Control strategy model construction method based on big data

Abstract

The invention discloses a control strategy model construction method based on big data, and relates to the technical field of intelligent control, a traffic situation deduction and emergency strategy generation module uses a multi-agent traffic simulation technology, simulates a traffic situation in combination with real-time traffic data, evaluates risks and early warns potential crisis by integrating multiple factors, and provides a control strategy model for the traffic situation deduction and emergency strategy generation module. And after an early warning is received, an emergency strategy is generated by using an intelligent algorithm, the strategy is evaluated and optimized by using virtual rehearsal, a result is fed back to a management department, and a synergistic effect with other modules is achieved. Traditional traffic detection equipment data is integrated with multi-source heterogeneous data such as mobile phone signaling, shared bicycle and online hailed bicycle GPS, intelligent vehicle-mounted equipment and public traffic operation, more comprehensive and accurate information such as traffic flow, flow direction, travel behavior track and the like is provided, and rich and accurate basis is provided for traffic management decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and specifically to a method for constructing a control strategy model based on big data. Background Art

[0002] With the growth of urbanization and the number of motor vehicles, the urban traffic pressure has increased sharply, and problems such as traffic congestion have become prominent. The development of traffic control technology has experienced a process from manual command, simple automation to intelligent transportation systems (ITS). ITS can collect and analyze traffic data in real time and dynamically adjust signal timing, and use regional coordinated control to improve traffic efficiency.

[0003] However, there are still problems in the existing technology: First, the data sources are limited, and the integrity and accuracy are insufficient. Due to the small amount of data, it is difficult to comprehensively reflect the traffic status, making the basis for formulating control strategies insufficient. Second, most control strategies are based on local optimization and lack global optimization of the urban traffic network, which may lead to local improvements causing congestion in other areas and unable to achieve overall optimality. Third, the popularization of technologies such as the Internet has changed travel modes and demands, and new traffic patterns have made traffic flow distribution more complex. The existing technology is difficult to adapt to and effectively manage and regulate. The rise of big data technology brings new opportunities to solve the above problems. It can process a large amount of traffic data and help formulate more scientific traffic control strategies. Constructing a traffic control strategy model based on big data is expected to realize intelligent and refined traffic control, improve the overall operation efficiency of urban traffic, relieve congestion and reduce accidents.

[0004] The applicant found through retrieval that the Chinese patent discloses "Traffic Control Method and Device" with the publication number "CN115762194 A". This patent mainly generates a propagation data packet according to the event level, the control area range, the traffic control information, the data propagation link direction and the attenuation factor, and transmits the propagation data packet to the adjacent edge computing units within the control area range. By applying the technical solution of the present application, the network load can be reduced and the response time of traffic control can be reduced. However, the method of collecting traffic data in this method is relatively single, and the prediction result and the traffic control strategy are not perfect. Therefore, we propose a method for constructing a control strategy model based on big data. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for constructing a control strategy model based on big data.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for constructing a control strategy model based on big data, including the collaborative work of multiple modules. The method for constructing a control strategy model based on big data includes a data collection module, a data preprocessing module, a data analysis and mining module, a traffic flow prediction module, a traffic control strategy generation module, a traffic control execution module, an effect evaluation and feedback module, and a traffic situation deduction and emergency strategy generation module. The traffic situation deduction and emergency strategy generation module uses multi-agent traffic simulation technology, combines real-time traffic data to simulate the traffic situation, evaluates risks and warns of potential crises considering various factors. After receiving a warning, it uses intelligent algorithms to generate emergency strategies, and uses virtual rehearsal to evaluate and optimize the strategies, and feeds the results back to the management department to cooperate with the other modules.

[0007] As a further solution of the present invention: The data collection module includes multi-source data integration and data transmission and storage. The multi-source data integration integrates the data of traditional traffic detection devices and obtains the basic information of road section traffic flow, vehicle speed, and vehicle type in real time. At the same time, it accesses and collects mobile phone signal data, GPS data of shared bicycles and online car-hailing, intelligent vehicle-mounted device data, and public transportation operation data to master the travel behaviors and trajectories of traffic participants. It uses optical fiber communication technology to transmit the collected data to the data center in real time, and uses a distributed storage system to store the collected data.

[0008] As a further solution of the present invention: The data preprocessing module includes data cleaning, data standardization, and missing value processing. The data cleaning is to perform denoising processing on the originally collected data, and remove duplicate, incorrect, and abnormal data records. Set the vehicle speed range and flow threshold as needed, identify and eliminate obviously incorrect data, uniformly convert data from different sources and different formats into a standard format, and for the missing values in the data, use interpolation methods, statistical models, and machine learning algorithms to fill them to complete the data.

[0009] As a further solution of the present invention: The data analysis and mining module includes traffic feature extraction, anomaly detection, and correlation analysis. The traffic feature extraction is to extract the spatio-temporal features, travel patterns, and congestion propagation rules of traffic flow from the preprocessed data using association data mining algorithms and clustering analysis. Anomaly detection is to use anomaly detection algorithms in machine learning, and the anomaly detection algorithms include Isolation Forest and One-Class SVM, and it monitors the abnormal times in traffic data in real time and issues alarms in a timely manner. The correlation analysis is to analyze the correlations between different traffic factors, including the relationships between traffic flow, weather conditions, holidays, and large-scale events, to provide a basis for formulating traffic control strategies.

