Traffic light intelligent adjustment method and system fused with Internet of Vehicles information

Abstract

The invention discloses a traffic light intelligent adjustment method and system fused with Internet of Vehicles information, and relates to the technical field of intelligent traffic control, and the method comprises the steps: collecting multi-source traffic data through roadside equipment, and carrying out the time-space alignment and abnormal value filtering; constructing a deep learning prediction model based on the filtered data, and predicting a traffic demand in combination with historical and real-time Internet of Vehicles information; generating a dynamic timing strategy according to prediction, and synchronizing the dynamic timing strategy to an associated intersection through V2X; deploying an edge computing node to adjust traffic light parameters in real time, and sending a suggested speed to the vehicle; a closed-loop feedback mechanism is established, and strategies and models are corrected on line based on trajectory data and efficiency indexes. The technical problem that an existing traffic light timing system cannot fuse multi-source Internet of Vehicles data in real time to carry out accurate demand prediction and adaptive adjustment in a dynamic traffic environment is solved, and the technical effects of remarkably improving the traffic efficiency and shortening the emergency response delay in a complex traffic scene are achieved.

Description

Technical Field

[0001] This application relates to the field of intelligent traffic control technology, and in particular, to a method and system for intelligent adjustment of traffic lights integrating vehicle networking information. Background Art

[0002] In traditional urban traffic management, traffic light control systems mostly adopt fixed timing or static optimization strategies based on wired detection devices (such as induction loops, cameras), and it is difficult to perceive the dynamic changes of complex traffic flows in real time. Existing technologies rely on historical traffic flow statistical data for cycle adjustment and cannot effectively integrate multi-dimensional dynamic information such as vehicle real-time positions, pedestrian densities, and emergency vehicle passing requirements. Especially in a high-density vehicle networking environment, there are problems such as spatio-temporal asynchrony and inconsistent standards in the massive data generated by heterogeneous data sources (such as roadside units, in-vehicle terminals, mobile terminals), resulting in limited accuracy of traffic state prediction. In addition, local congestion or emergency vehicle passing requirements caused by sudden traffic events (such as accidents, large-scale events) often lead to a sharp drop in the passing efficiency of the regional road network due to system response delays and the lack of multi-intersection coordination. Therefore, how to achieve real-time fusion of multi-source vehicle networking data and accurate prediction of dynamic traffic demands, and accordingly construct an adaptive traffic light coordination control mechanism, has become a key technical bottleneck in improving the intelligence of urban traffic management.

[0003] In the related technologies at the present stage, there is a technical problem that the traffic light timing system cannot fuse multi-source vehicle networking data in real time for accurate demand prediction and adaptive adjustment in a dynamic traffic environment. Summary of the Invention

[0004] This application provides a method and system for intelligent adjustment of traffic lights integrating vehicle networking information, and solves the technical problem that the existing traffic light timing system cannot fuse multi-source vehicle networking data in real time for accurate demand prediction and adaptive adjustment in a dynamic traffic environment.

[0005] This application provides a method for intelligent adjustment of traffic lights integrating vehicle networking information, including:

[0006] Collect multi-source traffic data in real time through roadside perception devices, perform spatio-temporal alignment and outlier filtering on the multi-source traffic data, and obtain the filtered multi-source traffic data; construct a multi-modal traffic state prediction model based on deep learning with the filtered multi-source traffic data, combine historical traffic data and real-time vehicle networking information, and predict the traffic demands in each direction within a future variable time window; generate a dynamic timing strategy according to the prediction results, and synchronize the strategy parameters to associated intersections through the V2X communication protocol; deploy edge computing nodes to perform real-time dynamic adjustment on traffic light controllers, generate traffic light cycle parameters, phase difference parameters, and green signal ratio parameters based on the dynamic timing strategy, and send recommended driving speed instructions to vehicle terminals through the vehicle-road collaborative system; establish a closed-loop feedback mechanism, and perform online parameter correction and model iterative update on the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators.

[0007] This application provides a traffic light intelligent adjustment system integrating vehicle networking information, including:

[0008] A multi-source traffic data acquisition module, which is used to collect multi-source traffic data in real time through roadside perception devices, perform spatio-temporal alignment and outlier filtering on the multi-source traffic data, and obtain the filtered multi-source traffic data; a traffic demand prediction module, which is used to construct a multi-modal traffic state prediction model based on deep learning with the filtered multi-source traffic data, combine historical traffic data and real-time vehicle networking information, and predict the traffic demands in each direction within a future variable time window; a dynamic timing strategy generation module, which is used to generate a dynamic timing strategy according to the prediction results, and synchronize the strategy parameters to associated intersections through the V2X communication protocol; an edge computing node deployment module, which is used to deploy edge computing nodes to perform real-time dynamic adjustment on traffic light controllers, generate traffic light cycle parameters, phase difference parameters, and green signal ratio parameters based on the dynamic timing strategy, and send recommended driving speed instructions to vehicle terminals through the vehicle-road collaborative system; a closed-loop feedback mechanism construction module, which is used to establish a closed-loop feedback mechanism, and perform online parameter correction and model iterative update on the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators.

[0009] The intelligent traffic light adjustment method and system integrating vehicle networking information proposed in this application first collect multi-source traffic data through roadside devices, perform spatio-temporal alignment and outlier filtering; build a deep learning prediction model based on the filtered data, and combine historical and real-time vehicle networking information to predict traffic demand; generate a dynamic timing strategy according to the prediction, and synchronize it to related intersections through V2X; deploy edge computing nodes to adjust traffic light parameters in real time and send recommended speeds to vehicles; establish a closed-loop feedback mechanism to online correct the strategy and model based on trajectory data and efficiency indicators, achieving the technical effect of significantly improving traffic efficiency and shortening emergency response delays in complex traffic scenarios. Brief Description of the Drawings

[0010] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the application. It should be understood that the operations before or below do not necessarily need to be executed precisely in sequence. On the contrary, according to the need, they can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0011] Figure 1 It is a schematic flowchart of the intelligent traffic light adjustment method integrating vehicle networking information provided by an embodiment of the present application;

[0012] Figure 2 It is a schematic structural diagram of the intelligent traffic light adjustment system integrating vehicle networking information provided by an embodiment of the present application.

