Adaptive control method and device for traffic signal controller, storage medium and electronic device

Abstract

The invention discloses a traffic signal controller adaptive control method and device, a storage medium and an electronic device. The method comprises the following steps: performing multi-source data fusion and spatial-temporal position sensing graph convolution initialization on real-time traffic monitoring, historical traffic data, meteorological data and urban activity information data; a space-time position sensing graph convolutional network is adopted, geographical and time information is converted into mathematical vectors to construct a dynamic traffic flow prediction model, space-time relevance among traffic nodes is analyzed through the graph convolutional network, and the future traffic flow state is predicted; performing deep reinforcement learning and signal optimization; self-adaptive feedback and system iteration updating are carried out; monitoring and evaluating system performance; the problems that an existing traffic signal controller control method lacks real-time data comprehensive utilization capacity, dynamic traffic condition prediction is insufficient, and signal control strategy optimization is insufficient are solved.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation, and in particular, to a method, device, storage medium and electronic device for adaptive control of traffic signals. Background Art

[0002] In modern urban traffic management, with the acceleration of urbanization and the surge in the number of vehicles, traditional traffic control systems are facing more and more challenges. Traditional traffic signal systems are mainly based on fixed time control or inductive control. Fixed time control sets the signal cycle according to historical traffic flow data, while inductive control adjusts the signal lights after detecting vehicles through sensors (such as geomagnetic and infrared sensors). These systems are simple and easy to use, but lack flexibility and adaptability, often causing unnecessary waiting during non-peak hours or insufficient response during peak traffic hours.

[0003] Existing traffic control systems face several key technical issues when dealing with complex and dynamic traffic environments, including: Lack of comprehensive utilization of real-time data. Traditional traffic control systems mostly rely on preset control plans or simple sensor feedback, which limits the flexibility and efficiency of the system's response to real-time traffic conditions. These systems are usually unable to process and integrate information from different data sources, such as real-time traffic flow, historical data, meteorological conditions, and special events, resulting in an inability to make rapid and effective adjustments when faced with emergencies or unusual changes in traffic patterns.

[0004] Insufficient prediction of dynamic traffic conditions. Existing traffic signal systems often lack the ability to dynamically predict future traffic flows, especially during peak hours and special events (such as major sporting events, public holidays, etc.). Most of these systems are based on static traffic models and cannot accurately predict short-term and long-term changes in traffic flows. Therefore, it is difficult to formulate effective signal control strategies to adapt to these changes, and the result is often traffic congestion and excessive waiting.

[0005] Insufficient optimization of signal control strategies. Although some advanced adaptive traffic control systems can adjust signals based on real-time data, the decisions of these systems usually still rely on traditional optimization algorithms, which are often limited to solving specific optimization problems and are difficult to adapt to the complex and changing actual traffic environment. In addition, the existing systems often do not fully consider factors such as driver behavior and changes in traffic rules during the signal optimization process, resulting in a deviation between the implementation effect and expectations.

[0006] In view of the problems in related technologies, the existing traffic signal control methods lack the ability to comprehensively utilize real-time data, insufficient prediction of dynamic traffic conditions, and insufficient optimization of signal control strategies, and no effective solutions have been proposed yet. Summary of the invention

[0007] The embodiments of the present application provide a traffic signal adaptive control method, device, storage medium and electronic device to at least solve the problems in the related art that the existing traffic signal control method lacks the ability to comprehensively utilize real-time data and is insufficient in predicting dynamic traffic conditions.

[0008] According to an embodiment of the present application, a traffic signal adaptive control method is provided, comprising: Perform multi-source data fusion and spatiotemporal location-aware graph convolution initialization on real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Using a spatiotemporal location-aware graph convolutional network, we convert geographic and temporal information into mathematical vectors to build a dynamic traffic flow prediction model. We use the graph convolutional network to analyze the spatiotemporal correlation between traffic nodes and predict future traffic flow status. Deep reinforcement learning and signal optimization; Adaptive feedback and system iterative updates; System performance monitoring and evaluation.

[0009] Optionally, the multi-source data fusion of real-time traffic monitoring data, historical traffic data, meteorological data and urban activity information data includes: Integrate real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Clean, standardize, and synchronize the above data; The specific data fusion method is: D = synchronize(normalize(clean(T_realtime, T_history, M, C))) in: D: The final integrated and synchronized dataset, used as input into the subsequent graph convolutional network; clean: data cleaning function to remove invalid or missing data; normalize: data standardization function, unifying the format and range of different data; synchronize: data synchronization function, ensuring that all data is aligned on the timeline.