[0010] As a further solution of the present invention: The traffic flow prediction module includes a deep learning model, multi-dimensional prediction, and prediction error evaluation and correction. The deep learning model is to construct a traffic flow prediction model based on deep learning, including long short-term memory networks, gated recurrent units, and combined models of convolutional neural networks and recurrent neural networks, and use historical traffic data to train and optimize the model to learn the time series characteristics and spatial dependence relationships of traffic flow. The multi-dimensional prediction is to consider the influence of time, space, and external factors and perform multi-dimensional traffic flow prediction, providing prediction results for 15 min - 60 min, 60 min - 180 min, and 180 min - 24 h. The prediction error evaluation and correction is to regularly evaluate the prediction model and use a feedback mechanism to adjust and correct the model.

[0011] As a further solution of the present invention: The traffic control strategy generation module is to establish a strategy library containing traffic control strategies, evaluate and record the applicable scenarios and effects of each strategy, and use the particle swarm optimization algorithm and simulated annealing algorithm to select a combination of traffic control strategies from the strategy library according to the traffic flow prediction results and real-time traffic conditions, and display the generated traffic control strategies to traffic management personnel. Adjust according to the marked signal timing plan, lane change information, and induced route on the traffic control strategy. The traffic control execution module specifically docks with the existing traffic signal control system in the city, sends the optimized signal timing plan to the signal controllers at each intersection in real time, automatically adjusts the traffic lights, and uses the traffic induction module to send traffic induction information to drivers in real time through variable message signs, mobile phone APPs, and radio channels to guide vehicles to reasonably select driving routes, and integrates with the system of the traffic law enforcement department to convey the restricted and prohibited measures in the traffic control strategy to law enforcement personnel to achieve the coordinated work of traffic control and law enforcement.

[0012] As a further solution of the present invention: The effect evaluation and feedback module includes evaluation index setting, real-time monitoring and evaluation, and feedback and adjustment. The evaluation index setting is to set evaluation indexes, including average delay time, traffic capacity, vehicle queue length, and user satisfaction, to measure the implementation effect of traffic control strategies and record them for backup. The real-time monitoring and evaluation is to use the real-time collected traffic data to monitor and evaluate the execution effect of traffic control strategies in real time, analyze the changes in various indexes before and after the implementation of the strategies, feedback to the traffic control strategy generation module according to the changes, and put forward improvement suggestions to dynamically adjust and optimize traffic control strategies to form a closed-loop management.

[0013] As a further solution of the present invention: The specific steps of using the traffic situation deduction and emergency strategy generation module are as follows:

[0014] Using multi-agent based traffic simulation technology, vehicles, pedestrians, and traffic facilities on the road are regarded as independent agents. Each agent makes decisions and takes actions based on its own behavior rules and surrounding environment information. A large-scale traffic simulation model is constructed, and real-time collected traffic data, including traffic flow, vehicle speed, and vehicle distribution, is input. The urban traffic situation is dynamically simulated. During the simulation process, the impacts of traffic accidents, road construction, and sudden severe weather traffic scenarios and uncertainty factors on traffic flow are considered. On this basis, a risk assessment model is established, and the current traffic situation is quantitatively evaluated for risk according to the traffic congestion level, accident probability, and traffic pressure on key sections. The congestion index is calculated, which is the ratio of the average vehicle speed of a section to the unobstructed vehicle speed, and the accident risk coefficient is determined by combining historical accident data and current traffic behavior characteristics. Based on this, the risk level of the traffic system is determined. When the risk level exceeds the set threshold, a warning message is sent in a timely manner to remind the traffic management department to pay attention to potential traffic crises;

[0015] The congestion index is calculated and used to measure the congestion level of a section. The formula is:

[0016]

[0017] Among them, CI represents the congestion index, represents the average vehicle speed within the section, which is obtained by weighted averaging the speeds of all vehicles within one hour in the section:

[0018]

[0019] Among them, u i is the speed of the i-th vehicle, t i is the driving time of the i-th vehicle on this section, and n is the total number of vehicles passing through this section within the statistical time period, is the unobstructed vehicle speed;

[0020] Calculation of the accident risk coefficient:

[0021] The accident risk coefficient is determined by combining historical accident data and current traffic behavior characteristics. The calculation formula is:

[0022]

[0023] Among them, ARC represents the accident risk coefficient, α and β are weight coefficients, and α + β = 1, N a is the number of historical accidents on this section within the statistical time period, N t is the total number of vehicle passages on this section within the same time period, ω j is the weight of the j-th traffic behavior characteristic factor, f jis the quantization value of the j-th traffic behavior characteristic factor, and m is the total number of traffic behavior characteristic factors considered;

[0024] Traffic flow prediction error evaluation:

[0025] In the traffic flow prediction module, the root mean square error formula is introduced:

[0026]

[0027] where y k is the actual traffic flow value, is the traffic flow value predicted by the model, n is the number of samples for evaluation, continuously monitor the RMSE value, and use the feedback mechanism to continuously adjust and correct the traffic flow prediction model, and use the gradient descent method to adjust the parameters of the deep learning model;

[0028] Comprehensive traffic risk index calculation:

[0029] The formula for calculating the comprehensive traffic risk index is:

[0030]

[0031] where γ and δ are weight coefficients, and the analytic hierarchy process is used to determine that γ + δ = 1, e l is the quantization value of other factors affecting traffic risk, including the degree of influence of bad weather and the scope of road construction influence, and θ l is the weight of the corresponding factor.