[0013] Description of the reference numerals: Multi-source traffic data acquisition module 10, traffic demand prediction module 20, dynamic timing strategy generation module 30, edge computing node deployment module 40, closed-loop feedback mechanism construction module 50. Detailed Embodiments

[0014] The above description is only an overview of the technical solutions of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically illustrates the detailed embodiments of this application.

[0015] In order to make the purpose, technical solutions and advantages of this application clearer, the present application will be further described in detail below in conjunction with the drawings. The described embodiments should not be regarded as limitations of this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

[0016] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first / second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application.

[0017] The embodiments of this application provide a method for intelligently adjusting traffic lights by integrating vehicle networking information, as Figure 1 shown, the method includes:

[0018] Step S100, real-time collect multi-source traffic data through roadside perception devices, perform spatio-temporal alignment and outlier filtering on the multi-source traffic data, and obtain the filtered multi-source traffic data. Specifically, to obtain the filtered multi-source traffic data, it is necessary to first select and deploy roadside perception devices, such as cameras, radars, sensors, and V2X communication devices, reasonably arrange their positions according to the road layout and monitoring requirements, and ensure smooth communication. Then, carry out multi-source traffic data collection, and use different devices to collect vehicle flow, pedestrian density, vehicle driving speed, and V2X communication data respectively. After that, perform spatio-temporal alignment. In terms of time, synchronize and calibrate the clocks of each device based on the GPS clock, and interpolate or resample data with different sampling frequencies; in terms of space, unify the geographic coordinate system, and match the data with the actual road positions with the help of GIS technology. Finally, perform outlier filtering, adopt statistical analysis, machine learning methods, and data consistency checks, process numerical data, identify abnormal patterns, and verify conflicting data, and finally obtain reliable data for subsequent traffic analysis and decision-making.

[0019] In a possible implementation, multi-source traffic data is collected in real time by roadside sensing devices. The multi-source traffic data is subjected to spatio-temporal alignment and outlier filtering to obtain filtered multi-source traffic data. Step S100 further includes step S110 of grouping the original vehicle network data in time series, calculating the interquartile range and standard deviation of each group of data, and obtaining a sequence of group data variances. Specifically, to obtain the sequence of group data variances, it is necessary to first clarify the purpose of the solution, that is, by grouping the original vehicle network data in time series and calculating the interquartile range and standard deviation, analyze the dispersion degree and distribution characteristics of the data in different time periods, and provide support for subsequent data processing. Then data preparation is carried out, collecting the original data from each data source of the vehicle network, and cleaning and removing error data and handling missing values. Then determine an appropriate time interval according to the analysis requirements, and use the data processing library to group the original data in time series. After that, for each group, sort the data and calculate the lower quartile Q1 and the upper quartile Q3 to obtain the interquartile range, and at the same time calculate the standard deviation according to the standard deviation formula. Finally, combine the interquartile range and standard deviation of each group, store them in a new DataFrame, and form a sequence of group data variances for subsequent in-depth analysis.

[0020] Step S120, based on the sliding window mechanism, identify the outlier data points in the sequence of group data variances, and use the cubic spline interpolation method to compensate for the missing data. Specifically, when processing the sequence of group data variances, identify the outlier data points and compensate for the missing data. First, identify the outliers based on the sliding window mechanism, determine the window size and step size, set the sliding window and calculate the statistical features, and determine the outlier judgment criterion according to the threshold method or in combination with the interquartile range, and mark the outlier points. Then use the cubic spline interpolation method to compensate for the missing data, locate the missing data points, select adjacent known data points, construct a cubic spline interpolation function and substitute the abscissa of the missing point to calculate the compensation value. Finally, conduct result verification and optimization, verify whether the outlier points are reasonable, adjust the parameters and re-identify; evaluate the interpolation effect, when it is not ideal, adjust the range of known data points or switch to other interpolation methods, so as to improve the data quality and usability and provide a reliable basis for subsequent analysis and decision-making.

[0021] Step S130: For the conflicting data collected by the radar and the camera, perform data fusion and credibility verification through the Kalman filtering algorithm to obtain filtered multi-source traffic data. Specifically, process the conflicting data collected by the radar and the camera. First, perform data preprocessing to synchronize the time of the two sets of data and remove noise and outliers. Then, establish a state model and an observation model, define the traffic target state vector, establish the state transition equation based on physical laws, and construct the observation equations for the radar and the camera respectively. Next, implement the Kalman filtering algorithm, initialize the state vector, covariance matrix, and noise covariance matrix, and perform the prediction and update steps to update the state vector and covariance matrix in combination with the observed values. After that, perform credibility verification. Evaluate the observation credibility based on the observation noise covariance matrix, evaluate the fusion result credibility through the state covariance matrix, and set a threshold to judge the reliability of the result. Finally, output and store the fused and verified state vector as the filtered data, and continuously update it as new data is collected. In addition, the algorithm parameters can be adaptively adjusted, and other sensor data can also be fused to optimize and improve the solution.

[0022] Step S200: Construct a multi-modal traffic state prediction model based on deep learning using the filtered multi-source traffic data, and combine historical traffic data and real-time vehicle networking information to predict the traffic demand in each direction within a future variable time window. Specifically, first perform data preparation and integration, prepare the filtered multi-source traffic data, historical traffic data, and real-time vehicle networking information, and fuse them into a complete data set after format conversion and standardization processing. Then, construct a multi-modal traffic state prediction model based on deep learning, determine an architecture including CNN, RNN, or LSTM and an attention mechanism, and design the input, hidden, and output layers. Subsequently, divide the data set, define the loss function and select the optimization algorithm, and train the model. Adjust the hyperparameters in combination with the validation set during training. After the model is trained, deploy it to the actual system, input real-time data for prediction, and post-process the results to assist in decision-making. Then, evaluate the model using evaluation metrics. If the performance does not meet the standard, analyze the reasons and optimize it. Finally, continuously improve, continuously update the data, and iterate the model according to actual feedback and new requirements to adapt to traffic changes and accurately predict the traffic demand in each direction within a future variable time window.