[0010] Optionally, the spatiotemporal position-aware graph convolution initialization includes: Construct a space-time graph to represent the transportation network; Through location embedding technology, the geographical and temporal attributes of each node (i.e., intersection, road segment, etc.) are converted into a mathematically tractable vector format; Spatial position embedding (V_s): Spatial position embedding reflects the location of each node in the geographic space and captures the spatial relationship between nodes through position embedding; V_s = W_s × E_s(node_positions): The spatial encoding function E_s is used to convert the geographic location information of each node into an embedding vector, which is then adjusted through the weight matrix W_s to adapt to the specific network architecture; Temporal position embedding (V_t): Temporal position embedding captures the periodicity and sequentiality of time; V_t = W_t × E_t(time_steps): Similarly, the time encoding function E_t converts time information (such as hours or weeks) into an embedding vector, and the weight matrix W_t is used to adjust the embedding vector to match the network structure.

[0011] Optionally, a spatiotemporal location-aware graph convolutional network is used to convert geographic and temporal information into mathematical vectors to construct a dynamic traffic flow prediction model. The spatiotemporal correlation between various traffic nodes is analyzed through the graph convolutional network, and the future traffic flow state is predicted, including: Using spatiotemporal location-aware relational reasoning to adaptively infer the relationship weights between nodes in different time and space, capturing and quantifying the interaction strength between nodes that varies with geographic location and time; Calculate the dynamic relationship weights between nodes based on the spatial position embedding (V_s) and temporal position embedding (V_t) of the nodes; W_ij(t) = softmax(exp(-norm(V_si - V_tj)^2 / sigma^2)) Where: Wij(t): the relationship weight from node i to node j at time point t; Vsi: the spatial position embedding of node i; Vtj: the temporal position embedding of node j at time point t; norm(...): calculates the Euclidean distance between two node position embedding vectors; sigma: the standard deviation parameter for adjusting distance sensitivity; softmax: ensures that the sum of all weights to node j is 1, providing normalized weights for subsequent graph convolution operations; After obtaining the dynamic relationship weights between nodes, the weights are used to update and aggregate the features of the nodes. The spatiotemporal graph convolutional network iteratively updates the states of the nodes through multi-layer graph convolution operations. The features of each node are updated by weighted accumulation of the features of adjacent nodes (multiplied by the corresponding edge weights and layer parameters). H^(l+1) = sigma(sum(theta_l × H^(l)(k) × W_ik for k in N(i))) Where: H^(l+1): feature vector of node i in the l+1th layer; H^(l)(k): feature vector of neighboring node k in the lth layer; theta_l: learning parameter of the lth layer; W_ik: weight between node i and its neighbor k; sigma: nonlinear activation function; N(i): set of neighboring nodes of node i.

[0012] Optionally, the deep reinforcement learning and signal optimization include: reinforcement learning training, using deep reinforcement learning to find the optimal traffic signal control strategy, learning from the environment through interactive learning and optimizing the decision-making process of signal control; constructing a reward function to enhance the performance of the traffic signal control system, R = -(alpha × delay + beta × queue_length), where R: reward value, designed to be a negative value; delay: total delay time caused by the traffic signal; queue_length: length of the vehicle queue formed at the traffic light; alpha and beta: coefficients for adjusting delay and queue length in the reward function; after obtaining the optimized reward function, applying the learned strategy to actually control the traffic signal, and using the learned Q function-action value function to select the optimal action to execute at each decision point, S_new = argmax(Q(S, A, R, S')), where: S_new: new signal state, Q(S, A, R, S'): Q function, representing the expected utility when taking action A in state S, obtaining reward R and transferring to the new state S', argmax: selecting action A that maximizes the Q function.

