A traffic risk dynamic assessment model construction system

Abstract

The present invention relates to the technical field of water traffic risk assessment, and in particular provides a system for constructing a dynamic assessment model for traffic risk, which solves the problem that the degree of intelligence and dynamic performance of traffic risk assessment needs to be further improved. The system includes: a first modeling component constructs a risk propagation model using meta-learning; a second modeling component constructs a multi-agent collaborative risk assessment model; an active intervention component; and generates an active intervention strategy. The present invention improves traffic safety through real-time risk perception, accurate assessment, and active intervention. The system can effectively reduce the incidence of traffic accidents; optimizes traffic management dynamic risk maps and intervention strategies to provide scientific decision-making basis for traffic management departments and improve management efficiency; enhances user experience through multi-channel early warning and path induction, which can provide drivers with a safer and more convenient travel experience; and promotes the development of smart transportation and provides technical support for the realization of smart transportation systems.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic risk assessment, and in particular to a traffic risk dynamic assessment model construction system. Background Art

[0002] The background technology of dynamic assessment of traffic risks mainly involves cross-applications in multiple fields, aiming to improve traffic safety and efficiency through real-time data analysis and intelligent algorithms; provide scientific decision-making support for traffic management departments, thereby effectively reducing the incidence of traffic accidents and improving road traffic efficiency. However, the intelligence and dynamics of existing technologies are relatively poor, which restricts the further improvement of the accuracy of dynamic assessment results of traffic risks.

[0003] Prior art 1, Chinese patent, application number: 202110643166.5 discloses a method for constructing a dynamic assessment model for water traffic risk. By collecting water traffic elements and analyzing the internal structural characteristics of the water traffic system, the internal information set of the water traffic system is established; based on the internal information flow of the water traffic system, the dynamic characteristics of water traffic risk are analyzed, and the water traffic risk operation indicators are extracted; the risk operation indicators are used as Bayesian network nodes to determine the Bayesian network structure; then the mutual information method is used to sort the correlation of the risk operation indicators and screen the Bayesian network nodes; finally, the Bayesian network node parameters are input to obtain the dynamic assessment model for water traffic risk. Although providing a method for establishing a real-time dynamic assessment model can provide theoretical and technical support for water traffic safety warning and pre-control, and effectively improve water traffic safety and water traffic management efficiency; however, it is limited to water traffic, not only the scope of application is limited, but also the interpretability of the Bayesian network model is poor, which may affect the decision support effect in practical applications.

[0004] Prior art 2, Chinese patent, application number 202410644262.5 discloses a method for extracting traffic risk elements based on emotional arousal of drivers and passengers, the method is: integrated data collection of people, vehicles and environment; emotional arousal status assessment of drivers and passengers; construction of a model associating emotional arousal with traffic environment; step 4, extraction of traffic risk elements. Although the evaluation of the emotional state of drivers and passengers is realized, it provides a basis for the extraction of traffic risk elements, and establishes a mapping from traffic environment information to the emotional state of drivers and passengers, and then extracts traffic risk elements that induce emotional arousal of drivers and passengers through the attention weights in the model, and from a people-oriented perspective, points out traffic risk elements that affect the sense of security of drivers and passengers in a data-driven way; however, its functions are relatively complex and overly dependent on the accuracy of the model.

[0005] Prior art three, Chinese patent, application number 202411527257.2 discloses a method and device for identifying highway traffic risks based on disaster impacts, the method includes the following steps: constructing top events and intermediate events of highway traffic operation risks under the impact of disasters; identifying sub-events under each intermediate event according to the intermediate events, and identifying the basic events in sequence from the sub-event characteristics downward, and then obtaining the direct risk factors or indirect risk factors of highway traffic risks under the impact of disasters; preliminarily calibrating the evaluation indicators of basic events; screening the evaluation indicators of direct risk factors or indirect risk factors, and obtaining the screened risk factors, which are used for highway traffic risk identification. Although it can provide a basis for highway risk management under disasters, managers can control the development of risk factors on this basis to avoid or reduce economic and property losses caused by high-risk situations; however, there are many steps to determine the evaluation indicators, and it is not specific to evaluate traffic risks, and its implementation performance is poor.