[0023] In a possible implementation, a multi-modal traffic state prediction model based on deep learning is constructed with the filtered multi-source traffic data. Combining historical traffic data and real-time vehicle networking information, the traffic demand in each direction within a future variable time window is predicted. Step S200 further includes step S210 of establishing a spatio-temporal graph convolutional network to process road topology features. Specifically, to establish a spatio-temporal graph convolutional network to process road topology features, data preparation needs to be carried out first. Road topology and traffic state data are collected from GIS, traffic sensors, etc., and are preprocessed through cleaning, normalization, etc. Then a graph is constructed. Intersections and road segments are abstracted as nodes, and the connection relationships are represented by edges, which can be represented by an adjacency matrix or an edge list. When designing the spatio-temporal graph convolutional network, graph convolutional operations are used spatially and stacked in multiple layers to learn complex spatial dependencies. Temporally, the time dimension is introduced and fused with spatial features. When training the model, the dataset is divided, the loss function is defined according to the task, a good optimization algorithm is selected, the parameters are adjusted through batch training, and the hyperparameters are adjusted in combination with the validation set. Then the model is evaluated with the test set. If the requirements are met, it is applied to scenarios such as traffic flow prediction. Finally, according to the evaluation and application feedback, the model is improved from aspects such as structural adjustment, feature engineering, and algorithm optimization to enhance its performance and prediction accuracy.

[0024] Step S220, using a long short-term memory network to model the temporal dependencies of traffic flow. Specifically, to use a long short-term memory network to model the temporal dependencies of traffic flow, first, traffic flow data from multiple channels such as road sensors and surveillance cameras need to be collected, covering indicators such as traffic volume and vehicle speed and recording timestamps. Then data preprocessing is carried out, including cleaning noise and outliers, using interpolation methods to handle missing values, and using normalization methods to unify the data magnitude. Then the training set, validation set, and test set are divided proportionally. When constructing the LSTM model, the input layer is designed to receive the preprocessed data, the LSTM layer is used as the core to process long-term dependencies through gating mechanisms, and the output layer is designed according to the task objectives; at the same time, parameters such as the number of hidden units and time steps are set, and a suitable activation function is selected. When training, loss functions such as mean squared error are defined, optimization algorithms such as Adam are selected, the data is input batch by batch for forward propagation, calculating the loss and backward propagation to update the parameters, and the hyperparameters are adjusted in combination with the validation set. When evaluating the model, indicators such as MAE and RMSE are selected, and the indicator values are calculated with the test set and compared with the threshold or the benchmark model. If the performance is not good, it can be optimized from aspects such as adjusting the model structure, improving data preprocessing, and optimizing hyperparameters. Finally, the model that meets the requirements is applied to actual scenarios such as traffic flow prediction to provide decision support.

[0025] Step S230: Dynamically weight the vehicle intention information in V2X communication data by integrating an attention mechanism. Specifically, to dynamically weight the vehicle intention information in V2X communication data by integrating an attention mechanism, first collect V2X communication data containing vehicle basic information and intention information from the vehicle networking system and store it. Then perform preprocessing, including cleaning noise and outliers, encoding intention information, and normalizing numerical data. Next, design an attention mechanism, represent the preprocessed information as a sequence of feature vectors, define queries, keys, and values, calculate attention scores and weights, and perform weighted summation on the value vectors. After that, select basic models such as RNN and LSTM, and integrate the attention mechanism into them. Divide the dataset, define a loss function according to the task, select an appropriate optimization algorithm, input data in batches for training, and adjust hyperparameters in combination with the validation set. Evaluate the model using the test set, select metrics such as accuracy and MSE, and compare with thresholds or baseline models. If the performance is not good, optimize from aspects such as adjusting attention mechanism parameters, improving the basic model structure, and optimizing data preprocessing. Finally, apply the model that meets the requirements to actual scenarios such as intelligent traffic signal control, and dynamically adjust traffic management strategies based on real-time V2X data.

[0026] Step S240: Optimize the multi-objective loss function through a reinforcement learning framework, balance the traffic efficiency and safety indicators, and generate a multi-modal traffic state prediction model. Specifically, to optimize the multi-objective loss function through a reinforcement learning framework and balance the traffic efficiency and safety indicators to generate a multi-modal traffic state prediction model, first clarify the quantization methods of traffic efficiency and safety indicators, collect multi-source data such as traffic sensors, cameras, and V2X communication and perform preprocessing. Then construct a multi-modal traffic state prediction model, combine CNN, RNN, etc. to process different modal data and fuse features, and perform preliminary training with the initial loss function. Next, build a reinforcement learning framework, abstract the traffic system as an environment, define a multi-objective reward function containing traffic efficiency and safety indicators, and select an appropriate reinforcement learning algorithm. After that, combine reinforcement learning with the prediction model, calculate the multi-objective loss function according to the prediction results and actual situations, and update the model parameters through policy optimization. Then evaluate the model using the test dataset, judge the performance based on the evaluation metrics, and tune the model structure, algorithm parameters, etc. when the requirements are not met. Finally, deploy the optimized model to the actual traffic management system, predict the traffic state based on real-time data and provide decision support, and continuously collect data to further optimize the model.

[0027] Step S300: Generate a dynamic timing strategy based on the prediction results and synchronize the strategy parameters to the associated intersections via the V2X communication protocol. Specifically, first obtain the multi-modal traffic state prediction results and preprocess them to extract the key indicators related to the dynamic timing strategy. Then, generate the main road green wave coordination, emergency vehicle priority passage, and pedestrian crossing protection plans respectively based on these indicators, integrate and optimize the strategy parameters. After that, select a suitable V2X communication protocol and configure it, encapsulate the optimized strategy parameters into a data format compliant with the protocol, and send them to the traffic signal control devices at the associated intersections via this protocol. After receiving and parsing for verification, if it passes, update the local signal timing parameters. During the implementation of the strategy, monitor the changes in traffic status in real time, adjust the strategy according to the monitoring results. If the actual situation deviates greatly from the prediction or the effect is not good, regenerate and synchronize the strategy. At the same time, establish a long-term evaluation mechanism to continuously optimize the strategy generation and adjustment methods.