[0013] Optionally, the adaptive feedback and system iterative update include: adaptive performance feedback collection, the system monitors and evaluates its performance in real time, the system collects traffic data in real time and analyzes the effect of the current signal control strategy through an evaluation function; The specific method of performance feedback is: Feedback = evaluate(H, T_realtime) Among them: Feedback: performance feedback, quantitative expression of evaluation results, evaluate: evaluation function, used to analyze the effect of signal control strategy, H: historical control strategy and performance data, T_realtime: real-time traffic data; the evaluation function combines historical data and real-time data to quantify the effectiveness of the current signal control strategy. By comparing historical and real-time data, the system can identify which strategies are effective and which need to be improved; After collecting adaptive performance feedback, adjust the model parameters according to the adaptive performance feedback, and use the gradient descent method to adjust the model parameters, Theta_new = Theta_old + gamma × gradient(Feedback) Among them: Theta_new: updated model parameters; Theta_old: current model parameters; gamma: learning rate, controlling the step size of parameter update; gradient (Feedback): the gradient calculated based on feedback, guiding the direction and amplitude of parameter update.

[0014] Optionally, the system performance monitoring and evaluation includes: Evaluate the effectiveness of signal control strategies by continuously monitoring traffic flow speed, delay time, queue length, and accident rate indicators; The specific method of performance evaluation is: Performance = f(H_eval, Metrics) in: Performance: Performance evaluation results, a quantitative score or a collection of multiple scores, expressing the current operating efficiency and effectiveness of the system; f: evaluation function, used to extract and calculate performance scores from historical performance data and real-time measured indicators; H_eval: historical evaluation data, including previous performance evaluation records and system operation data; Metrics: Current key performance indicators, including traffic flow speed, delay time, etc.

[0015] According to another embodiment of the present application, there is also provided a traffic signal adaptive control device, comprising: Data fusion and processing module: used to perform multi-source data fusion and spatiotemporal location-aware graph convolution initialization on real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Dynamic spatiotemporal relationship modeling and prediction: It uses spatiotemporal location-aware graph convolutional networks to convert geographic and time information into mathematical vectors to build a dynamic traffic flow prediction model. It uses graph convolutional networks to analyze the spatiotemporal correlation between various traffic nodes and predict future traffic flow status. Learning and optimization module: used for deep reinforcement learning and signal optimization; Feedback and update module: used for adaptive feedback and system iterative update; Monitoring and evaluation module: used for system performance monitoring and evaluation.

[0016] According to another aspect of the embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium includes a stored program, wherein the program executes the above-mentioned traffic signal adaptive control method when running.

[0017] According to another aspect of an embodiment of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the traffic signal adaptive control method through the computer program.

[0018] In the embodiments of the present application, the traffic signal adaptive control method of the present application has significant advantages over the prior art, especially in terms of data processing capability, prediction accuracy and flexibility of signal optimization strategy. The following are the main advantages of the present invention over the background technology: more advanced data fusion capability, compared with the limitation of the prior art that usually only relies on simple sensor data or a single data source, the present invention provides more comprehensive data support by comprehensively utilizing multi-source data such as real-time traffic monitoring, historical traffic data, meteorological information and urban activities. This advanced data fusion capability enables the present invention to more comprehensively understand and respond to the complex dynamics of the traffic system. The method of the present application can predict and respond to possible changes in traffic flow in advance, thereby optimizing traffic signal scheduling and reducing traffic congestion caused by failure to respond in time; dynamic traffic prediction based on deep learning, the present application uses a spatiotemporal graph convolutional network (STPGCN) to model and predict traffic flow. Compared with the linear model or simple prediction algorithm commonly used in traditional methods, STPGCN can more accurately capture and analyze the complex spatiotemporal relationship between traffic nodes, and provide more accurate traffic flow prediction. This prediction is not only based on historical data, but also reflects the current traffic conditions in real time, so that traffic signal control is more refined and forward-looking, and effectively prevents and alleviates traffic pressure during peak hours and emergencies; adaptive signal control strategy optimization, the present application automatically learns and optimizes traffic signal control strategies by applying deep reinforcement learning technology, which provides higher adaptability and intelligence level than the existing technology. In traditional systems, signal control often relies on preset programs or limited response rules, and lacks sufficient flexibility to cope with rapidly changing traffic conditions; and the method of the present invention can continuously learn the most effective control strategy, adjust the signal phase and duration in real time, so as to more effectively cope with complex and changeable traffic conditions and achieve optimization of signal control. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 is a flow chart of a traffic signal adaptive control method according to an embodiment of the present application; Figure 2 It is a structural block diagram of adaptive control of a traffic signal according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] In this embodiment, a method is provided. Figure 1 is a flow chart of a traffic signal adaptive control method according to an embodiment of the present application. Figure 1 As shown, the process includes the following steps: S101 performs multi-source data fusion and spatiotemporal location-aware graph convolution initialization on real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; S102 uses a spatiotemporal location-aware graph convolutional network to convert geographic and temporal information into mathematical vectors to build a dynamic traffic flow prediction model. It uses the graph convolutional network to analyze the spatiotemporal correlation between various traffic nodes and predict future traffic flow status. S103 Deep reinforcement learning and signal optimization; S104 adaptive feedback and system iterative update; S105 system performance monitoring and evaluation.

