Traffic signal lamp control method capable of predicting traffic flow in real time

Abstract

The invention relates to the technical field of traffic control, and discloses a traffic signal lamp control method capable of performing real-time prediction on traffic flow, and the method comprises the steps: S1, collecting multi-source traffic flow real-time monitoring data through roadside sensing equipment, S2, carrying out the data preprocessing and feature fusion processing of the multi-source traffic flow real-time monitoring data, and S3, carrying out the feature fusion processing of the multi-source traffic flow real-time monitoring data. Generating a standardized traffic flow feature vector; by verifying the consistency of multi-source sensor data in real time and detecting a data deviation problem caused by illumination shielding or equipment abnormity, the system can eliminate sensor input distortion, ensure the input accuracy of a traffic flow prediction model, further reduce signal control errors, improve the overall control reliability, and improve the traffic flow prediction accuracy. Through adaptive adjustment of model weight parameters and real-time correction of prediction deviation, the system can prevent a control strategy from lagging behind traffic evolution, ensure timely update of a signal phase sequence, reduce the risk of queuing overflow at an intersection, and further enhance adaptive control capability.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic control, and in particular to a traffic signal light control method capable of performing real-time prediction of vehicle flow. Background Art

[0002] Traffic flow is based on the number of vehicles passing through a certain road section in unit time, which is the number of vehicles passing through the road section within a certain period of time. Traffic lights are signal lights that direct traffic operations, generally consisting of red lights, green lights, and yellow lights. Traffic control uses traffic control facilities managed by electronic computers to limit, regulate, induce, and divert traffic flows in order to reduce the total amount of traffic, ease traffic, and ensure traffic safety and smoothness.

[0003] At present, in urban traffic signal control systems, due to the high dynamics and spatiotemporal heterogeneity of traffic flow, existing technologies have defects in achieving real-time prediction of traffic flow: when collecting traffic flow data, roadside sensors are affected by weather conditions, obstructions and equipment aging, and are unable to perceive subtle changes in traffic flow status in real time, resulting in systematic deviations in the collected traffic flow parameters; abnormal fluctuations in sensor data may cause distortion of the input of the traffic flow prediction model, which in turn may cause a serious deviation between the signal timing plan and the actual traffic demand, making it impossible to guarantee the accuracy of signal control; at the same time, when establishing a traffic flow prediction model, traditional methods are difficult to be effective. The system cannot effectively integrate multi-source heterogeneous data and cannot capture the nonlinear propagation characteristics of traffic flow in complex road networks, resulting in insufficient accuracy in short-term traffic flow prediction. When traffic conditions suddenly change, the prediction model lacks an adaptive correction mechanism and cannot adjust the signal phase sequence in a timely manner, causing the control strategy to lag behind the actual traffic evolution and exacerbating the risk of queue overflow at intersections. In addition, when performing coordinated control of multiple intersections, the existing system cannot achieve global optimization of signal timing parameters due to the lack of quantitative analysis of the spatiotemporal correlation of regional traffic flows. It can only perform local single-point adjustments, resulting in low traffic efficiency on arterial roads and unsatisfactory green wave belt coordination effects, further reducing the overall operating efficiency of the road network.

[0004] Therefore, a traffic light control method that can perform real-time prediction of traffic flow is proposed to solve the above problems. Summary of the Invention

[0005] (1) Technical problems solved In view of the shortcomings of the existing technology, the present invention provides a traffic light control method that can predict traffic flow in real time, solving the problems raised in the above background technology.