[0028] In a possible implementation manner, when generating a dynamic timing strategy based on the prediction results and synchronizing the strategy parameters to the associated intersections via the V2X communication protocol, step S300 further includes step S310: Divide the traffic scenarios into three modes: normal traffic, peak congestion, and emergency events. Specifically, to divide the traffic scenarios into these three modes, data collection and analysis need to be carried out first. Collect data from multiple aspects such as traffic flow, video surveillance, special events, and meteorology. After preprocessing such as cleaning and normalization, perform time series and clustering analysis. Then define the three modes. In the normal traffic mode, the traffic flow is stable and the driving is smooth, and the judgment criteria can be set according to indicators such as traffic volume and vehicle speed; the peak congestion mode occurs during specific periods with a large traffic volume and slow driving speed, and there are also corresponding judgment thresholds; the emergency event mode requires special scheduling due to special circumstances and is determined based on the alarm information from relevant departments. Then build a real-time monitoring system, design a pattern recognition algorithm, collect and analyze data in real time, and update and feedback in a timely manner when the mode changes. Finally, verify the accuracy of the mode division through historical data and actual applications, and optimize the mode definition, judgment criteria, and recognition algorithm according to the results to adapt to the development and changes of the traffic system.

[0029] Step S320, adopt the green wave band optimization method for the normal traffic mode. Specifically, for the normal traffic mode, adopt the green wave band optimization method based on the genetic algorithm. First, collect data such as road infrastructure, traffic flow, and signal light settings and preprocess them, removing noise, filling missing values, and normalizing. Then construct a green wave band model, determine the goal of improving vehicle passing efficiency, define decision variables such as signal light cycle duration, and set constraints such as traffic flow, time, and safety to form a mathematical model. Then design a genetic algorithm, encode the decision variables to form chromosomes, randomly generate an initial population, define a fitness function to evaluate the quality of the solution, design genetic operations such as selection, crossover, and mutation, and set termination conditions. Run the algorithm to solve, evaluate the output solution, and if it is not ideal, adjust the algorithm parameters and re-optimize. Finally, apply the optimized solution to the actual signal light control system, establish a monitoring system, evaluate the effect according to real-time traffic data, and adjust and optimize in time when there is a deviation.

[0030] Step S330, enable the multi-intersection collaborative control mechanism for the peak congestion mode. Specifically, for the peak congestion mode, enable the multi-intersection collaborative control mechanism based on game theory. First, collect data such as traffic flow, intersection topology, and signal light information and preprocess them, cleaning noise and outliers and normalizing the data. Then model the multi-intersection collaborative control problem, take the signal light control strategies of each intersection as participants, define a revenue function that comprehensively considers passing efficiency, fairness, and stability, and formulate the interaction rules of each participant. Then select a suitable game model and design an analytical or numerical solution algorithm to solve the strategy. Apply the solved strategy to the actual signal light control system, establish a real-time monitoring system to evaluate the effect, and adjust the strategy in time according to the feedback. Regularly evaluate the effect of the mechanism using indicators such as average delay time, and optimize the game model, algorithm, and strategy adjustment method accordingly. Finally, integrate this mechanism with the existing traffic management system and conduct compatibility tests to ensure stable operation in coordination with other traffic control means.

[0031] Step S340: Build a dynamic allocation model for right of way for the emergency event mode. Specifically, to build a dynamic allocation model for right of way for the emergency event mode, first, collect information such as the type of emergency event, location, surrounding traffic conditions, and the position of emergency vehicles through the alarm system, traffic monitoring equipment, and V2X communication technology, and evaluate the severity of the event and analyze the traffic conditions. Then build the model, set the goal of ensuring the rapid arrival of emergency vehicles and reducing interference with normal traffic, define the path selection and signal timing adjustment as decision variables, set constraints such as traffic flow, safety, and laws, and solve using mathematical programming or intelligent algorithms. Input the relevant information into the model for solution to generate specific allocation strategies, including navigating for emergency vehicles and adjusting signal timing. During the implementation of the strategy, monitor the driving of emergency vehicles and the execution of signals in real time, and dynamically adjust the strategy according to the results. After the event ends, evaluate the effect using indicators such as the arrival time of emergency vehicles, and optimize the model based on this. Finally, integrate the model with the existing system, train relevant personnel, and improve the collaborative work efficiency and the ability of personnel to respond to emergency events.

[0032] Step S400: Deploy edge computing nodes to make real-time dynamic adjustments to the traffic light controller, generate traffic light cycle parameters, phase difference parameters, and green signal ratio parameters based on the dynamic timing strategy, and send a recommended driving speed instruction to the vehicle terminal through the vehicle-road collaborative system. Specifically, collect data such as traffic flow and geographical information and preprocess them, reasonably deploy edge computing nodes, and build a vehicle-road collaborative system. Then identify the traffic scenario based on the collected data, select a suitable strategy from the strategy library and optimize it, and generate traffic light cycle, phase difference, and green signal ratio parameters based on this. The edge computing node sends these parameters to the traffic light controller, and the controller updates the control strategy. At the same time, the node monitors and feeds back in real time for dynamic adjustment. After that, the node fuses and analyzes traffic, traffic light, and vehicle position information, calculates the recommended driving speed of the vehicle, generates an instruction and sends it to the vehicle terminal, and the vehicle responds and feeds back. Finally, regularly evaluate the system performance using indicators such as the average delay time, and optimize the dynamic timing strategy, etc. based on the results to improve the system performance and traffic operation efficiency.