[0025] Optionally, the multi-source data fusion of real-time traffic monitoring data, historical traffic data, meteorological data and urban activity information data includes: Integrate real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Real-time traffic monitoring (T_realtime), historical traffic data (T_history), meteorological data (M), and urban activity information (C). Each data source provides information from different perspectives and dimensions of the operation of the traffic system: Real-time traffic monitoring (T_realtime): including traffic flow, speed, accident reports, etc., which directly affects signal control decisions.

[0026] Historical traffic data (T_history): Provides past traffic patterns and trends, which helps optimize the prediction model.

[0027] Meteorological data (M): Weather conditions have a significant impact on traffic flow, for example, rain or snow may cause traffic to slow down.

[0028] City event information (C): Special events such as sporting events or concerts can cause traffic rush.

[0029] The above data are cleaned, standardized, and synchronized; data cleaning removes erroneous or incomplete data records; data standardization ensures the consistency of data formats from different sources; data synchronization ensures that all data timestamps are aligned to provide a unified time frame for subsequent analysis.

[0030] D = synchronize(normalize(clean(T_realtime, T_history, M, C))) in: D: The final integrated and synchronized dataset, used as input into the subsequent graph convolutional network; clean: data cleaning function to remove invalid or missing data; normalize: data standardization function, unifying the format and range of different data; synchronize: data synchronization function, ensuring that all data is aligned on the timeline.

[0031] Optionally, the spatiotemporal position-aware graph convolution initialization includes: Construct a space-time graph to represent the transportation network; Through location embedding technology, the geographical and temporal attributes of each node (i.e., intersection, road segment, etc.) are converted into a mathematically tractable vector format; Spatial position embedding (V_s): Spatial position embedding reflects the location of each node in the geographic space and captures the spatial relationship between nodes through position embedding; V_s = W_s × E_s(node_positions): The spatial encoding function E_s is used to convert the geographic location information of each node into an embedding vector, which is then adjusted through the weight matrix W_s to adapt to the specific network architecture; Temporal position embedding (V_t): Temporal position embedding captures the periodicity and sequentiality of time; V_t = W_t × E_t(time_steps): Similarly, the time encoding function E_t converts time information (such as hours or weeks) into an embedding vector, and the weight matrix W_t is used to adjust the embedding vector to match the network structure.

[0032] Through this multi-source data fusion and spatiotemporal graph construction method, traffic conditions can be monitored and predicted in real time, and traffic signals can be adjusted dynamically. For example, during the morning rush hour on weekdays, the system can predict the congestion of specific sections of the road based on real-time data and historical trends, and adjust the green signal duration of the traffic lights in advance to alleviate congestion. Similarly, when encountering emergencies (such as large-scale sporting events), the system can use urban activity information and meteorological data to adjust traffic flows and optimize traffic conditions throughout the city. This approach not only improves the efficiency and flexibility of the transportation system, but also improves the level of intelligent urban management.

[0033] Multi-source data fusion and spatiotemporal location-aware graph convolution initialization are the foundation of the entire system, responsible for collecting and integrating data from different sources and converting them into a format that can be used for further analysis. This step ensures that the system has sufficient and accurate data to support subsequent modeling and prediction work. In addition, by converting geographic and temporal information into mathematically processable vectors, this step provides basic data input for modeling dynamic spatiotemporal relationships.

[0034] Optionally, a spatiotemporal location-aware graph convolutional network is used to convert geographic and temporal information into mathematical vectors to construct a dynamic traffic flow prediction model. The spatiotemporal correlation between various traffic nodes is analyzed through the graph convolutional network, and the future traffic flow state is predicted, including: Spatiotemporal location-aware relational reasoning is used to adaptively infer the relationship weights between nodes in different time and space, capturing and quantifying the interaction strength between nodes that varies with geographic location and time. The degree of mutual influence between nodes is evaluated by calculating the distance between spatial and temporal embedding vectors. Nodes that are closer have a greater influence on each other, which is reflected in higher weight values. By adjusting the sigma parameter, we can control the sensitivity of the system to distance changes, thereby adapting to different traffic scenarios and needs.