[0032] As a further solution of the present invention: The specific operation steps of the method for constructing a control strategy model based on big data are as follows:

[0033] Enable traffic detection devices and access multi-source data, store the collected data in the data center using optical fiber communication, then preprocess the data by using data cleaning algorithms, unified standards and various methods to fill in missing values, analyze and mine the data by using algorithms such as association rule mining, construct a deep learning model for multi-dimensional traffic flow prediction and evaluation and adjustment, establish a strategy library, use algorithms to screen strategy combinations and visualize the display, execute traffic control according to the actual situation, set evaluation indicators, evaluate the effect based on real-time traffic data and give feedback, improve and dynamically adjust the strategy for problems, at the same time, the traffic situation deduction and emergency strategy generation module conducts real-time traffic situation dynamic simulation, risk assessment and early warning, intelligent generation of emergency strategies, and strategy rehearsal and optimization work, and collaborates with other modules to improve the overall efficiency of the traffic control strategy model.

[0034] Using the above technical solutions, compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] 1. The present invention realizes a comprehensive perception of traffic conditions through multi-source data fusion, integrating data from traditional traffic detection devices with multi-source heterogeneous data such as mobile phone signaling, GPS of shared bicycles and online car-hailing, intelligent in-vehicle devices, and public transportation operations, providing more comprehensive and accurate information on traffic flow, flow direction, travel behavior trajectories, etc., providing rich and accurate basis for traffic management decisions, making the formulation of control strategies more in line with actual traffic needs. Through the application of deep learning and intelligent optimization algorithms, and constructing deep learning models such as the combination of LSTM, GRU, CNN, and RNN, training with a large amount of historical traffic data to accurately capture the spatio-temporal characteristics of traffic flow, realizing accurate flow prediction at different time scales. At the same time, using algorithms such as particle swarm optimization and simulated annealing, screening the optimal control strategy combination from the strategy library according to the prediction results and real-time traffic conditions, greatly improving the accuracy of traffic flow prediction and the scientific nature of traffic control strategies;

[0036] 2. The present invention enables the system to have good scalability and maintainability through modular design. Modules such as data collection, preprocessing, analysis and mining, flow prediction, strategy generation, execution, and effect evaluation and feedback are independent of each other and work together. This facilitates flexible functional upgrade and optimization of specific modules according to different traffic needs and development changes in the city. When new traffic detection devices are added or new traffic patterns emerge, the corresponding modules can be quickly adjusted to adapt to the new situation. Through real-time monitoring and feedback mechanisms, it is ensured that traffic control strategies can promptly adapt to changes in traffic conditions. By continuously collecting key traffic data with a large number of real-time monitoring devices, comparing with the expected effects, once a deviation occurs, it is quickly fed back to the strategy generation module, and the strategy is adjusted and optimized accurately and in a timely manner in combination with the latest data, forming a closed-loop management, significantly improving traffic management efficiency and real-time performance, and ensuring the efficient operation of urban traffic;

[0037] 3. The present invention can improve the ability of the traffic control strategy model to handle complex situations and enhance the overall effectiveness by using the traffic situation deduction and emergency strategy generation module and interacting it with the other modules, assisting traffic management departments in making scientific decisions and achieving more efficient and intelligent traffic control. By dynamically simulating traffic situations, comprehensively considering complex scenarios such as accidents, construction, and bad weather, potential traffic crises are identified in advance, and a risk assessment model is used to quantify the risk level. Once the threshold is exceeded, an early warning is given in a timely manner, enabling traffic management departments to quickly detect the crisis and layout countermeasures in advance, significantly improving the response speed and handling ability to sudden traffic conditions. With the help of intelligent algorithms such as mixed integer programming, emergency strategies that take into account multiple objectives such as minimizing traffic delays, ensuring smoothness in key areas, and reducing the impact range of accidents are generated according to the real-time traffic state and future trends. These strategies are formulated considering the traffic network capacity and rule constraints, and are more scientific and reasonable than traditional methods, effectively improving the accuracy and effectiveness of emergency traffic control. Brief Description of the Drawings

[0038] Figure 1 This is a schematic diagram of the working process of the control strategy model in the embodiments of the present invention. Specific embodiments

[0039] The following further describes the specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0040] In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0041] Please refer to the attached Figure 1 , a method for constructing a control strategy model based on big data according to the present invention includes the collaborative work of multiple modules to achieve comprehensive collection, in-depth analysis, accurate prediction of traffic data, and generation and execution of intelligent traffic control strategies. The method for constructing a control strategy model based on big data includes a data collection module, a data preprocessing module, a data analysis and mining module, a traffic flow prediction module, a traffic control strategy generation module, a traffic control execution module, an effect evaluation and feedback module, and a traffic situation deduction and emergency strategy generation module. The traffic situation deduction and emergency strategy generation module uses multi-agent traffic simulation technology, combines real-time traffic data to simulate traffic situations, evaluates risks and warns of potential crises considering various factors. After receiving a warning, it uses intelligent algorithms to generate emergency strategies, and uses virtual previews to evaluate and optimize the strategies, and feeds the results back to the management department to cooperate with the other modules.

[0042] In an embodiment of the present invention: The data collection module includes multi-source data integration and data transmission and storage. The multi-source data integration integrates the data of traditional traffic detection devices and obtains the basic information of road section traffic flow, vehicle speed, and vehicle type in real time. At the same time, it accesses and collects mobile phone signal data, GPS data of shared bicycles and online car-hailing, intelligent in-vehicle device data, and public transportation operation data to master the travel behaviors and trajectories of traffic participants, and uses optical fiber communication technology to transmit the collected data to the data center in real time, and uses a distributed storage system to store the collected data.