[0006] At present, the existing technologies 1, 2 and 3 have the problem that the intelligent degree and dynamic performance of traffic risk assessment need to be further improved. Therefore, the present invention provides a traffic risk dynamic assessment model construction system. Summary of the invention

[0007] In order to achieve the above object, the present invention adopts the following technical scheme:

[0008] One aspect of the present invention provides a traffic risk dynamic assessment model construction system, comprising:

[0009] The first modeling component is configured to use spatiotemporal graph neural networks and spatiotemporal attention mechanisms to capture the coupling relationship between traffic risks in time and space, and perceive the real-time risks of complex transportation networks; build a risk propagation model using meta-learning to simulate the propagation and diffusion process of risks in transportation networks; generate dynamic risk maps to display the risk level and change trend of each area in the transportation network in real time;

[0010] The second modeling component is configured to collect driver behavior data, vehicle status data, road condition data and environmental data; build a multi-agent collaborative risk assessment model to analyze the interactive relationship between people, vehicles, roads and the environment; use deep reinforcement learning to simulate the decision-making behavior of drivers in different environments and predict potential risks; and analyze the transmission path and key nodes of risks in the traffic system;

[0011] The active intervention component is configured to establish a multi-level early warning mechanism based on risk levels; automatically generate early warning information using natural language and push it in real time through multiple channels; and generate active intervention strategies.

[0012] The present invention improves traffic safety through real-time risk perception, accurate assessment and active intervention, and the system can effectively reduce the incidence of traffic accidents; optimizes traffic management dynamic risk maps and intervention strategies to provide scientific decision-making basis for traffic management departments and improve management efficiency; enhances user experience through multi-channel early warning and path guidance, and the system can provide drivers with a safer and more convenient travel experience; promotes the development of smart transportation, provides technical support for the realization of smart transportation systems, and promotes traffic management to develop in an intelligent and refined direction. Through the above analysis, it can be seen that the various components of the traffic risk dynamic assessment model construction system cooperate with each other technically, and jointly realize the comprehensive perception, accurate assessment and active intervention of traffic risks, providing strong support for improving traffic safety and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0014] Figure 1 A system block diagram for constructing a dynamic assessment model for traffic risks provided in Example 1 of the present invention;

[0015] Figure 2 This is a block diagram of the first modeling component provided in Embodiment 2 of the present invention;

[0016] Figure 3 This is a block diagram of a real-time display module provided in Embodiment 3 of the present invention;

[0017] Figure 4 This is a block diagram of the second modeling component provided in Embodiment 4 of the present invention;

[0018] Figure 5 This is a block diagram of a decision-making hierarchical module provided in Embodiment 5 of the present invention;

[0019] Figure 6 This is a block diagram of the reasoning and hierarchical modeling submodule provided in Example 6 of the present invention;

[0020] Figure 7 This is a block diagram of a risk quantification module provided in Example 7 of the present invention;

[0021] Figure 8 This is a block diagram of the path and key node identification submodule provided in Example 8 of the present invention;

[0022] Fig. 9 This is a block diagram of a quantization output unit provided in Embodiment 9 of the present invention;

[0023] Fig.10 This is a block diagram of the active intervention component provided in Embodiment 10 of the present invention;

[0024] Fig.11 A block diagram of an electronic device provided by the present invention;

[0025] Fig.12 A block diagram of a computer-readable storage medium provided for the present invention. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present invention will be described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0027] In the following, the terms "first", "second", etc. are used only for convenience of description and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0028] In the present invention, unless otherwise clearly specified and limited, the term "connection" should be understood in a broad sense, for example, "connection" can be a fixed mechanical connection, or a detachable mechanical connection, or integrated; or, "connection" can be a direct connection, or an indirect connection through an intermediate medium. In addition, unless otherwise clearly specified and limited, the term "coupling" should be understood in a broad sense, for example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components, and can also be understood as electrical connection between different components in a circuit structure through physical lines such as copper foil or wires on a printed circuit board (PCB) that can transmit electrical signals to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in an air-spaced / non-contact manner, for example, two components are electrically connected by capacitive coupling to transmit electrical signals.