[0035] Calculate the dynamic relationship weights between nodes based on the spatial position embedding (V_s) and temporal position embedding (V_t) of the nodes; W_ij(t) = softmax(exp(-norm(V_si - V_tj)^2 / sigma^2)) Where: Wij(t): the relationship weight from node i to node j at time point t; Vsi: the spatial position embedding of node i; Vtj: the temporal position embedding of node j at time point t; norm(...): calculates the Euclidean distance between two node position embedding vectors; sigma: the standard deviation parameter for adjusting distance sensitivity; softmax: ensures that the sum of all weights to node j is 1, providing normalized weights for subsequent graph convolution operations; After obtaining the dynamic relationship weights between nodes, the weights are used to update and aggregate the features of the nodes. The spatiotemporal graph convolutional network iteratively updates the states of the nodes through multi-layer graph convolution operations. The features of each node are updated by weighted accumulation of the features of adjacent nodes (multiplied by the corresponding edge weights and layer parameters). H^(l+1) = sigma(sum(theta_l × H^(l)(k) × W_ik for k in N(i))) Where: H^(l+1): feature vector of node i in the l+1th layer; H^(l)(k): feature vector of neighboring node k in the lth layer; theta_l: learning parameter of the lth layer; W_ik: weight between node i and its neighbor k; sigma: nonlinear activation function; N(i): set of neighboring nodes of node i.

[0036] The spatiotemporal graph convolutional network uses geographic location information and time labels to construct a spatiotemporal graph. By learning the complex dependencies between traffic nodes, it can accurately predict the changing trends of traffic flow. The spatiotemporal graph convolutional network is trained using historical traffic data, while verification and testing are carried out in combination with real-time traffic data.

[0037] The spatiotemporal graph convolutional network iteratively updates the state of nodes through multi-layer graph convolution operations, so that the features of each node not only contain its own information, but also integrate the information of neighboring nodes. The features of each node are updated by weighted accumulation of the features of adjacent nodes (multiplied by the corresponding edge weights and layer parameters). This makes the new features of each node not only contain its own historical information, but also integrate the information of nodes related to it in space and time. This integration of information enables the network to capture complex spatiotemporal dependencies and improve the accuracy of traffic flow prediction. For example, in the city's traffic management center, by deploying this dynamic spatiotemporal relationship modeling and prediction system, managers can monitor and predict the traffic conditions at major intersections in the city in real time. For example, before the morning rush hour begins, the system can predict the congestion that may occur at a specific intersection, and adjust the timing of the traffic lights in advance, or push recommended routes to drivers to effectively avoid or alleviate traffic congestion. In addition, the system can also quickly adjust signal control strategies during emergencies such as accidents or large-scale events to ensure smooth traffic flow.

[0038] Optionally, the deep reinforcement learning and signal optimization include: reinforcement learning training, using deep reinforcement learning to find the optimal traffic signal control strategy, learning from the environment through interactive learning and optimizing the decision-making process of signal control; constructing a reward function to enhance the performance of the traffic signal control system, R = -(alpha × delay + beta × queue_length), where R: reward value, designed to be a negative value; delay: total delay time caused by the traffic signal; queue_length: length of the vehicle queue formed at the traffic light; alpha and beta: coefficients for adjusting delay and queue length in the reward function; after obtaining the optimized reward function, applying the learned strategy to actually control the traffic signal, and using the learned Q function-action value function to select the optimal action to execute at each decision point, S_new = argmax(Q(S, A, R, S')), where: S_new: new signal state, Q(S, A, R, S'): Q function, representing the expected utility when taking action A in state S, obtaining reward R and transferring to the new state S', argmax: selecting action A that maximizes the Q function.