[0033] In a possible implementation, edge computing nodes are deployed to dynamically adjust traffic lights in real time. Based on the dynamic timing strategy, traffic light cycle parameters, phase difference parameters, and green signal ratio parameters are generated, and a recommended driving speed instruction is sent to vehicle terminals through the vehicle-road collaborative system. Step S400 further includes step S410, which designs a lightweight model compression algorithm to reduce the number of parameters of the prediction model. Specifically, when designing a lightweight model compression algorithm to reduce the number of parameters of the prediction model to less than 30% of the original model, preliminary preparations need to be done first, including collecting the prediction model and its structure information and a representative data set and preprocessing them, and determining evaluation indicators such as the number of parameters and accuracy. Then, common algorithms such as pruning, quantization, and knowledge distillation are investigated and selected according to the model characteristics, application requirements, and hardware limitations. After that, customized design is carried out to determine pruning strategies, quantization schemes, knowledge distillation forms, etc., and the hyperparameter search optimization algorithm is used. The model compression is implemented, and then the compressed model is trained or fine-tuned with the data set. The performance is evaluated using the test data set, and it is checked whether the number of parameters meets the standard. If not, the parameters are adjusted or the algorithm is redesigned. The model is optimized according to the evaluation results, and different algorithm combinations can be tried. Finally, the optimized compressed model is deployed to the actual application environment, appropriate hardware and software frameworks are selected, and the performance and the number of parameters are continuously monitored during the application, and regular evaluation and optimization are carried out to maintain stable performance.

[0034] Step S420 realizes data interaction between the local node and the cloud platform through the API interface. The local node processes millisecond-level real-time control instructions, and the cloud platform performs hour-level policy optimization. Specifically, to realize data interaction between the local node and the cloud platform through the API interface so that the local node processes millisecond-level real-time control instructions and the cloud platform performs hour-level policy optimization, the following steps need to be taken: First, perform requirement analysis and planning, clarify the data types and functional requirements of the interaction between the two parties, and formulate API interface specifications, covering URL structure, request method, data format, and parameter definition. Then, develop the local node and the cloud platform separately. The local node implements the API client and optimizes the real-time processing ability, and the cloud platform builds the API server and realizes hour-level policy optimization. After that, conduct data interaction tests, including unit tests, integration tests, and security tests, to ensure that the system functions, performance, and security meet the standards. Then, deploy the system to the production environment, establish a monitoring and maintenance mechanism, and record logs to troubleshoot problems and conduct audits. Finally, continuously optimize, improve the system through data analysis and user feedback, pay attention to industry trends, and introduce new technologies for upgrading and innovation.

[0035] Step S430: Establish a communication quality assessment module to dynamically switch the transmission channels between LTE-V2X direct communication and 5G cellular communication. Specifically, to establish a communication quality assessment module for dynamically switching the transmission channels between LTE-V2X direct communication and 5G cellular communication, first, it is necessary to conduct a requirements analysis and planning, clarify the communication quality requirements for different application scenarios, consider the device characteristics, and determine the evaluation indicators and thresholds. Then, design and develop the communication quality assessment module, which includes a data acquisition unit to accurately collect indicator data in real time, a data analysis and evaluation unit to preprocess the data and comprehensively evaluate the communication quality, and then integrate and optimize each unit. Next, formulate a dynamic switching mechanism, develop a reasonable switching strategy based on the evaluation results, and design a switching process including steps such as triggering, requesting, accessing, and migrating to ensure smooth and reliable switching. After that, conduct system testing and optimization. First, test in a laboratory simulation scenario, and then conduct on-site testing in an actual scenario. Adjust the system according to the results. Finally, complete the deployment and maintenance. Deploy the system to actual devices and networks, train personnel, establish a monitoring mechanism, promptly handle faults, upgrade and optimize the system to adapt to new requirements.

[0036] Step S500: Establish a closed-loop feedback mechanism to perform online parameter correction and model iteration update on the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators. Specifically, to establish a closed-loop feedback mechanism for performing online parameter correction and model iteration update on the dynamic timing strategy, it is necessary to first collect high-precision trajectory data and traffic efficiency indicator data, collect and store them using sensor devices and communication networks, and perform preprocessing such as cleaning and normalization. Then, establish a dynamic timing strategy model, determine the structure and initialize the parameters. Design a closed-loop feedback mechanism, use the traffic efficiency indicator as the feedback indicator, regularly compare the actual value with the target value, and trigger an adjustment when the threshold is exceeded. Then, perform online parameter correction and model iteration update. Adjust the parameters according to the feedback results, update the model with new data, and evaluate the performance. Regularly evaluate the system using indicators such as the improvement degree of traffic efficiency, and accordingly optimize the feedback mechanism, model, and parameter correction method. Finally, deploy the system to the actual traffic management system, train personnel, establish a monitoring mechanism, promptly handle faults, upgrade and optimize to adapt to traffic changes.

[0037] In a possible implementation, a closed-loop feedback mechanism is established to perform online parameter correction and model iterative update on the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators. Step S500 further includes step S510, which defines a comprehensive traffic efficiency evaluation indicator, including three-dimensional parameters of average delay time, number of stops, and fuel consumption. Specifically, to define a comprehensive traffic efficiency evaluation indicator including three-dimensional parameters of average delay time, number of stops, and fuel consumption, it is first necessary to clarify the basis for determining the indicator. These three dimensions can comprehensively evaluate traffic efficiency from the perspectives of traffic flow, obstruction, and energy consumption. Then, data is collected through multiple channels. Data on average delay time and number of stops are obtained using sensors and video monitoring, and data on fuel consumption is estimated from the OBD interface or through a vehicle dynamics model. Then, the indicator is quantified, with the three indicators measured in seconds, integers, and liters respectively, and standardized using min-max normalization or Z-score normalization. Next, the weights of each indicator are reasonably determined through the expert scoring method or the analytic hierarchy process. Finally, the standardized parameters of each dimension are multiplied by the corresponding weights and summed to construct a comprehensive traffic efficiency evaluation indicator. The formula is E = w1×x1 + w2×x2 + w3×x3, which provides a strong basis for traffic system evaluation and optimization.

