space and time must be taken as a whole. LSTM is used as a time response model for traffic flow prediction. S2: Establish a traffic simulation scenario based on the actual scenario and initialize the traffic simulation scenario; In the said S2, the process of vehicle-road collaborative simulation includes the following three key steps: full-time and space traffic information extraction technology, microscopic traffic flow simulation model optimization technology and multi-intersection signal light collaborative control technology, to obtain vehicle position information, obtain regional traffic flow information, process and analyze this information, and use collaborative control algorithms to achieve collaborative control of single intersection and trunk signals to reduce vehicle travel time and average delay. Each simulation step will generate simulation data, and the simulation step is set at 33-50ms. After the simulation data is obtained, the data is converted, and mathematical methods are used to abstractly create simulation scenarios for urban road traffic systems. The simulation scenarios are initialized according to the investigated traffic flow data and empirical data, and the standard expansion code provided by the programmer is used. Code, define the functional interface function. When the simulation starts, the system will automatically enter the trigger function. When a vehicle leaves the OD matrix, the system will enter the function and enter the application through TIMESTEP to obtain vehicle information. The standard mandatory code defines the mandatory execution function, and rewrites its internal default safety distance model through the speed of the front vehicle and the speed of the vehicle. The information acquisition standard code defines the function of acquiring information. Through the mutual use of these functions, various information during the simulation run can be obtained. The signal light control information is obtained through the acquisition information standard code function. The information standard code is set to set the information that can be updated in real time in the simulation road network. The vehicle speed information function is set. By writing plug-ins, the real-time data acquired by the collection code is set to set the simulation scene data. S3: Extract road, vehicle and signal information based on the plug-in; S4: Calculate the optimal solution for vehicle speed and traffic light matching based on the real-time data model; S5: input the obtained real-time data into the simulation software; S6: Analyze the data generated by the simulation to verify the feasibility of the algorithm.

2. The traffic prediction modeling and online simulation method for vehicle-road cooperative management according to claim 1 is characterized in that: In the said S2, traffic flow data is extracted in real time, the freely generated traffic flow data is integrated, and the simulated road network is controlled by a specific control strategy. By setting the required parameters, the simulation saved data file is obtained at the end of the simulation. The simulation parameters include: road network initialization time, time length of data collection, time interval of data collection and data type to be collected. The collection time interval is set to 10-15 minutes. The collected data types include vehicle speed, vehicle ID, and vehicle type information. These data are used to evaluate traffic quality. The road network initialization time.

3. The traffic prediction modeling and online simulation method for vehicle-road cooperative management according to claim 1 is characterized in that: In the S4, trunk adaptive coordinated control is adopted, the traffic system status is continuously measured, and trunk signal control based on phase difference fuzzy adjustment is adopted. For large intersections, the time of each phase is allocated according to the "phase equal saturation" principle, and the import flow of large intersections is collected in real time. Each intersection calculates the signal timing based on the flow in four directions. The central control unit selects the intersection that needs coordinated control based on the road information, and then realizes the green wave band based on the phase difference. This is for urban intersection signals that adopt a four-phase control method.

4. The traffic prediction modeling and online simulation method for vehicle-road cooperative control according to claim 1 is characterized in that: In the S5, the main modules of the simulation software include modeler, batch processor, analyzer, programmer, emission monitor and matrix estimator. The modeler provides three basic operation processes, including model establishment, traffic simulation and statistical data output. In the modeler, the simulation road network and the nodes of the road network are drawn, the road type, the number of lanes of the road, the speed limit of the road, etc. are drawn, the road turns, the allocation of vehicle travel areas, and the number of vehicles traveling in the road network are adjusted by changing the number of vehicles traveling in the area. The modeler adds an information release beat to display vehicle speed limit information and path selection information, and controls the traffic lights at the intersection. The batch processor is a tool for the core model to execute batch traffic simulation. It allows users to simulate the road network without displaying visual traffic, supports parallel computing, and simulates multiple road networks and scenarios at the same time, reducing user downtime and accelerating model development speed. It is used for sensitivity and option testing. It allows users to perform simulation calculations in a batch processing manner. The programmer provides a comprehensive application program interface for extension plug-in development. Through the use of the API interface, simulation key data can be extracted in real time. Some path guidance development is done through the API, and different guidance methods are given according to the actual traffic flow.

5. The traffic prediction modeling and online simulation method for vehicle-road cooperative control according to claim 1 is characterized in that: In the S5, the main modules in the simulation software include a modeler, a batch processor, an analyzer, a programmer, an emission monitor and a matrix estimator. The emission monitor uses the API module developed by the programmer to collect exhaust emission data directly from the simulation system. During the simulation process, it can calculate the amount of exhaust emissions of all vehicles in the simulated traffic network and display it visually, and users can define the exhaust emission level data themselves; the analyzer is used to display the simulation results output by the modeler or processor, and is used to process multiple groups of data. The output evaluation indicators include: vehicle delay, vehicle travel time, vehicle parking time, road traffic flow, average number of stops, maximum vehicle queue length, and average vehicle speed. Users can customize these output evaluation indicators according to their needs. It has a special Excel wizard for filtering data and outputting statistical indicators and comparison results of the simulation at different stages. The matrix estimator is used to estimate the OD matrix at the micro level. It is fully compatible with the OD matrix, model and simulated vehicle path selection, provides an open and visual interface, and allows users to add their own prior knowledge and experience to the system kernel of the estimator. It provides an API interface so that researchers can customize the estimation program.

6. The traffic prediction modeling and online simulation method for vehicle-road cooperative control according to claim 1 is characterized in that: In the S6, the module analyzes the log files generated by the simulation to evaluate the feasibility of the control algorithm. Two experiments are conducted by injecting the following model of this article and the traditional safety distance control model, with a warm-up time of 3-6 minutes, data collection of 54-57 minutes, and a data collection interval of 5-8 minutes. Vehicle delay, parking time, and driving time are used as evaluation indicators for comparison. In different simulation time periods, the simulation data based on the vehicle following model of this article is better than the simulation data of the traditional minimum safety distance following model.

Images