[0039] In the traffic control problem, the goal is to reduce delay and queue length, which directly affect the smoothness of traffic flow and the waiting time of vehicles. By adjusting the values ​​of alpha and beta, the impact of delay time or queue length on overall traffic efficiency can be emphasized. For example, if an intersection is particularly congested, a higher weight may need to be given to queue length to relieve traffic pressure. This reward function encourages the learning algorithm to find signal control strategies that reduce these two parameters, thereby optimizing traffic flow. The deep reinforcement learning model continuously collects and analyzes data from traffic sensors. For example, during the morning rush hour, the system can automatically adjust the green letter duration of the signal light to adapt to the increase in traffic flow, thereby avoiding serious traffic congestion. In addition, when large-scale events or emergencies occur nearby, the system can quickly adjust the signal control strategy to optimize traffic flow and distribution to ensure safe and smooth traffic. By monitoring traffic conditions in real time and automatically adjusting signal lights, urban transportation systems can more effectively manage traffic loads and improve the overall efficiency and safety of urban transportation.

[0040] Optionally, the adaptive feedback and system iterative update include: adaptive performance feedback collection, the system monitors and evaluates its performance in real time, the system collects traffic data in real time and analyzes the effect of the current signal control strategy through an evaluation function; Feedback = evaluate(H, T_realtime) Among them: Feedback: performance feedback, quantitative expression of evaluation results, evaluate: evaluation function, used to analyze the effect of signal control strategy, H: historical control strategy and performance data, T_realtime: real-time traffic data; the evaluation function combines historical data and real-time data to quantify the effectiveness of the current signal control strategy. By comparing historical and real-time data, the system can identify which strategies are effective and which need to be improved; The evaluation function combines historical and real-time data to quantify the effectiveness of the current signal control strategy. By comparing historical and real-time data, the system can identify which strategies are effective and which need improvement. For example, if real-time data shows that delays at a specific intersection are increasing, the system may need to adjust the signal duration or phase at that intersection.

[0041] After collecting adaptive performance feedback, adjust the model parameters according to the adaptive performance feedback, and use the gradient descent method to adjust the model parameters, Theta_new = Theta_old + gamma × gradient(Feedback) where: Theta_new: updated model parameters; Theta_old: current model parameters; gamma: learning rate, controlling the step size of parameter update; gradient(Feedback): the gradient calculated based on the feedback, guiding the direction and magnitude of parameter update. By calculating the gradient of performance feedback, the system can understand how to adjust parameters to improve performance. For example, if feedback shows that reducing the green letter time at a certain intersection can reduce delays, the system will adjust the relevant parameters to achieve this goal.

[0042] Adaptive feedback and system iterative update mechanisms ensure that traffic control strategies can continue to adapt to urban development and daily changes. For example, if during holidays, traffic pressure in a certain area of ​​the city increases abnormally, the system can quickly identify problems through real-time data monitoring and feedback collection and alleviate traffic pressure through parameter adjustments, such as increasing green letter duration or optimizing phase allocation. This rapid response and optimization capability not only improves traffic efficiency, but also enhances the city's ability to respond to emergencies, thereby ensuring public safety and the smooth operation of urban traffic.

[0043] Optionally, the system performance monitoring and evaluation includes: Evaluate the effectiveness of signal control strategies by continuously monitoring traffic flow speed, delay time, queue length, and accident rate indicators; Performance = f(H_eval, Metrics) in: Performance: Performance evaluation results, a quantitative score or a collection of multiple scores, expressing the current operating efficiency and effectiveness of the system; f: evaluation function, used to extract and calculate performance scores from historical performance data and real-time measured indicators; H_eval: historical evaluation data, including previous performance evaluation records and system operation data; Metrics: Current key performance indicators, including traffic flow speed, delay time, etc.

[0044] The evaluation function f comprehensively evaluates the system performance based on historical data and real-time data. For example, the function may use statistical analysis, machine learning models, or deep learning techniques to predict and analyze the long-term and short-term effects of different traffic strategies. By comparing historical performance with real-time performance, the system can detect and diagnose existing problems, such as traffic signal duration is not enough to handle peak traffic, or signal timing is not adapted to changes in traffic patterns caused by specific events.

[0045] In the urban traffic management system, the system performance monitoring and evaluation module runs continuously, collects traffic data in real time and evaluates the effectiveness of the traffic signal control system. Assume that the main roads of the city experience a significant increase in traffic volume due to the opening of a new commercial center. The traffic management system will monitor this change and calculate the performance evaluation results through Performance = f(H_eval,Metrics). If it is found that the traffic delay and queue length exceed the preset threshold, the system will prompt the traffic engineer or automatically adjust the signal timing strategy to relieve traffic pressure. In addition, the system can also predict possible traffic pressure and adjust the signal settings in advance before special events, such as sports events or large conferences, to ensure smooth traffic. Through this continuous monitoring and evaluation, the urban traffic signal control system can not only respond to immediate changes in traffic demand in a timely manner, but also adjust strategies based on historical data and long-term trends, optimize traffic flow, improve road use efficiency, reduce environmental impact, and improve the quality of life of urban residents.