[0006] (2) Technical solution To achieve the above objectives, the present invention provides the following technical solution: a traffic light control method capable of real-time prediction of traffic flow, the method comprising the following steps: S1. Collect multi-source traffic flow real-time monitoring data through roadside sensing equipment; S2. performing data preprocessing and feature fusion processing on the multi-source traffic flow real-time monitoring data to generate a standardized traffic flow feature vector; S3. Build a time series prediction model, perform traffic flow prediction processing for future periods based on the standardized traffic flow feature vector, and generate traffic flow prediction parameters; S4. Performing signal light phase conflict analysis based on the traffic flow prediction parameters to generate phase conflict matrix data; S5. Calculating and processing a signal timing scheme based on the phase conflict matrix data and a traffic optimization objective function to generate dynamic signal timing parameters; S6. Executing the dynamic signal timing parameters through the traffic signal controller to achieve dynamic allocation of signal light phase time; S7: Monitor traffic flow status changes in real time and feed back to step S1 to form a closed-loop control mechanism.

[0007] Preferably, said S1 comprises the following steps: S11. Collect traffic video stream data through roadside camera equipment and use the YOLOv7 target detection algorithm to extract vehicle position and speed parameters; S12, collecting lane-level vehicle presence status data through a geomagnetic sensor array to generate a vehicle occupancy time series; S13. Collect vehicle trajectory data through microwave radar equipment to generate a traffic flow speed distribution feature vector.

[0008] Preferably, said S2 comprises the following steps: S21, performing background difference noise reduction processing on the traffic video stream data to generate denoised video frame data; S22, performing motion trajectory smoothing processing on the vehicle trajectory data using a Kalman filter algorithm to generate standardized trajectory data; S23. Perform spatiotemporal alignment and fusion of the vehicle occupancy time series, the standardized trajectory data, and the vehicle speed parameters to generate a five-dimensional traffic flow feature tensor.

[0009] Preferably, said S3 comprises the following steps: S31. Construct a bidirectional LSTM prediction model based on the attention mechanism, and set the number of input layer nodes to 128 dimensions; S32, inputting the five-dimensional traffic flow feature tensor lasting 15 minutes into the prediction model to perform time series sliding prediction; S33. Output the predicted traffic flow value and confidence parameter for each lane in the next 5 minutes. The prediction model updates the hidden state through the LSTM forget gate calculation unit. The calculation formula is:

[0010] in is the output value of the forget gate at time step t, ranging from [0,1] and representing the forget weight of historical information, is the Sigmoid activation function, is the weight matrix of the forget gate, For the previous time step The hidden state vector of is the input feature vector of the current time step t, is the bias vector of the forget gate; The output of the forget gate is used to control the weight of retaining historical information and calculate the traffic flow prediction value based on it.

[0011] Preferably, said S4 comprises the following steps: S41, analyzing the road network topology data and establishing a relationship map of intersection conflict points; S42, calculating the traffic pressure index of each phase according to the traffic flow prediction value; S43. Generate a phase-conflict binary matrix based on the conflict point relationship map and the traffic pressure index.

[0012] Preferably, the S5 comprises the following steps: S51. Set the traffic optimization objective function as maximizing comprehensive utility, and balance delay and traffic volume through weighted coefficients:

[0013] in is the comprehensive utility value, is the traffic weight coefficient, is the delay sensitivity coefficient, is the total delay value, is the traffic conversion coefficient, To predict traffic volume, It is a natural constant and is suitable for constructing an exponential decay relationship; S52, using the NSGA-II multi-objective optimization algorithm to solve the Pareto optimal solution set, with the comprehensive utility value as the fitness function; S53. Filter the optimal signal timing scheme from the Pareto solution set using the entropy weight method, and output the green light duration, phase difference, and cycle duration parameters.

[0014] Preferably, the S6 comprises the following steps: S61, encoding the dynamic signal timing parameters into a Siemens TLS-ATC standard control instruction set; S62, transmitting the command to the intersection signal controller via the optical fiber private network; S63: Real-time monitoring of the signal execution deviation. When the deviation exceeds 5%, the timing parameters are recalculated.

[0015] Preferably, it also includes: S71. Establish a regional collaborative control mechanism to obtain queue length data at adjacent intersections through V2X communication equipment; S72. When a queue overflow risk is detected, the green wave band collaborative control mode is started; S73. Calculate the coordinated phase offset based on the fleet discrete model and generate regional coordinated control parameters.