[0038] Step S520, develop a simulation verification platform based on digital twin to conduct virtual intersection stress tests on the new strategy. Specifically, to develop a simulation verification platform based on digital twin to conduct virtual intersection stress tests on the new strategy, it is first necessary to carry out preliminary planning and requirements analysis to clarify the test objectives, requirements, simulation scope, and accuracy. Then, collect data on the geometry, traffic facilities, traffic flow, etc. of the physical intersection and clean and preprocess it. Then, construct a digital twin model, including physical entity modeling and realizing data mapping and synchronization. After that, carry out the development of the simulation verification platform, design the architecture and develop functional modules such as new strategy import, simulation scenario setting, calculation, result analysis, and visualization. Implement virtual intersection stress tests, design different scenarios, import the new strategy for testing and compare and analyze it with the benchmark strategy. Then, verify the platform, compare it with actual historical data, and adjust and optimize it if the difference is large. Further optimize the platform according to the test and feedback. Finally, compile a user manual, test report, and technical documentation, deliver the platform and documentation to users, and provide training and technical support.

[0039] Step S530: Build a self-healing system for abnormal working conditions. When it is detected that the deviation of policy execution exceeds the threshold, automatically roll back to the previous stable version. Specifically, to build a self-healing system for abnormal working conditions to achieve automatic rollback to the previous stable version when the deviation of policy execution exceeds the threshold, the following steps need to be taken: First, conduct system requirement analysis and planning, clarify the application scenario and objectives, define the policy and deviation indicators, and set the threshold. Then, build a data collection and processing system, deploy collection devices, establish a transmission network, preprocess and store the data. Next, develop a deviation detection module, select a detection algorithm and implement optimization, and conduct real-time monitoring and alarm. Build a policy version management system, define the version identifier, determine the stable version, and establish a backup and recovery mechanism. Implement an automatic rollback mechanism, set the trigger condition, design the rollback process and test and verify it. After that, conduct system integration and testing, covering functional and performance testing. Finally, deploy the system to the production environment, establish a monitoring and maintenance mechanism, and continuously improve the system to enhance the self-healing ability and reliability.

[0040] In a possible implementation, establish a closed-loop feedback mechanism. Based on high-precision trajectory data and traffic efficiency indicators, perform online parameter correction and model iteration update on the dynamic signal timing strategy. Step S500 further includes step S540: Establish a hierarchical response mechanism, and divide the traffic state into four levels of early warning: red, orange, yellow, and blue. Specifically, to establish a hierarchical response mechanism that divides the traffic state into four levels of early warning: red, orange, yellow, and blue, the following steps need to be taken: First, collect and integrate multi-source data. Use devices such as geomagnetic sensors and video surveillance to collect traffic flow data. At the same time, collect relevant information such as weather, special events, and road construction, and build a data storage platform for management. Then, determine the early warning thresholds for each level. By mining historical data and simulating predictions, set the blue early warning as less than 30% of the normal daily average traffic flow, the yellow early warning as 30%-60%, the orange early warning as 60%-90%, and the red early warning as more than 90%. Next, develop an early warning monitoring system, design the architecture, implement the algorithm, and conduct test optimization. Then, formulate a response mechanism, clarify the measures corresponding to different levels of early warning, such as strengthening monitoring for blue early warning and launching a comprehensive emergency for red early warning. At the same time, determine the responsibility division and coordination mechanism for each department. After that, release standardized early warning information through diversified channels such as traffic radio and electronic display screens, and establish a public feedback and communication mechanism. Finally, regularly evaluate the operation effect of the system, formulate and implement improvement measures according to the results, and continuously improve the scientificity and effectiveness of the mechanism.

[0041] Step S550, initiate regional linkage control for the red warning status, and coordinate all signal lights within the predefined range to enter the emergency mode. Specifically, to initiate regional linkage control for the red warning status within a 5-kilometer surrounding area and coordinate the signal lights to enter the emergency mode, the following steps need to be taken: conduct data collection and analysis in the early stage, covering traffic and geographical information data, evaluate and upgrade the signal light system, and build a communication network; set up a red warning trigger mechanism, determine the warning threshold, and have the intelligent monitoring system judge in real time; design the emergency mode, optimize the signal light timing, establish a regional coordination control and a special vehicle priority control mechanism; integrate the emergency strategy into the control system and test it. After the warning is triggered, initiate the linkage control, and at the same time send traffic police to the scene for command; establish an evaluation index system to evaluate the effect in real time and provide feedback, and continuously optimize the control plan; conduct training for relevant personnel to improve the emergency response ability, and publicize through the media to improve the cooperation degree, and jointly deal with traffic congestion.

[0042] Step S560, develop a driver behavior prediction module, combine historical trajectory data to predict risk behaviors such as running a red light and cutting in line, and adjust the phase in advance. Specifically, to develop a driver behavior prediction module to combine historical trajectory data to predict risk behaviors such as running a red light and cutting in line and adjust the phase in advance, the following steps need to be taken: first, conduct requirement analysis and planning, clarify the business objectives, functional requirements, and formulate a project plan; then collect historical trajectory and risk behavior annotation data, clean and extract features, and divide the data set; then select a suitable machine learning or deep learning model, train it using the training set, and optimize the model in combination with the validation set; integrate the trained module into the signal light control system, and conduct functional and performance tests; finally, deploy the system, establish a real-time monitoring and feedback mechanism, and continuously optimize the model and system according to the actual operation situation to adapt to the changing traffic environment and driver behavior patterns.

[0043] The embodiment of the present application adopts the method of collecting multi-source traffic data through roadside devices, performing spatio-temporal alignment and outlier filtering; constructing a deep learning prediction model based on the filtered data, combining historical and real-time vehicle networking information to predict traffic demand; generating a dynamic timing strategy according to the prediction, synchronizing it to associated intersections through V2X; deploying edge computing nodes to adjust traffic light parameters in real time and send recommended speeds to vehicles; establishing a closed-loop feedback mechanism, and online correcting the strategy and model based on trajectory data and efficiency indicators, achieving the technical effect of significantly improving traffic efficiency and shortening the emergency response delay in complex traffic scenarios.

[0044] In the above text, with reference to Figure 1 the intelligent traffic light adjustment method integrating vehicle networking information according to the embodiment of the present invention is described in detail. Next, with reference to Figure 2 the intelligent traffic light adjustment system integrating vehicle networking information according to the embodiment of the present invention will be described.