[0046] According to another embodiment of the present application, Figure 2 As shown, a traffic signal adaptive control device is also provided, comprising: Data fusion and processing module 201: used for multi-source data fusion and spatiotemporal location-aware graph convolution initialization of real-time traffic monitoring, historical traffic data, meteorological data and urban activity information data; Dynamic spatiotemporal relationship modeling and prediction 202: It is used to use spatiotemporal location-aware graph convolutional networks to convert geographic and time information into mathematical vectors to build a dynamic traffic flow prediction model. The graph convolutional network is used to analyze the spatiotemporal correlation between various traffic nodes and predict future traffic flow status. Learning and optimization module 203: used for deep reinforcement learning and signal optimization; Feedback and update module 204: used for adaptive feedback and system iterative update; Monitoring and evaluation module 205: used for system performance monitoring and evaluation.

[0047] According to another aspect of the embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium includes a stored program, wherein the program executes the above-mentioned traffic signal adaptive control method when running.

[0048] According to another aspect of an embodiment of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the traffic signal adaptive control method through the computer program.

[0049] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0050] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A traffic signal adaptive control method, characterized in that: include: Perform multi-source data fusion and spatiotemporal location-aware graph convolution initialization on real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Using a spatiotemporal location-aware graph convolutional network, we convert geographic and temporal information into mathematical vectors to build a dynamic traffic flow prediction model. We use the graph convolutional network to analyze the spatiotemporal correlation between traffic nodes and predict future traffic flow status. Deep reinforcement learning and signal optimization; Adaptive feedback and system iterative updates; System performance monitoring and evaluation.

2. The method according to claim 1, characterized in that The multi-source data fusion of real-time traffic monitoring data, historical traffic data, meteorological data and urban activity information data includes: Integrate real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Clean, standardize, and synchronize the above data; The specific data fusion method is: D = synchronize(normalize(clean(T_realtime, T_history,M, C))) in: D: The final integrated and synchronized dataset, used as input into the subsequent graph convolutional network; clean: data cleaning function to remove invalid or missing data; normalize: data standardization function, unifying the format and range of different data; synchronize: data synchronization function, ensuring that all data is aligned on the timeline.

3. The method according to claim 2, characterized in that The spatiotemporal position-aware graph convolution initialization includes: Construct a space-time graph to represent the transportation network; Through location embedding technology, the geographical and temporal attributes of each node are converted into a mathematically processable vector format; Spatial position embedding (V_s): Spatial position embedding reflects the location of each node in the geographic space and captures the spatial relationship between nodes through position embedding; V_s = W_s × E_s(node_positions): The spatial encoding function E_s is used to convert the geographic location information of each node into an embedding vector, which is then adjusted through the weight matrix W_s to adapt to the specific network architecture; Temporal position embedding (V_t): Temporal position embedding captures the periodicity and sequentiality of time; V_t = W_t × E_t(time_steps): Similarly, the time encoding function E_t converts time information (such as hours or weeks) into an embedding vector, and the weight matrix W_t is used to adjust the embedding vector to match the network structure.

4. The method according to claim 1, characterized in that: The spatiotemporal location-aware graph convolutional network is used to convert geographic and time information into mathematical vectors to build a dynamic traffic flow prediction model. The graph convolutional network is used to analyze the spatiotemporal correlation between various traffic nodes and predict future traffic flow status, including: Using spatiotemporal location-aware relational reasoning to adaptively infer the relationship weights between nodes in different time and space, capturing and quantifying the interaction strength between nodes that varies with geographic location and time; Calculate the dynamic relationship weights between nodes based on the spatial position embedding (V_s) and temporal position embedding (V_t) of the nodes; W_ij(t) = softmax(exp(-norm(V_si - V_tj)^2 / sigma^2)) Where: Wij(t): the relationship weight from node i to node j at time point t; Vsi: the spatial position embedding of node i; Vtj: the temporal position embedding of node j at time point t; norm(...): calculates the Euclidean distance between two node position embedding vectors; sigma: the standard deviation parameter for adjusting distance sensitivity; softmax: ensures that the sum of all weights to node j is 1, providing normalized weights for subsequent graph convolution operations; After obtaining the dynamic relationship weights between nodes, the weights are used to update and aggregate the features of the nodes. The spatiotemporal graph convolutional network iteratively updates the states of the nodes through multi-layer graph convolution operations. The features of each node are updated by weighted accumulation of the features of adjacent nodes (multiplied by the corresponding edge weights and layer parameters). H^(l+1) = sigma(sum(theta_l × H^(l)(k) × W_ik for k in N(i))) Where: H^(l+1): feature vector of node i in the l+1th layer; H^(l)(k): feature vector of neighboring node k in the lth layer; theta_l: learning parameter of the lth layer; W_ik: weight between node i and its neighbor k; sigma: nonlinear activation function; N(i): set of neighboring nodes of node i.