[0016] Preferably, an exception handling mechanism is also included: S81, real-time monitoring of equipment failure rate, when the data loss rate exceeds 20% switch to the historical mode database; S82, using the KNN algorithm to match backup control schemes from historical similar scenarios; S83: Start the redundant control channel and switch to the backup communication link to transmit the control instruction.

[0017] Preferably, it also includes a traffic flow adaptive optimization mechanism: S91, collecting dynamic change parameters from multi-source traffic flow monitoring data in real time to generate a traffic flow change rate indicator; S92. When the traffic flow change rate index exceeds a preset threshold, triggering a phase response adjustment process to calculate a green light time increment parameter for each phase; S93, updating dynamic signal timing parameters based on the green light time increment parameter, including optimizing the green light duration distribution weight using a gradient descent algorithm, correcting the phase difference offset based on historical intersection traffic efficiency data, generating an adaptive timing update instruction, and transmitting it to the signal controller; S94. After executing the adaptive timing update, evaluate the traffic delay improvement rate and queue length reduction rate in real time; S95. When the improvement rate is lower than the target value, go back to step S92 to perform iterative optimization until the preset traffic optimization target is met.

[0018] (3) Beneficial effects Compared with the existing technology, the present invention provides a traffic light control method that can predict traffic flow in real time, which has the following beneficial effects: 1. In the present invention, by setting up a data calibration terminal, when collecting vehicle flow data, the consistency of multi-source sensor data is verified in real time, and data deviation problems caused by light obstruction or equipment abnormalities are detected. This enables the system to eliminate sensor input distortion, ensure the input accuracy of the vehicle flow prediction model, further reduce signal control errors, and improve overall control reliability.

[0019] 2. In the present invention, by setting up a model optimization end, when the traffic flow prediction model is running, the model weight parameters are adaptively adjusted to respond to sudden changes in traffic flow and correct the prediction deviation in real time, so that the system can avoid the control strategy lagging behind traffic evolution, ensure the timely update of the signal phase sequence, reduce the risk of queue overflow at the intersection, and further enhance the adaptive control capability.

[0020] 3. In the present invention, by setting up a collaborative control terminal, when optimizing the signal timing of multiple intersections, global coordination is achieved by quantifying the spatiotemporal correlation of regional traffic flows and integrating local single-point data, so that the system can eliminate the problem of collaborative failure, ensure the coordination effect of the green wave belt, improve the traffic efficiency of the road network, and optimize the smoothness of the main traffic flow. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The present invention is a flow chart of a traffic signal light control method capable of performing real-time prediction of vehicle flow. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figure 1 A traffic light control method capable of real-time prediction of traffic flow comprises the following steps: S1. Collect multi-source traffic flow real-time monitoring data through roadside sensing equipment; S2. Perform data preprocessing and feature fusion processing on multi-source traffic flow real-time monitoring data to generate standardized traffic flow feature vectors; S3. Build a time series prediction model, perform traffic flow prediction processing for future periods based on the standardized traffic flow feature vector, and generate traffic flow prediction parameters; S4. Perform signal light phase conflict analysis based on traffic flow prediction parameters to generate phase conflict matrix data; S5. Calculate and process the signal timing scheme based on the phase conflict matrix data and the traffic optimization objective function to generate dynamic signal timing parameters; S6. Implement dynamic signal timing parameters through the traffic signal controller to achieve dynamic allocation of signal light phase time; S7: Monitor traffic flow status changes in real time and feed back to step S1 to form a closed-loop control mechanism.

[0024] S1 includes the following steps: S11. Collect traffic video stream data through roadside camera equipment and use the YOLOv7 target detection algorithm to extract vehicle position and speed parameters; S12, collecting lane-level vehicle presence status data through a geomagnetic sensor array to generate a vehicle occupancy time series; S13. Collect vehicle trajectory data through microwave radar equipment to generate a traffic flow speed distribution feature vector.