[0043] In an embodiment of the present invention: The data preprocessing module includes data cleaning, data standardization, and missing value processing. Data cleaning is to perform denoising processing on the originally collected data, and remove duplicate, incorrect, and abnormal data records. Set the vehicle speed range and flow threshold as needed, identify and eliminate obviously incorrect data, uniformly convert data from different sources and different formats into a standard format, and for the missing values in the data, use interpolation methods, statistical models, and machine learning algorithms to fill them to complete the data.

[0044] In an embodiment of the present invention: The data analysis and mining module includes traffic feature extraction, anomaly detection, and correlation analysis. Traffic feature extraction is to extract the spatio-temporal features of traffic flow, travel patterns, and congestion propagation rules from the preprocessed data using association data mining algorithms and clustering analysis. Anomaly detection is to use anomaly detection algorithms in machine learning. The anomaly detection algorithms include Isolation Forest and One-Class SVM, and it monitors the abnormal time in traffic data in real time and issues alarms in a timely manner. Correlation analysis is to analyze the correlations between different traffic factors, including the relationships between traffic flow, weather conditions, holidays, and large events, providing a basis for formulating traffic control strategies.

[0045] In an embodiment of the present invention: The traffic flow prediction module includes a deep learning model, multi-dimensional prediction, and prediction error evaluation and correction. The deep learning model is to construct a traffic flow prediction model based on deep learning, including Long Short-Term Memory networks, Gated Recurrent Units, and combined models of Convolutional Neural Networks and Recurrent Neural Networks, and use historical traffic data to train and optimize the model to learn the time series features and spatial dependence relationships of traffic flow. Multi-dimensional prediction is to consider the influences of time, space, and external factors and perform multi-dimensional traffic flow prediction, providing prediction results for 15 min - 60 min, 60 min - 180 min, and 180 min - 24 h. Prediction error evaluation and correction is to regularly evaluate the prediction model and use a feedback mechanism to adjust and correct the model.

[0046] In an embodiment of the present invention: The traffic control strategy generation module is to establish a strategy library containing traffic control strategies, evaluate and record the applicable scenarios and effects of each strategy, use the Particle Swarm Optimization algorithm and the Simulated Annealing algorithm to select a combination of traffic control strategies from the strategy library according to the traffic flow prediction results and real-time traffic conditions, and display the generated traffic control strategies to traffic management personnel. Adjust according to the marked signal timing plan, lane change information, and induced route on the traffic control strategy. The traffic control execution module specifically docks with the existing traffic signal control system in the city, sends the optimized signal timing plan to the signal controllers at each intersection in real time, automatically adjusts the traffic lights, and uses the traffic guidance module to send traffic guidance information to drivers in real time through variable message signs, mobile APPs, and radio channels, guiding vehicles to reasonably select driving routes, and integrating with the system of the traffic law enforcement department to convey the restricted and prohibited measures in the traffic control strategy to law enforcement officers to achieve the coordinated work of traffic control and law enforcement.

[0047] In one embodiment of the present invention: The effect evaluation and feedback module includes evaluation index setting, real-time monitoring and evaluation, and feedback and adjustment. The evaluation index setting is to set evaluation indexes, including average delay time, traffic capacity, vehicle queue length, and user satisfaction, so as to measure the implementation effect of the traffic control strategy and record it for backup. Real-time monitoring and evaluation is to use the real-time collected traffic data to monitor and evaluate the implementation effect of the traffic control strategy in real time, analyze the changes of various indexes before and after the implementation of the strategy, feedback to the traffic control strategy generation module according to the changes, and put forward improvement suggestions to dynamically adjust and optimize the traffic control strategy, forming a closed-loop management.

[0048] In one embodiment of the present invention: The specific steps of using the traffic situation deduction and emergency strategy generation module are as follows:

[0049] Using multi-agent-based traffic simulation technology, regarding the vehicles, pedestrians, and traffic facilities on the road as independent agents, each agent makes decisions and actions based on its own behavior rules and surrounding environment information. Using this, a large-scale traffic simulation model is constructed, and the real-time collected traffic data, including traffic flow, vehicle speed, and vehicle distribution, is input, and the urban traffic situation is dynamically simulated. During the simulation process, consider the traffic scenarios and uncertainty factors such as traffic accidents, road construction, and sudden bad weather that affect traffic flow. On this basis, a risk assessment model is established, and the current traffic situation is quantitatively evaluated for risk according to the traffic congestion degree, accident probability, and traffic pressure on key sections, calculate the congestion index, that is, the ratio of the average vehicle speed of the section to the unobstructed vehicle speed and the accident risk coefficient, which is determined by combining historical accident data and current traffic behavior characteristics, so as to determine the risk level of the traffic system. When the risk level exceeds the set threshold, a warning message is sent in time to remind the traffic management department to pay attention to potential traffic crises;

[0050] The calculation of the congestion index, which is used to measure the congestion degree of the section, and its formula is:

[0051]

[0052] Among them, CI represents the congestion index, represents the average vehicle speed within the section, which is obtained by weighted averaging the speeds of all vehicles within one hour in the section:

[0053]

[0054] Among them, u i is the speed of the i-th vehicle, t i is the driving time of the i-th vehicle on this section, and n is the total number of vehicles passing through this section during the statistical time period, is the unobstructed vehicle speed;

[0055] Calculation of accident risk coefficient:

[0056] The accident risk coefficient is determined by combining historical accident data and current traffic behavior characteristics, and the calculation formula is:

[0057]

[0058] Among them, ARC represents the accident risk coefficient, α and β are weight coefficients, and α + β = 1, n a is the number of historical accidents that occurred on this section during the statistical time period. The historical accidents on this section, N t is the total number of vehicle passages on this section during the same time period, ω j is the weight of the jth traffic behavior characteristic factor, f j is the quantization value of the jth traffic behavior characteristic factor, and m is the total number of traffic behavior characteristic factors considered;