[0029] In the embodiments of the present invention, directional terms such as "up", "down", "left" and "right" may be defined including but not limited to the orientation relative to the schematic placement of the components in the drawings. It should be understood that these directional terms may be relative concepts, which are used for relative description and clarification, and may change accordingly according to the change of the orientation of the components in the drawings.

[0030] The embodiments of the present invention can be used for real-time traffic risk perception and prediction. Through the spatiotemporal graph neural network and spatiotemporal attention mechanism, it can capture the spatiotemporal coupling relationship of risks in the traffic network in real time, and dynamically perceive the potential risks in complex traffic scenes such as urban roads, highways, and intersections; help traffic management departments to grasp the changing trends of traffic risks in a timely manner, take preventive measures in advance, and reduce the incidence of traffic accidents. Dynamic risk evolution simulation uses meta-learning technology to simulate the propagation and diffusion process of traffic risks in the network, generate dynamic risk maps, and display the risk level and changing trends of each area in real time; provide a scientific basis for traffic planning and management, help optimize traffic resource allocation, and improve the overall safety and efficiency of the traffic network. Multi-dimensional data fusion and risk assessment collect driver behavior, vehicle status, road conditions and environmental data, build a multi-agent collaborative risk assessment model, analyze the interactive relationship between people-vehicles-roads-environment; comprehensively evaluate the sources of risks in the traffic system, identify potential high-risk scenarios, and provide data support for precise intervention. Driver behavior simulation and risk prediction use deep reinforcement learning technology to simulate the decision-making behavior of drivers in different environments, predict potential risks, and analyze the transmission path and key nodes of risks in the traffic system; help traffic management departments predict high-risk behaviors (such as fatigue driving, speeding) in advance, and formulate targeted intervention measures. Multi-level warning and active intervention, establish a multi-level warning mechanism based on risk level, automatically generate warning information through natural language generation technology, and push it in real time through multiple channels such as vehicle terminals, mobile APPs, and traffic broadcasts; improve the risk awareness of drivers and traffic managers, and reduce traffic accidents caused by information lags. Improve the intelligence level of the traffic system. Through the in-depth application of AI technology, the system can realize intelligent assessment, warning and intervention of traffic risks, promote the transformation of traffic management from passive response to active prevention; provide important support for the construction of smart cities, and improve the safety, efficiency and sustainability of urban transportation. Application scenarios: Urban traffic management monitors traffic risks on urban roads in real time, optimizes signal light control and traffic flow distribution; highway management predicts potential risks on highways (such as congestion, accidents) and issues early warning information in a timely manner; intersection safety optimization analyzes the risk transmission path of intersections, optimizes traffic signals and vehicle guidance strategies; driver assistance systems provide drivers with personalized risk warnings and driving suggestions to improve driving safety.

[0031] Embodiment 1:

[0032] like Figure 1 As shown, an embodiment of the present invention provides a traffic risk dynamic assessment model construction system, comprising:

[0033] The first modeling component is configured to use spatiotemporal graph neural networks and spatiotemporal attention mechanisms to capture the coupling relationship between traffic risks in time and space, and to perceive the real-time risks of complex transportation networks (such as urban roads, highways, and intersections); to build a risk propagation model using meta-learning to simulate the propagation and diffusion process of risks in the transportation network; and to generate a dynamic risk map to display the risk level and change trend of each area in the transportation network in real time;

[0034] The second modeling component is configured to collect driver behavior data (such as fatigue status, driving habits), vehicle status data (such as vehicle speed, acceleration), road condition data (such as road slipperiness, visibility) and environmental data (such as weather, lighting); build a multi-agent collaborative risk assessment model to analyze the interactive relationship between people, vehicles, roads and the environment; use deep reinforcement learning to simulate the decision-making behavior of drivers in different environments and predict potential risks; analyze the transmission path and key nodes of risks in the traffic system;