[0025] S2 includes the following steps: S21, performing background difference noise reduction processing on the traffic video stream data to generate denoised video frame data; S22, performing motion trajectory smoothing processing on the vehicle trajectory data using a Kalman filter algorithm to generate standardized trajectory data; S23. Perform spatiotemporal alignment and fusion of the vehicle occupancy time series, standardized trajectory data, and vehicle speed parameters to generate a five-dimensional traffic flow feature tensor.

[0026] Specifically, S3 includes the following steps: S31. Construct a bidirectional LSTM prediction model based on the attention mechanism, and set the number of input layer nodes to 128 dimensions; S32, inputting the 15-minute five-dimensional traffic flow feature tensor into the prediction model to perform time series sliding prediction; S33. Output the predicted traffic flow value and confidence parameter for each lane in the next 5 minutes. The prediction model updates the hidden state through the LSTM forget gate calculation unit. The calculation formula is:

[0027] in is the output value of the forget gate at time step t, ranging from [0,1] and representing the forget weight of historical information, is the Sigmoid activation function, is the weight matrix of the forget gate, For the previous time step The hidden state vector of is the input feature vector of the current time step t, is the bias vector of the forget gate; The output of the forget gate is used to control the weight of retaining historical information and calculate the traffic flow prediction value based on it.

[0028] S4 includes the following steps: S41, analyzing the road network topology data and establishing a relationship map of intersection conflict points; S42, calculating the traffic pressure index of each phase according to the traffic flow prediction value; S43. Generate a phase-conflict binary matrix based on the conflict point relationship map and the traffic pressure index.

[0029] S5 includes the following steps: S51. Set the traffic optimization objective function as maximizing comprehensive utility, and balance delay and traffic volume through weighted coefficients:

[0030] in is the comprehensive utility value, is the traffic weight coefficient, is the delay sensitivity coefficient, is the total delay value, is the traffic conversion coefficient, To predict traffic volume, It is a natural constant and is suitable for constructing an exponential decay relationship; S52, using the NSGA-II multi-objective optimization algorithm to solve the Pareto optimal solution set, with the comprehensive utility value as the fitness function; S53. Filter the optimal signal timing scheme from the Pareto solution set using the entropy weight method, and output the green light duration, phase difference, and cycle duration parameters.

[0031] S6 includes the following steps: S61, encoding the dynamic signal timing parameters into a Siemens TLS-ATC standard control instruction set; S62, transmitting the command to the intersection signal controller via the optical fiber private network; S63: Real-time monitoring of the signal execution deviation. When the deviation exceeds 5%, the timing parameters are recalculated.

[0032] Also includes: S71. Establish a regional collaborative control mechanism to obtain queue length data at adjacent intersections through V2X communication equipment; S72. When a queue overflow risk is detected, the green wave band collaborative control mode is started; S73. Calculate the coordinated phase offset based on the fleet discrete model and generate regional coordinated control parameters.

[0033] It also includes exception handling mechanisms: S81, real-time monitoring of equipment failure rate, when the data loss rate exceeds 20% switch to the historical mode database; S82, using the KNN algorithm to match backup control schemes from historical similar scenarios; S83: Start the redundant control channel and switch to the backup communication link to transmit the control instruction.

[0034] It also includes traffic flow adaptive optimization mechanism: S91, collecting dynamic change parameters from multi-source traffic flow monitoring data in real time to generate a traffic flow change rate indicator; S92. When the traffic flow change rate index exceeds a preset threshold, triggering a phase response adjustment process to calculate a green light time increment parameter for each phase; S93, updating dynamic signal timing parameters based on the green light time increment parameter, including optimizing the green light duration distribution weight using a gradient descent algorithm, correcting the phase difference offset based on historical intersection traffic efficiency data, generating an adaptive timing update instruction, and transmitting it to the signal controller; S94. After executing the adaptive timing update, evaluate the traffic delay improvement rate and queue length reduction rate in real time; S95. When the improvement rate is lower than the target value, go back to step S92 to perform iterative optimization until the preset traffic optimization target is met.