[0059] Evaluation of traffic flow prediction error:

[0060] In the traffic flow prediction module, the root mean square error formula is introduced:

[0061]

[0062] Among them, y k is the actual traffic flow value, is the traffic flow value predicted by the model. N is the number of samples used for evaluation. Continuously monitor the RMSE value and use the feedback mechanism to continuously adjust and correct the traffic flow prediction model. The parameters of the deep learning model are adjusted using the gradient descent method;

[0063] Calculation of comprehensive traffic risk index:

[0064] The calculation formula of the comprehensive traffic risk index is:

[0065]

[0066] Among them, γ and δ are weight coefficients, and γ + δ = 1 is determined using the analytic hierarchy process. e l is the quantization value of other factors affecting traffic risk, including the degree of influence of bad weather and the scope of road construction influence, θ l is the weight of the corresponding factor.

[0067] In an embodiment of the present invention: The specific operation steps of the control strategy model construction method based on big data are as follows:

[0068] Enable traffic detection devices and access multi-source data. Store the collected data in the data center using optical fiber communication. Then, preprocess the data by applying data cleaning algorithms, unified standards, and various methods to fill in missing values. Analyze and mine the data using algorithms such as association rule mining. Build a deep learning model for multi-dimensional traffic flow prediction, evaluation, and adjustment. Establish a strategy library, use algorithms to screen strategy combinations and visualize them. Implement traffic control according to the actual situation, set evaluation indicators, evaluate the effect based on real-time traffic data and give feedback. Improve and dynamically adjust strategies for problems. At the same time, the traffic situation deduction and emergency strategy generation module conducts real-time traffic situation dynamic simulation, risk assessment and warning, intelligent generation of emergency strategies, and strategy rehearsal and optimization work, collaborating with other modules to enhance the overall effectiveness of the traffic control strategy model.

[0069] In one embodiment of the present invention: Once the traffic situation deduction and emergency strategy generation module receives a risk warning, the module uses intelligent algorithms to generate targeted emergency traffic control strategies. Using the mixed integer programming algorithm, with the optimization goals of minimizing traffic delays, ensuring smooth traffic in key areas, reducing the impact scope of accidents, etc., and considering conditions such as the capacity limit of the traffic network and traffic rule constraints at the same time. When a traffic accident causes local road congestion, the algorithm quickly calculates the optimal traffic diversion plan, including temporarily adjusting the signal timing, extending the green light duration of the sections around the accident, and guiding vehicles to bypass the accident area. Under the influence of bad weather, strategies such as restricting the passage of specific vehicle types and adjusting bus operation routes are formulated. The generated emergency strategies not only consider the current traffic state but also predict the traffic development trend in a period of time in the future to ensure the effectiveness and forward-looking of the strategies. To ensure that the generated emergency strategies are practical and achieve the best results, each strategy is virtually rehearsed. In the traffic simulation environment, simulate the traffic operation situation after implementing the emergency strategy. By comparing the traffic indicators before and after the rehearsal, including the average delay time, traffic capacity, vehicle queue length, etc., evaluate the implementation effect of the strategy. Use methods such as sensitivity analysis to adjust and optimize the key parameters in the strategy, including the green light duration in the signal timing plan, the recommended priority of the induced route, etc., to further enhance the effectiveness of the emergency strategy. The rehearsal results and optimization suggestions are fed back to the traffic management department to help it make more scientific decisions.

[0070] In one embodiment of the present invention: During the calculation of the congestion index, the free-flow speed is the average driving speed of vehicles when the traffic flow is extremely low and the road is completely unobstructed. This value is usually determined according to historical traffic data and road design parameters. In the comprehensive traffic risk index, taking the influence degree of bad weather as an example, the quantitative value e can be obtained through the fuzzy comprehensive evaluation method according to meteorological parameters such as rainfall, snowfall, and wind force. l, the comprehensive traffic risk index organically combines the congestion index, accident risk coefficient and other influencing factors, which can more accurately reflect the overall risk status of the traffic system and provide a more comprehensive basis for traffic situation deduction and emergency strategy generation.

[0071] Example 1. Please refer to the appendix Figure 1 , optimization during the daily traffic peak period;

[0072] In the main urban area of a first-tier city, the traffic flow is huge during the morning rush hour on weekdays;

[0073] Data collection module: Traditional traffic detection equipment real-time feedback shows that the traffic flow on the main roads reaches thousands of vehicles per hour, and the average vehicle speed is only 20 km / h. At the same time, by accessing mobile phone signaling data, it is found that a large number of citizens move from surrounding residential areas to the central business district. Shared bicycles and online car-hailing are concentrated in hot spots. Public transportation operation data shows that the full-load rate of some bus lines exceeds 80%. These data are quickly transmitted to the data center for storage through optical fiber communication;

[0074] Data preprocessing module: According to the vehicle speed range and flow threshold, abnormal vehicle speed data caused by equipment failures are cleaned, and traffic data from different sources are unified in format. For a small amount of missing traffic flow data, linear interpolation method is used to supplement it completely;

[0075] Data analysis and mining module: Through the Apriori algorithm, it is found that during the morning rush hour on weekdays and sunny days, there is a correlation between the traffic flow on several main roads and the surrounding branch roads. The travel mode in a certain area shows a concentrated commuting characteristic. The Isolation Forest algorithm detects small-scale fluctuations and anomalies in traffic flow on specific sections, which are analyzed to be caused by the narrowing of lanes due to road construction enclosures. Correlation analysis shows that traffic flow is highly correlated with the morning rush hour;