5. The method according to claim 1, characterized in that The deep reinforcement learning and signal optimization include: reinforcement learning training, using deep reinforcement learning to find the optimal traffic signal control strategy, learning from the environment through interactive learning and optimizing the decision-making process of signal control; constructing a reward function to strengthen the performance of the traffic signal control system, and the reward value calculation method is: R = -(alpha × delay + beta × queue_length), where R: reward value, designed to be a negative value; delay: total delay time caused by the traffic signal; queue_length: length of the vehicle queue formed at the traffic light; alpha and beta: coefficients for adjusting delay and queue length in the reward function; after obtaining the optimized reward function, applying the learned strategy to actually control the traffic signal, and using the learned Q function-action value function to select the optimal action to execute at each decision point, S_new = argmax(Q(S, A, R, S')), where: S_new: new signal state, Q(S, A, R, S'): Q function, representing the expected utility when taking action A in state S, obtaining reward R and transferring to a new state S', argmax: selecting action A that maximizes the Q function.

6. The method according to claim 1, characterized in that The adaptive feedback and system iterative update include: Adaptive performance feedback collection, the system monitors and evaluates its performance in real time, the system collects traffic data in real time and analyzes the effect of the current signal control strategy through evaluation functions; The specific calculation method for performance feedback is: Feedback = evaluate(H, T_realtime) Among them: Feedback: performance feedback, quantitative expression of evaluation results, evaluate: evaluation function, used to analyze the effect of signal control strategy, H: historical control strategy and performance data, T_realtime: real-time traffic data; the evaluation function combines historical data and real-time data to quantify the effectiveness of the current signal control strategy. By comparing historical and real-time data, the system can identify which strategies are effective and which need to be improved; After collecting the adaptive performance feedback, the model parameters are adjusted according to the adaptive performance feedback, and the gradient descent method is used to adjust the model parameters. The specific adjustment method of the model parameters is: Theta_new = Theta_old + gamma × gradient(Feedback) Among them: Theta_new: updated model parameters; Theta_old: current model parameters; gamma: learning rate, controlling the step size of parameter update; gradient (Feedback): the gradient calculated based on feedback, guiding the direction and amplitude of parameter update.

7. The method according to claim 1, characterized in that The system performance monitoring and evaluation includes: Evaluate the effectiveness of signal control strategies by continuously monitoring traffic flow speed, delay time, queue length, and accident rate indicators; The specific method of performance evaluation is: Performance = f(H_eval, Metrics) in: Performance: Performance evaluation results, a quantitative score or a collection of multiple scores, expressing the current operating efficiency and effectiveness of the system; f: evaluation function, used to extract and calculate performance scores from historical performance data and real-time measured indicators; H_eval: historical evaluation data, including previous performance evaluation records and system operation data; Metrics: Current key performance indicators, including traffic flow speed, delay time, etc.

8. A traffic signal adaptive control device, characterized in that: include: Data fusion and processing module: used to perform multi-source data fusion and spatiotemporal location-aware graph convolution initialization on real-time traffic monitoring, historical traffic data, meteorological data, and urban activity information data; Dynamic spatiotemporal relationship modeling and prediction: It uses spatiotemporal location-aware graph convolutional networks to convert geographic and time information into mathematical vectors to build a dynamic traffic flow prediction model. It uses graph convolutional networks to analyze the spatiotemporal correlation between various traffic nodes and predict future traffic flow status. Learning and optimization module: used for deep reinforcement learning and signal optimization; Feedback and update module: used for adaptive feedback and system iterative update; Monitoring and evaluation module: used for system performance monitoring and evaluation.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images