[0035] The active intervention component is configured to establish a multi-level warning mechanism based on risk levels (such as low-risk prompts, medium-risk warnings, and high-risk alerts); automatically generate warning information using natural language and push it in real time through multiple channels (such as vehicle-mounted terminals, mobile phone apps, and traffic broadcasts); and generate active intervention strategies (such as dynamic speed limits, traffic light optimization, and path guidance).

[0036] In the above embodiment, the first modeling component can perceive the risk distribution in a complex traffic network in real time by capturing the coupling relationship between traffic risks in time and space; combined with the graph neural network and the attention mechanism, the model can dynamically learn the correlation between different nodes (such as road sections and intersections) in the traffic network, and predict the risk changes in combination with time series data; the meta-learning risk propagation model simulates the process of risk propagation and diffusion in the traffic network, and can quickly adapt to the changes in different traffic scenarios and risk patterns; through meta-learning, the model can quickly adjust parameters with a small amount of data to improve the prediction accuracy; the dynamic risk map displays the risk level and change trend of each area in the traffic network in real time, providing intuitive risk information for traffic management departments and drivers. Significance: Improve risk perception capabilities, through spatiotemporal modeling and dynamic evolution models, it can more accurately identify and predict traffic risks, and provide a scientific basis for traffic management; Enhance the introduction of adaptive meta-learning technology so that the system can quickly adapt to new traffic scenarios and risk patterns, and improve the generalization ability of the model; Visual support dynamic risk maps provide decision makers with intuitive risk distribution information, which is convenient for formulating targeted management measures. The second modeling component collects driver behavior data, vehicle status data, road condition data and environmental data to build a comprehensive risk assessment foundation; the multi-agent collaborative risk assessment model analyzes the interactive relationship between people, vehicles, roads and the environment, and identifies the main sources and transmission paths of risks; simulates the decision-making behavior of drivers in different environments through deep reinforcement learning to predict potential risks; risk transmission path analysis identifies key nodes and high-risk areas in the traffic system to provide a basis for risk intervention. Significance: Comprehensive risk assessment can more comprehensively assess traffic risks through multi-source data fusion and avoid the limitations of a single data source; accurate prediction deep reinforcement learning technology can simulate the decision-making behavior of drivers, predict potential risks, and improve the accuracy of warnings; key node identification can identify weak links in the traffic system by analyzing risk transmission paths, providing support for optimizing traffic management. The active intervention component provides differentiated warning information according to the risk level (low risk prompt, medium risk warning, high risk alarm); automatically generates warning information and pushes it in real time through multiple channels such as vehicle terminals, mobile phone APPs, and traffic broadcasts to ensure timely communication of information; generates intervention strategies such as dynamic speed limits, signal light optimization, and path induction to reduce the possibility of risk. Significance: The multi-level warning mechanism can provide differentiated warning information according to the risk level to avoid excessive or insufficient warning; efficient information communication pushes warning information through multiple channels to ensure that drivers and traffic management departments can obtain risk information in a timely manner; active intervention can proactively reduce traffic risks and improve overall traffic safety through strategies such as dynamic speed limit and traffic light optimization.

[0037] In summary, this embodiment improves traffic safety through real-time risk perception, accurate assessment and active intervention, and the system can effectively reduce the incidence of traffic accidents; optimizes traffic management dynamic risk maps and intervention strategies to provide scientific decision-making basis for traffic management departments and improve management efficiency; enhances user experience through multi-channel early warning and path guidance, and the system can provide drivers with a safer and more convenient travel experience; promotes the development of smart transportation and provides technical support for the realization of smart transportation systems, and promotes traffic management towards intelligent and refined directions. Through the above analysis, it can be seen that the various components of the traffic risk dynamic assessment model construction system cooperate with each other technically, and jointly realize the comprehensive perception, accurate assessment and active intervention of traffic risks, providing strong support for improving traffic safety and efficiency.