[0076] Traffic flow prediction module: Using the Long Short-Term Memory network combined with the Convolutional Neural Network model, considering time (morning rush hour), space (locations of each section) and external factors (weekdays), it is predicted that within the next 1h - 3h, the traffic flow will continue to remain high, and the congested sections will gradually spread from the main roads in the inbound direction to the roads around the core business district;

[0077] Traffic control strategy generation module: From the strategy library, using the Particle Swarm Optimization algorithm and the Simulated Annealing algorithm, combined with the prediction results, an optimized strategy is generated. At the upstream intersection of the congested section, the green light duration is extended, and the variable lane is set as dedicated for the inbound direction;

[0078] Traffic control execution module: Send the optimized signal timing plan to the corresponding intersection signal controller, and at the same time, push real-time guidance information to drivers through the mobile phone APP and variable message signs to guide vehicles to choose appropriate routes in advance;

[0079] Effect evaluation and feedback module: After the implementation of traffic control, the average delay time was shortened by 15 minutes, the traffic capacity was increased by 20%, and the user satisfaction rate increased from 60% to 75%. Based on the evaluation results, the traffic light timing and induction information content were further fine-tuned;

[0080] Traffic situation simulation and emergency strategy generation module: Through multi-agent traffic simulation technology, the traffic flow change trend is simulated, and it is assessed in advance that if the control strategy is not adjusted in time, the roads around the core business district will be seriously congested in 2 hours. Emergency strategies are planned in advance, temporary parking spots are set up around the congested areas, and online ride-hailing vehicles are guided to park in an orderly manner to avoid random parking that aggravates congestion. After virtual rehearsal, the strategy effectively alleviates potential congestion, and the rehearsal results are fed back to the traffic management department to provide a reference for subsequent peak response. This embodiment shows that the system can accurately grasp the characteristics of daily peak traffic, plan in advance and effectively alleviate congestion.

[0081] Example 2, please refer to the attached Figure 1 , emergency response to sudden traffic accidents;

[0082] On a city expressway, a truck collided with a car, causing half of the road to be closed;

[0083] Data acquisition module: Video surveillance quickly captures the accident scene. Geomagnetic sensors and microwave radars detect that traffic flow on the accident section has dropped sharply and vehicle speeds are approaching 0. Mobile phone signaling data shows that surrounding vehicles are beginning to detour. The number of shared bicycles and online ride-hailing orders in the area around the accident has decreased. Public transportation operation data shows that related bus routes have been affected. Data is transmitted to the data center in real time.

[0084] Data preprocessing module: cleans up abnormal data fluctuations caused by accidents, standardizes various accident-related data, and ensures that each module can be used effectively;

[0085] Data analysis and mining module: Traffic feature extraction analyzes the impact range and propagation direction of the accident on the surrounding traffic flow, anomaly detection confirms the accident anomaly, and correlation analysis finds that the accident has a significant impact on the congestion of surrounding roads;

[0086] Traffic flow prediction module: Based on the gated recurrent unit model and combined with real-time traffic data, it is predicted that the road sections around the accident will be severely congested within half an hour, and the congestion will last for 2h-3h;

[0087] Traffic control strategy generation module: quickly select emergency strategies for accidents from the strategy library, set up diversion points on the upstream section of the accident, and adjust the timing of traffic lights to guide vehicles to detour;

[0088] Traffic control execution module: Immediately send the signal timing adjustment plan to relevant intersections, and release accident information and detour routes to drivers through radio and mobile APPs. Traffic law enforcement officers quickly go to the accident scene and diversion points to maintain order;

[0089] Effect evaluation and feedback module: After implementing emergency control, the average delay time of the roads around the accident was shortened by 30 minutes compared with the expectation. Although the traffic capacity decreased due to the semi-closure of the road, it remained within an acceptable range. The satisfaction of users with accident handling reached 80%. According to the evaluation, the detour route indication was further optimized;

[0090] Traffic situation deduction and emergency strategy generation module: Use multi-agent traffic simulation technology to simulate the traffic situation after the accident, evaluate the effects of existing emergency strategies, and find that if additional temporary traffic guides are arranged to direct at key intersections, the traffic efficiency can be further improved. Generate new optimized strategies and preview them. The preview results show that the average delay time can be shortened by another 10 minutes. Feed the new strategy back to the traffic management department and implement it, effectively improving the accident emergency handling ability.

[0091] Example 3. Please refer to the appendix Figure 1 , traffic control under bad weather;

[0092] In a coastal city, a typhoon hit, and strong winds and heavy rains caused waterlogging on some roads and reduced visibility;

[0093] Data collection module: Video monitoring shows that some sections are severely waterlogged and vehicles are moving slowly. Data from geomagnetic sensors and microwave radars reflect a significant decrease in traffic flow, and the vehicle speed is generally lower than 15 km / h. Mobile signaling data shows that the willingness of citizens to travel has decreased, and the order volume of shared bicycles and online car-hailing has decreased sharply. Public transportation operation data shows that some bus lines have been temporarily adjusted or suspended. The data is quickly aggregated to the data center through optical fiber communication;

[0094] Data preprocessing module: Clean the abnormal data caused by bad weather interference, and process various types of data according to the standard format of bad weather scenarios;

[0095] Data analysis and mining module: Through association rule mining, strong associations between bad weather and traffic flow, vehicle speed, and accident incidence are found. Anomaly detection monitors abnormal situations where some sections of the road are paralyzed due to waterlogging;