[0045] The intelligent traffic light adjustment system integrating vehicle networking information according to an embodiment of the present invention is used to solve the technical problem that the existing traffic light timing system cannot fuse multi-source vehicle networking data in real time for accurate demand prediction and adaptive adjustment in a dynamic traffic environment, achieving the technical effect of significantly improving the traffic efficiency and shortening the emergency response delay in complex traffic scenarios. The intelligent traffic light adjustment system integrating vehicle networking information includes: Description of reference numerals: Multi-source traffic data acquisition module 10, traffic demand prediction module 20, dynamic timing strategy generation module 30, edge computing node deployment module 40, closed-loop feedback mechanism construction module 50.

[0046] The multi-source traffic data acquisition module 10 is used to collect multi-source traffic data in real time through roadside sensing devices, perform spatio-temporal alignment and outlier filtering on the multi-source traffic data, and obtain the filtered multi-source traffic data.

[0047] The traffic demand prediction module 20 is used to construct a multi-modal traffic state prediction model based on deep learning with the filtered multi-source traffic data, and combine historical traffic data and real-time vehicle networking information to predict the traffic demand in each direction within a future variable time window.

[0048] The dynamic timing strategy generation module 30 is used to generate a dynamic timing strategy according to the prediction result, and synchronize the strategy parameters to associated intersections through the V2X communication protocol.

[0049] The edge computing node deployment module 40 is used to deploy edge computing nodes to perform real-time dynamic adjustment on traffic light controllers, generate traffic light cycle parameters, phase difference parameters, and green signal ratio parameters based on the dynamic timing strategy, and send recommended driving speed instructions to vehicle terminals through the vehicle-road coordination system.

[0050] The closed-loop feedback mechanism construction module 50 is used to establish a closed-loop feedback mechanism, and perform online parameter correction and model iteration update on the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators.

[0051] Next, the specific configuration of the multi-source traffic data acquisition module 10 will be described in detail. As described above, multi-source traffic data is collected in real time by roadside sensing devices, and the multi-source traffic data is subjected to spatio-temporal alignment and outlier filtering to obtain the filtered multi-source traffic data. The multi-source traffic data acquisition module 10 further includes: a grouped data bit difference sequence acquisition unit, which is used to group the original vehicle network data according to the time series, calculate the interquartile range and standard deviation of each grouped data, and obtain the grouped data bit difference sequence; an outlier data point identification unit, which is used to identify the outlier data points of the grouped data bit difference sequence based on the sliding window mechanism, and use the cubic spline interpolation method to compensate for the missing data; a data verification unit, which is used to perform data fusion and credibility verification on the conflicting data collected by the radar and the camera through the Kalman filter algorithm to obtain the filtered multi-source traffic data.

[0052] Next, the specific configuration of the traffic demand prediction module 20 will be described in detail. As described above, a multi-modal traffic state prediction model based on deep learning is constructed with the filtered multi-source traffic data, and the traffic demand in each direction within a future variable time window is predicted by combining historical traffic data and real-time vehicle network information. The traffic demand prediction module 20 further includes: a spatio-temporal graph convolutional network construction unit, which is used to establish a spatio-temporal graph convolutional network to process the road topology features; a temporal dependence relationship construction unit, which is used to model the temporal dependence relationship of traffic flow using a long short-term memory network; a vehicle intention information integration unit, which is used to integrate the vehicle intention information in the V2X communication data by dynamically weighting the attention mechanism; a multi-modal traffic state prediction model generation unit, which is used to optimize the multi-objective loss function through a reinforcement learning framework, balance the traffic efficiency and safety indicators, and generate a multi-modal traffic state prediction model.

[0053] Next, the specific configuration of the dynamic timing strategy generation module 30 will be described in detail. As described above, a dynamic timing strategy is generated according to the prediction results, and the strategy parameters are synchronized to the associated intersections through the V2X communication protocol. The dynamic timing strategy generation module 30 further includes: a mode division unit, which is used to divide the traffic scenario into three modes: normal traffic, peak congestion, and emergency events; a green wave band optimization unit, which is used to adopt the green wave band optimization method for the normal traffic mode; a cooperative control mechanism activation unit, which is used to enable the multi-intersection cooperative control mechanism for the peak congestion mode; a right-of-way dynamic allocation model construction unit, which is used to construct a priority right-of-way dynamic allocation model for the emergency event mode.

[0054] Next, the specific configuration of the edge computing node deployment module 40 will be described in detail. As described above, deploying edge computing nodes enables real-time dynamic adjustment of traffic signal controllers, generating traffic signal cycle parameters, phase difference parameters, and green signal ratio parameters based on the dynamic timing strategy, and sending recommended driving speed instructions to vehicle terminals through the vehicle-road cooperation system. The edge computing node deployment module 40 further includes: a lightweight model compression algorithm design unit for designing a lightweight model compression algorithm to reduce the number of parameters of the prediction model; a data interaction unit for realizing data interaction between the local node and the cloud platform through an API interface, where the local node processes millisecond-level real-time control instructions and the cloud platform performs hour-level policy optimization; a communication quality evaluation module construction unit for establishing a communication quality evaluation module to dynamically switch between LTE-V2X direct communication and 5G cellular communication transmission channels.

[0055] Next, the specific configuration of the closed-loop feedback mechanism construction module 50 will be described in detail. As described above, a closed-loop feedback mechanism is established to perform online parameter correction and model iteration update on the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators. The closed-loop feedback mechanism construction module 50 further includes: an evaluation index definition unit for defining a comprehensive traffic efficiency evaluation index, including three-dimensional parameters of average delay time, stop times, and fuel consumption; a simulation verification platform development unit for developing a simulation verification platform based on digital twin to perform virtual intersection stress tests on new strategies; an abnormal condition self-healing system construction unit for constructing an abnormal condition self-healing system that automatically rolls back to the previous stable version when it detects that the strategy execution deviation exceeds the threshold.

[0056] Among them, the closed-loop feedback mechanism construction module 50 further includes: a hierarchical response mechanism construction unit for establishing a hierarchical response mechanism to divide traffic states into four levels of red, orange, yellow, and blue early warnings; a regional linkage control activation unit for activating regional linkage control for the red early warning state to coordinate all traffic signals within a predefined range to enter the emergency mode; a driver behavior prediction module development unit for developing a driver behavior prediction module to predict risk behaviors such as running red lights and cutting in line in combination with historical trajectory data and adjust the phase in advance.