[0038] Embodiment 2:

[0039] like Figure 2 As shown, based on Example 1, the first modeling component provided by the embodiment of the present invention includes:

[0040] The relationship capture module is configured to collect real-time data from multiple data sources (such as traffic monitoring, sensors, vehicle-mounted equipment, etc.), including traffic flow, vehicle speed, accident records, and weather conditions; clean the collected real-time data, remove noise and outliers, and perform standardization processing; define each intersection or section in the traffic network as a node in the graph, and define the connection relationship between nodes as edges; use the spatiotemporal graph neural network to extract the spatiotemporal characteristics of the traffic network and capture the coupling relationship of traffic risks in time and space;

[0041] The model training module is configured to define the traffic risk evolution problem as a series of related tasks, each task corresponding to a specific time window or spatial region; train the model through a meta-learner to capture the dynamic characteristics of risk evolution;

[0042] The real-time display module is configured to build a risk propagation model based on the meta-learner to simulate the risk propagation and diffusion process in the transportation network; based on the output of the risk propagation model, a dynamic risk map is generated to display the risk level and change trend of each area in the transportation network in real time.

[0043] In the above embodiments, the relationship capture module collects data from multiple data sources (such as traffic monitoring, sensors, vehicle-mounted equipment, etc.) in real time to ensure the comprehensiveness and real-time nature of the data; through cleaning and standardization processing, noise and outliers are removed to improve the quality and reliability of the data; graph structure modeling defines intersections or sections in the traffic network as nodes, and defines connection relationships as edges to construct the graph structure of the traffic network. The modeling method can intuitively reflect the topological relationship of the traffic network; spatiotemporal feature extraction uses spatiotemporal graph neural network (STGNN) to extract spatiotemporal features in the traffic network and capture the coupling relationship of traffic risks in time and space. The feature extraction method can more accurately reflect the dynamic changes of traffic risks. Significance: Data-driven decision-making provides a high-quality data foundation for model training through the integration and cleaning of multi-source data; networked modeling can more realistically reflect the complexity of the traffic network and provide a basis for risk propagation analysis; the spatiotemporal graph neural network captured by spatiotemporal features can capture the spatiotemporal evolution of traffic risks and provide a scientific basis for risk prediction. The model training module decomposes the traffic risk evolution problem into a series of related tasks, each task corresponds to a specific time window or spatial area. The decomposition method can analyze the local characteristics of risk evolution in more detail; the model is trained by the meta-learner to capture the dynamic characteristics of risk evolution. The meta-learner can quickly adapt to different tasks and improve the generalization ability of the model. Significance: Through task decomposition, the evolution law of traffic risk in different time and space can be analyzed more accurately; the introduction of the meta-learner enables the model to quickly adapt to changes in traffic risks, improving the robustness and prediction accuracy of the model. The real-time display module builds a risk propagation model based on the meta-learner, simulates the propagation and diffusion process of risks in the traffic network, and can dynamically reflect the propagation path and impact range of risks; according to the output of the risk propagation model, a dynamic risk map is generated to display the risk level and change trend of each area in the traffic network in real time. The visualization method can intuitively reflect the distribution and evolution of traffic risks. Significance: Through dynamic risk maps, traffic management departments can grasp the distribution and changing trends of traffic risks in real time and take timely response measures; risk propagation models and dynamic risk maps provide a scientific basis for traffic management, which helps to optimize traffic resource allocation and emergency response strategies.

[0044] In summary, this embodiment can achieve comprehensive perception, accurate prediction and real-time display of traffic risks through the collaborative work of multiple modules. Its significance lies not only in improving the intelligent level of traffic management, but also in providing strong technical support for the safety and efficiency of urban transportation.