[0096] Traffic flow prediction module: Use a combined model of convolutional neural network and recurrent neural network to predict that the bad weather will continue to affect traffic for 4 - 6 hours, and the traffic conditions will further deteriorate as the intensity of wind and rain changes;

[0097] Traffic control strategy generation module: According to the prediction results, generate control strategies for adverse weather from the strategy library, restrict the passage of large trucks and motorcycles, implement traffic control on some severely waterlogged sections, and adjust the bus operation routes to avoid dangerous areas;

[0098] Traffic control execution module: Release adverse weather traffic control information to drivers through variable message signs and mobile APPs. The traffic signal control system adjusts the signal timing, extends the green light interval time to ensure that vehicles can safely pass through intersections. Traffic law enforcement officers strengthen road patrols to guide vehicles to drive safely;

[0099] Effect evaluation and feedback module: After the implementation of the control, the road accident incidence rate has decreased by 50% compared with that without control during adverse weather. Although the average delay time has increased, it is within an acceptable range. The user satisfaction with the traffic management during adverse weather reaches 70%. According to the evaluation, continuously optimize the content and method of traffic control information release;

[0100] Traffic situation deduction and emergency strategy generation module: Use multi-agent traffic simulation technology to simulate the evolution of traffic situations under adverse weather, evaluate the effects of existing strategies, and find that if warning signs are set in advance on easily waterlogged sections, the accident risk can be further reduced. Generate new strategies and conduct pre-drills. The pre-drill results are good and are fed back to the traffic management department for implementation. This embodiment shows that the system can generate effective control strategies through the cooperation of multiple modules in complex situations of adverse weather.

[0101] Through the above content and the content of the embodiment, the construction process of the traffic control strategy model based on big data can be clearly demonstrated. From enabling traffic detection devices to collect multi-source data, analyzing and predicting after preprocessing, to generating and executing traffic control strategies, and then optimizing the strategies through effect evaluation and feedback. At the same time, the traffic situation deduction and emergency strategy generation module can handle emergencies, and each module operates in coordination to ensure the systematicness and scientificity of the construction method of the traffic control strategy model based on big data.

[0102] Although the present invention is disclosed above in a preferred embodiment, it is not used to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for constructing a control strategy model based on big data, including the collaborative work of multiple modules, characterized in that: The method for constructing a control strategy model based on big data includes a data collection module, a data preprocessing module, a data analysis and mining module, a traffic flow prediction module, a traffic control strategy generation module, a traffic control execution module, an effect evaluation and feedback module, and a traffic situation deduction and emergency strategy generation module. The traffic situation deduction and emergency strategy generation module uses multi-agent traffic simulation technology, combines real-time traffic data to simulate the traffic situation, evaluates risks and warns of potential crises considering various factors. After receiving a warning, it uses intelligent algorithms to generate emergency strategies, and uses virtual rehearsal to evaluate and optimize the strategies, and feeds the results back to the management department to cooperate with the other modules.

2. The method for constructing a control strategy model based on big data according to claim 1, wherein: The data collection module includes multi-source data integration and data transmission and storage. Multi-source data integration integrates data from traditional traffic detection devices and obtains basic information on road section traffic flow, vehicle speed, and vehicle type in real time. At the same time, it accesses and collects mobile phone signal data, GPS data of shared bicycles and online car-hailing, intelligent in-vehicle device data, and public transportation operation data to master the travel behaviors and trajectories of traffic participants. It uses fiber optic communication technology to transmit the collected data to the data center in real time and stores the collected data using a distributed storage system.

3. The method for constructing a control strategy model based on big data according to claim 1, wherein: The data preprocessing module includes data cleaning, data standardization, and missing value processing. Data cleaning performs denoising on the originally collected data and removes duplicate, incorrect, and abnormal data records. Set the vehicle speed range and flow threshold as needed to identify and eliminate obviously incorrect data. Convert data from different sources and in different formats into a standard format uniformly. For the missing values in the data, use interpolation methods, statistical models, and machine learning algorithms to fill them and complete the data.

4. A method for constructing a control strategy model based on big data according to claim 1, characterized in that: The data analysis and mining module includes traffic feature extraction, anomaly detection, and correlation analysis. Traffic feature extraction uses association data mining algorithms and clustering analysis to extract spatio-temporal features of traffic flow, travel patterns, and congestion propagation rules from the preprocessed data. Anomaly detection uses anomaly detection algorithms in machine learning, including Isolation Forest and One-Class SVM, and monitors abnormal times in traffic data in real time to issue alarms in a timely manner. Correlation analysis analyzes the correlations between different traffic factors, including the relationships between traffic flow, weather conditions, holidays, and large events, providing a basis for formulating traffic control strategies.

5. The method for constructing a control strategy model based on big data according to claim 1, wherein: The traffic flow prediction module includes a deep learning model, multi-dimensional prediction, and prediction error evaluation and correction. The deep learning model is to construct a traffic flow prediction model based on deep learning, including long short-term memory networks, gated recurrent units, and a combined model of convolutional neural networks and recurrent neural networks, and uses historical traffic data to train and optimize the model to learn the time series characteristics and spatial dependence relationships of traffic flow. The multi-dimensional prediction takes into account the influences of time, space, and external factors, and conducts multi-dimensional traffic flow prediction, providing prediction results for 15 min - 60 min, 60 min - 180 min, and 180 min - 24 h. The prediction error evaluation and correction is to regularly evaluate the prediction model and use a feedback mechanism to adjust and correct the model.