[0035] Example 1: Dynamic Perception and Fusion Processing of Multi-Source Traffic Data A comprehensive perception system is deployed at intersections of urban arterial roads, comprising a collaborative data collection network consisting of high-definition vision units, geomagnetic sensing arrays, and millimeter-wave detection equipment. When traffic enters the transition phase of the evening rush hour, the system initiates a multimodal data fusion process: the vision unit continuously captures vehicle motion trajectories using a dynamic target recognition algorithm, the geomagnetic sensor simultaneously generates a sequence of lane occupancy states, and the millimeter-wave device outputs vehicle spatial position information. A spatiotemporal calibration engine maps heterogeneous data to a unified spatiotemporal coordinate system, eliminating timing misalignments caused by differences in device acquisition frequencies. Nonlinear filtering techniques are used to smooth sudden changes in trajectory, constructing a fused feature set encompassing vehicle velocity distribution, density gradients, and acceleration characteristics.

[0036] Under unusual climatic conditions, the system activates an adaptive environmental compensation mechanism. When the visual unit is disrupted by rain or fog, the system automatically increases the weighting of millimeter-wave data, maintaining the integrity of traffic flow feature extraction through a multi-source complementary mechanism. Dynamic benchmark correction technology also eliminates systematic errors caused by long-term sensor drift, ensuring spatial consistency across various traffic parameters. The resulting standardized traffic flow feature tensor provides a highly confident input basis for short-term forecasts, establishing a stable and reliable data supply chain.

[0037] Example 2: Adaptive Optimization of Traffic Flow Prediction Model Based on the real-time fused feature tensors, a pre-trained time series prediction model is loaded to perform traffic evolution calculations. The model utilizes a bidirectional memory network architecture and embeds an attention allocation mechanism to strengthen feature weights in key directions. In the event of a sudden traffic accident, the system immediately triggers the model's online learning mechanism upon detecting unusual fluctuations in the feature tensor. This mechanism intercepts traffic flow features within the latest time window for incremental training, freezing the base network layer parameters and only fine-tuning the dynamic weight matrix.

[0038] A cross-regional collaborative optimization framework was simultaneously launched, exchanging feature distribution parameters with adjacent intersections via encrypted channels. Knowledge distillation technology was used to compress the model size, preserving key feature recognition capabilities while reducing computational load. After multiple iterative updates, the prediction model accurately captured the patterns of traffic flow diffusion under the influence of accidents and output traffic flow evolution trends covering multiple signal cycles. This optimization process continuously monitors the stability of the prediction results and automatically rolls back to the previous stable version when the confidence level falls below a threshold, ensuring the reliability of the decision-making basis.

[0039] Example 3: Signal Timing Collaborative Optimization and Execution Based on the traffic distribution trends output by the prediction model, a multi-objective optimization engine is activated to generate a signal timing plan. First, the road network topology is analyzed to establish a phase conflict map. The priority of each phase's demand is calculated using the traffic pressure index. Then, guided by optimizing regional traffic efficiency, an evolutionary algorithm is employed to find the Pareto solution for multi-phase coordination, achieving a dynamic balance between minimizing delays and maximizing traffic volume.

[0040] The optimization engine features a built-in resilience adjustment mechanism that automatically adjusts the objective function weights when detecting special vehicle traffic demands. The generated timing plan is filtered through an entropy-weighted decision-making mechanism to form a set of control instructions, including phase duration, cycle parameters, and green wave offset. These instructions are encapsulated using multiple verification protocols and transmitted to the intersection signal via a dedicated communication link. The execution unit monitors the signal switching status in real time. When the laser detection unit detects that the actual queue length deviates from the expected value, it immediately triggers the recalculation of the timing parameters. The system also records execution deviation data to facilitate continuous iteration of the optimization algorithm.