[0057] The traffic signal intelligent adjustment system integrating vehicle networking information provided by the embodiments of the present invention can execute the traffic signal intelligent adjustment method integrating vehicle networking information provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0058] Although this application makes various references to certain modules in the system according to the embodiments of this application, however, any number of different modules can be used and run on the user terminal and / or the server. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual distinction and are not used to limit the protection scope of the present invention.

[0059] The above specific embodiments do not constitute a limitation on the protection scope of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A traffic light intelligent adjustment method integrating Internet of Vehicles information, characterized in that: The method comprises: Collect multi-source traffic data in real time through roadside sensing equipment, perform spatiotemporal alignment and outlier filtering on the multi-source traffic data, and obtain filtered multi-source traffic data; A multimodal traffic status prediction model based on deep learning is constructed with filtered multi-source traffic data, and historical traffic data and real-time Internet of Vehicles information are combined to predict traffic demand in all directions within a future variable time window; Generate dynamic timing strategies based on prediction results and synchronize strategy parameters to associated intersections via V2X communication protocol; Deploy edge computing nodes to dynamically adjust the traffic light controller in real time, generate traffic light cycle parameters, phase difference parameters and green-to-signal ratio parameters based on the dynamic timing strategy, and send recommended driving speed instructions to the vehicle terminal through the vehicle-road cooperative system; A closed-loop feedback mechanism is established to perform online parameter correction and model iterative update of the dynamic timing strategy based on high-precision trajectory data and traffic efficiency indicators.

2. The intelligent traffic light adjustment method integrating Internet of Vehicles information according to claim 1 is characterized in that: The outlier filtering includes: The original data of the Internet of Vehicles is grouped according to the time series, and the interquartile range and standard deviation of each group data are calculated to obtain the group data difference sequence; Identifying outlier data points of the grouped data difference sequence based on a sliding window mechanism, and compensating for missing data using a cubic spline interpolation method; The conflict data collected by radar and camera are fused and verified for credibility through Kalman filtering algorithm to obtain filtered multi-source traffic data.

3. The intelligent traffic light adjustment method integrating Internet of Vehicles information according to claim 1 is characterized in that: The construction of the multimodal traffic state prediction model includes: Establish a spatiotemporal graph convolutional network to process road topology features; Long short-term memory network is used to model the temporal dependency of traffic flow; Integrate attention mechanism to dynamically weight vehicle intention information in V2X communication data; Through the reinforcement learning framework, the multi-objective loss function is optimized, the traffic efficiency and safety indicators are balanced, and a multimodal traffic status prediction model is generated.

4. The intelligent traffic light adjustment method integrating Internet of Vehicles information according to claim 1 is characterized in that: The generation of the dynamic timing strategy includes: Divide traffic scenarios into three modes: normal traffic, peak congestion, and emergency events; A green wave band optimization method is used for normal traffic patterns; Enable multi-intersection coordinated control mechanism for peak congestion mode; Construct a dynamic right-of-way allocation model for emergency event modes.

5. The intelligent traffic light adjustment method integrating Internet of Vehicles information according to claim 1 is characterized in that: The deployment of the edge computing node includes: Design a lightweight model compression algorithm to reduce the number of prediction model parameters; The data interaction between the local node and the cloud platform is realized through the API interface. The local node processes millisecond-level real-time control instructions, and the cloud platform performs hour-level strategy optimization; Establish a communication quality assessment module to dynamically switch the LTE-V2X direct communication and 5G cellular communication transmission channels.

6. The intelligent traffic light adjustment method integrating Internet of Vehicles information according to claim 1 is characterized in that: The closed-loop feedback mechanism includes: Define comprehensive evaluation indicators for traffic efficiency, including three-dimensional parameters: average delay time, number of stops, and fuel consumption; Develop a simulation verification platform based on digital twins to conduct virtual intersection stress tests on new strategies; Build a self-healing system for abnormal conditions, and automatically roll back to the previous stable version when it detects that the policy execution deviation exceeds the threshold.

7. The intelligent traffic light adjustment method integrating Internet of Vehicles information according to claim 1 is characterized in that: The method further comprises: Establish a hierarchical response mechanism to divide traffic conditions into four levels of warning: red, orange, yellow, and blue; Initiate regional linkage control for red warning status and coordinate all traffic lights within a predefined range to enter emergency mode; Develop a driver behavior prediction module, combine historical trajectory data to predict risky behaviors such as cutting in and squeezing in, and adjust the phase in advance.

8. The intelligent traffic light adjustment system integrating vehicle network information is characterized by: The system is used to implement the intelligent traffic light adjustment method integrating Internet of Vehicles information according to any one of claims 1 to 7, and the system includes: A multi-source traffic data acquisition module, which is used to collect multi-source traffic data in real time through roadside sensing equipment, perform spatiotemporal alignment and outlier filtering on the multi-source traffic data, and obtain filtered multi-source traffic data; A traffic demand prediction module, which is used to construct a multimodal traffic state prediction model based on deep learning with filtered multi-source traffic data, and to predict traffic demand in various directions within a future variable time window by combining historical traffic data and real-time Internet of Vehicles information; A dynamic timing strategy generation module, the dynamic timing strategy generation module is used to generate a dynamic timing strategy according to the prediction results, and synchronize strategy parameters to the associated intersection through the V2X communication protocol; An edge computing node deployment module, which is used to deploy edge computing nodes to perform real-time dynamic adjustment on the traffic light controller, generate traffic light cycle parameters, phase difference parameters and green-to-signal ratio parameters based on the dynamic timing strategy, and send a recommended driving speed instruction to the vehicle terminal through the vehicle-road cooperative system; A closed-loop feedback mechanism building module is used to establish a closed-loop feedback mechanism, and based on high-precision trajectory data and traffic efficiency indicators, perform online parameter correction and model iterative update on the dynamic timing strategy.

Images