6. A method for constructing a control strategy model based on big data according to claim 1, characterized in that: The traffic control strategy generation module is to establish a strategy library containing traffic control strategies, evaluate and record the applicable scenarios and effects of each strategy, use the particle swarm optimization algorithm and simulated annealing algorithm to select a combination of traffic control strategies from the strategy library according to the traffic flow prediction results and real-time traffic conditions, and display the generated traffic control strategies to traffic management personnel, and adjust according to the marked signal timing plan, lane change information, and induced route on the traffic control strategy. The traffic control execution module specifically docks with the existing traffic signal control system in the city, sends the optimized signal timing plan to the signal controllers at each intersection in real time, automatically adjusts the traffic lights, and uses the traffic induction module to real-time publish traffic induction information to drivers through variable message signs, mobile phone APPs, and radio channels to guide vehicles to reasonably select driving routes, and integrates with the system of the traffic law enforcement department to convey the restricted and prohibited measures in the traffic control strategy to law enforcement personnel to achieve the collaborative work of traffic control and law enforcement.

7. A method for constructing a control strategy model based on big data according to claim 1, characterized in that: The effect evaluation and feedback module includes evaluation index setting, real-time monitoring and evaluation, and feedback and adjustment. The evaluation index setting is to set evaluation indexes, including average delay time, traffic capacity, vehicle queue length, and user satisfaction, to measure the implementation effect of traffic control strategies and record them for backup. The real-time monitoring and evaluation is to use the real-time collected traffic data to conduct real-time monitoring and evaluation of the implementation effect of traffic control strategies, analyze the changes in various indexes before and after the implementation of the strategies, feedback to the traffic control strategy generation module according to the changes, and put forward improvement suggestions to dynamically adjust and optimize traffic control strategies to form a closed-loop management.

8. A method for constructing a control strategy model based on big data according to claim 1, characterized in that: The specific steps of using the traffic situation deduction and emergency strategy generation module are as follows: Using multi-agent based traffic simulation technology, vehicles, pedestrians and traffic facilities on the road are regarded as independent agents. Each agent makes decisions and takes actions based on its own behavior rules and surrounding environment information. A large-scale traffic simulation model is constructed, and real-time collected traffic data, including traffic flow, vehicle speed and vehicle distribution, is input. The urban traffic situation is dynamically simulated. During the simulation process, the traffic scenarios and uncertainty factors of traffic accidents, road construction and sudden bad weather are considered to impact the traffic flow. On this basis, a risk assessment model is established. The current traffic situation is quantitatively evaluated for risks according to the traffic congestion level, accident occurrence probability and traffic pressure on key sections, and the congestion index is calculated, that is, the ratio of the average vehicle speed of the section to the unobstructed vehicle speed, and the accident risk coefficient, which is determined by combining historical accident data and current traffic behavior characteristics. Thus, the risk level of the traffic system is determined. When the risk level exceeds the set threshold, a warning message is sent in time to remind the traffic management department to pay attention to potential traffic crises; Congestion index calculation, which is used to measure the congestion level of a section. The formula is: Among them, CI represents the congestion index, represents the average vehicle speed within the road section, which is obtained by weighted averaging all vehicle speeds within one hour in the road section: where, u i is the speed of the i-th vehicle, t i is the travel time of the i-th vehicle on this section of the road, and n is the total number of vehicles passing through this section of the road within the statistical time period, is the unimpeded vehicle speed; Accident risk coefficient calculation: The accident risk coefficient is determined by combining historical accident data and current traffic behavior characteristics. The calculation formula is: Among them, ARC represents the accident risk coefficient, α and β are weight coefficients, and α + β = 1, N a is the number of historical accidents that occurred on this section during the statistical time period, N t is the total number of vehicle passages on this section during the same time period, ω j is the weight of the j-th traffic behavior characteristic factor, f j is the quantification value of the j-th traffic behavior characteristic factor, and m is the total number of traffic behavior characteristic factors considered; Traffic flow prediction error assessment: In the traffic flow prediction module, the root mean square error formula is introduced: Among them, y k is the actual traffic flow value, is the traffic flow value predicted by the model. For the number of samples used for evaluation, continuously monitor the RMSE value, and use the feedback mechanism to continue to adjust and correct the traffic flow prediction model. The gradient descent method is used to adjust the parameters of the deep learning model; Comprehensive traffic risk index calculation: The formula for calculating the comprehensive traffic risk index is: Among them, γ and δ are weight coefficients, and the analytic hierarchy process is used to determine that γ + δ = 1, and e l is the quantified value of other factors affecting traffic risk, including the degree of influence of bad weather and the scope of influence of road construction, θ l is the weight of the corresponding factor.

9. A method for constructing a control strategy model based on big data according to claim 1, characterized in that: The specific operation steps of the above-mentioned method for constructing a control strategy model based on big data are as follows: Enable traffic detection devices and access multi-source data. Store the collected data in the data center using optical fiber communication. Then preprocess the data by using data cleaning algorithms, unified standards and various methods to fill in missing values. Analyze and mine the data using algorithms such as association rule mining. Construct a deep learning model for multi-dimensional traffic flow prediction, evaluation and adjustment, and establish a strategy library. Use algorithms to screen strategy combinations and visually display them. Implement traffic control according to the actual situation, set evaluation indicators, evaluate the effect based on real-time traffic data and give feedback, improve and dynamically adjust the strategy for problems. At the same time, the traffic situation deduction and emergency strategy generation module conducts real-time traffic situation dynamic simulation, risk assessment and warning, intelligent generation of emergency strategies, and strategy rehearsal and optimization work, collaborating with other modules to improve the overall efficiency of the traffic control strategy model.

Images