[0041] Example 4: Global Exception Handling and Resilience Control In a regionally coordinated operating environment, a multi-level emergency response system is constructed. When a sensing device at a specific intersection experiences a sudden malfunction, the system automatically switches to historical scenario matching mode: searching a library of historical scenarios with similar spatiotemporal characteristics, mapping alternative data sources using a nearest neighbor algorithm, and simultaneously activating a backup communication channel to transmit downgraded control instructions.

[0042] In the event of a large-scale traffic disruption, a hierarchical coordinated response sequence is initiated: upstream intersections implement traffic interception to prevent backflow, and downstream intersections dynamically create emergency corridors. The vehicle-road cooperative system pushes optimized routes to navigation terminals, directing traffic around core congestion areas. Event features are encrypted and shared throughout the process using a federated learning framework, driving the simultaneous upgrade of prediction models in adjacent areas. After the incident is resolved, the system automatically backtracks on control effectiveness and generates a resilience assessment report encompassing response time, impact range, and recovery speed. This report guides iterative updates to the policy library and system self-improvement.

[0043] Example 5: All-weather closed-loop control verification Full-scale stress testing is initiated in scenarios where extreme weather conditions and special events overlap. When heavy rain and large-scale events occur simultaneously, the perception unit automatically switches to anti-interference mode, the prediction model loads climate compensation parameters, and the optimization engine activates the emergency objective function. By monitoring the road network saturation index in real time, the control strategy's aggressiveness is dynamically adjusted.

[0044] During execution, a dual-channel feedback mechanism is established: a primary channel transmits real-time control instructions, while a secondary channel simultaneously collects execution performance data. When a decrease in regional traffic efficiency is detected, a strategy retrospective analysis is automatically triggered to identify weak links in the control chain. Ultimately, a complete chain of evidence, from data collection to strategy optimization, is formed, validating the system's adaptability and robustness in complex environments and providing an empirical foundation for the continued evolution of the technical solution.

[0045] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0046] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A traffic light control method capable of real-time prediction of traffic flow, characterized by: The method comprises the following steps: S1. Collect multi-source traffic flow real-time monitoring data through roadside sensing equipment; S2. performing data preprocessing and feature fusion processing on the multi-source traffic flow real-time monitoring data to generate a standardized traffic flow feature vector; S3. Build a time series prediction model, perform traffic flow prediction processing for future periods based on the standardized traffic flow feature vector, and generate traffic flow prediction parameters; S4. Performing signal light phase conflict analysis based on the traffic flow prediction parameters to generate phase conflict matrix data; S5. Calculating and processing a signal timing scheme based on the phase conflict matrix data and a traffic optimization objective function to generate dynamic signal timing parameters; S6. Executing the dynamic signal timing parameters through the traffic signal controller to achieve dynamic allocation of signal light phase time; S7: Monitor traffic flow status changes in real time and feed back to step S1 to form a closed-loop control mechanism.

2. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 1, characterized in that: Said S1 comprises the following steps: S11. Collect traffic video stream data through roadside camera equipment and use the YOLOv7 target detection algorithm to extract vehicle position and speed parameters; S12, collecting lane-level vehicle presence status data through a geomagnetic sensor array to generate a vehicle occupancy time series; S13. Collect vehicle trajectory data through microwave radar equipment to generate a traffic flow speed distribution feature vector.

3. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 2, characterized in that: The S2 comprises the following steps: S21, performing background difference noise reduction processing on the traffic video stream data to generate denoised video frame data; S22, performing motion trajectory smoothing processing on the vehicle trajectory data using a Kalman filter algorithm to generate standardized trajectory data; S23. Perform spatiotemporal alignment and fusion of the vehicle occupancy time series, the standardized trajectory data, and the vehicle speed parameters to generate a five-dimensional traffic flow feature tensor.

4. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 3, characterized in that: The S3 includes the following steps: S31. Construct a bidirectional LSTM prediction model based on the attention mechanism, and set the number of input layer nodes to 128 dimensions; S32, inputting the five-dimensional traffic flow feature tensor lasting 15 minutes into the prediction model to perform time series sliding prediction; S33. Output the predicted traffic flow value and confidence parameter for each lane in the next 5 minutes. The prediction model updates the hidden state through the LSTM forget gate calculation unit. The calculation formula is: in is the output value of the forget gate at time step t, ranging from [0,1] and representing the forget weight of historical information, is the Sigmoid activation function, is the weight matrix of the forget gate, For the previous time step The hidden state vector of is the input feature vector of the current time step t, is the bias vector of the forget gate; The output of the forget gate is used to control the weight of retaining historical information and calculate the traffic flow prediction value based on it.

5. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 4, characterized in that: The S4 comprises the following steps: S41, analyzing the road network topology data and establishing a relationship map of intersection conflict points; S42, calculating the traffic pressure index of each phase according to the traffic flow prediction value; S43. Generate a phase-conflict binary matrix based on the conflict point relationship map and the traffic pressure index.

6. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 1, characterized in that: The S5 comprises the following steps: S51. Set the traffic optimization objective function as maximizing comprehensive utility, and balance delay and traffic volume through weighted coefficients: in is the comprehensive utility value, is the traffic weight coefficient, is the delay sensitivity coefficient, is the total delay value, is the traffic conversion coefficient, To predict traffic volume, It is a natural constant and is suitable for constructing an exponential decay relationship; S52, using the NSGA-II multi-objective optimization algorithm to solve the Pareto optimal solution set, with the comprehensive utility value as the fitness function; S53. Filter the optimal signal timing scheme from the Pareto solution set using the entropy weight method, and output the green light duration, phase difference, and cycle duration parameters.

7. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 1, characterized in that: The S6 comprises the following steps: S61, encoding the dynamic signal timing parameters into a Siemens TLS-ATC standard control instruction set; S62, transmitting the command to the intersection signal controller via the optical fiber private network; S63: Real-time monitoring of the signal execution deviation. When the deviation exceeds 5%, the timing parameters are recalculated.

8. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 1, characterized in that: Also includes: S71. Establish a regional collaborative control mechanism to obtain queue length data at adjacent intersections through V2X communication equipment; S72. When a queue overflow risk is detected, the green wave band collaborative control mode is started; S73. Calculate the coordinated phase offset based on the fleet discrete model and generate regional coordinated control parameters.

9. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 1, characterized in that: It also includes exception handling mechanisms: S81, real-time monitoring of equipment failure rate, when the data loss rate exceeds 20% switch to the historical mode database; S82, using the KNN algorithm to match backup control schemes from historical similar scenarios; S83: Start the redundant control channel and switch to the backup communication link to transmit the control instruction.

10. The traffic signal light control method capable of real-time prediction of traffic flow according to claim 1, characterized in that: It also includes traffic flow adaptive optimization mechanism: S91, collecting dynamic change parameters from multi-source traffic flow monitoring data in real time to generate a traffic flow change rate indicator; S92. When the traffic flow change rate index exceeds a preset threshold, triggering a phase response adjustment process to calculate a green light time increment parameter for each phase; S93, updating dynamic signal timing parameters based on the green light time increment parameter, including optimizing the green light duration distribution weight using a gradient descent algorithm, correcting the phase difference offset based on historical intersection traffic efficiency data, generating an adaptive timing update instruction, and transmitting it to the signal controller; S94. After executing the adaptive timing update, evaluate the traffic delay improvement rate and queue length reduction rate in real time; S95. When the improvement rate is lower than the target value, go back to step S92 to perform iterative optimization until the preset traffic optimization target is met.

Images