[0045] Embodiment 3:

[0046] like Figure 3 As shown, based on Example 2, the real-time display module provided by the embodiment of the present invention includes:

[0047] The meta-knowledge extraction submodule is configured to establish a multi-dimensional feature space, map the spatiotemporal features, environmental features, and event features in the traffic network into a unified high-dimensional space; dynamically identify the feature dimensions that have a significant impact on risk propagation through adaptive feature selection; construct a nonlinear association network between features to capture the potential coupling relationship between features;

[0048] The dynamic evolutionary learning submodule is configured to design a spatiotemporal attention mechanism to automatically capture the key spatiotemporal nodes of risk propagation; establish a differential equation model of risk propagation to simulate the diffusion process of risk in the transportation network; introduce memory enhancement to store and update the historical pattern of risk propagation;

[0049] The meta-learning optimization submodule is configured to build a multi-task learning framework, taking the risk evolution in different time windows and spatial regions as independent tasks; designing a knowledge transfer mechanism between tasks to achieve the cross-temporal and cross-spatial migration of risk propagation patterns; introducing an adaptive weight allocation strategy to dynamically adjust the contribution of different tasks to model training;

[0050] The risk propagation simulation submodule is configured to establish a risk propagation network based on a probability graph to simulate the probability of risk transfer between nodes; design an attenuation function for risk diffusion to simulate the attenuation characteristics of risk over time and space; construct a risk superposition model to simulate the superposition effect of multi-source risks in the transportation network; establish a real-time prediction mechanism for risk propagation, dynamically update the risk propagation path, and continuously optimize the risk propagation model based on real-time data according to feedback correction.

[0051] Among them, the multi-dimensional feature space mapping formula of the meta-knowledge extraction submodule is:

[0052] In the formula, Represents a high-dimensional feature space matrix with dimensions , represents the multi-dimensional features extracted from the traffic network; Represents the time feature matrix, with dimension , represents the temporal dynamics in the transportation network; Represents the spatial feature matrix, with dimension , represents the spatial distribution characteristics in the transportation network; Represents the environmental feature matrix, with dimension , represents the environmental influencing factors in the transportation network; Represents nonlinear basis functions (such as polynomials, kernel functions, etc.) to capture complex relationships between features; represents the weight matrix, which is used to adjust the importance of different basis functions; represents the bias term, which is used to adjust the overall offset of the feature space; k represents the number of nonlinear basis functions;

[0053] Adaptive feature selection formula:

[0054] In the formula, Represents the feature weight matrix, with dimension , indicating the importance of each feature dimension; Represents the adaptive weight matrix, with dimension , used to dynamically adjust the relationship between features; Represents the bias vector, with dimension , used to adjust the initial weight of the feature; through adaptive feature selection, dynamically identify the feature dimensions that have a significant impact on risk propagation and reduce the interference of redundant features;

[0055] Nonlinear association network formula:

[0056] In the formula, Represents the feature association matrix, with dimension , indicating the nonlinear relationship between features; Represents a nonlinear activation function (such as ReLU or Sigmoid), which is used to capture the complex coupling relationship between features; represents the dynamic coupling matrix, with dimension , used to adjust the interaction strength between features; construct a nonlinear association network between features, capture the potential coupling relationship between features, and provide support for the dynamic evolution of risk propagation;

[0057] Dynamic evolution learning submodule spatiotemporal attention mechanism formula:

[0058] In the formula, Representation Node i For Node j The attention weight of the node i For Node j degree of attention; represents the query vector, represents the node i Features; represents the key vector, representing the node j Features; and Represents a value vector, representing a node i and nodes j The key spatiotemporal nodes of risk propagation are automatically captured through the spatiotemporal attention mechanism, which improves the spatiotemporal modeling ability of the model. n It represents the total number of nodes involved in the calculation;

[0059] Risk propagation differential equation model formula:

[0060]

[0061] In the formula, Indicates time t the risk communication status at that time; represents the risk transmission rate; represents the risk attenuation rate; Represents the influence coefficient of environmental factors on risk propagation; Indicates time t The environmental characteristic vector at the time of the accident; establish a differential equation model of risk propagation, simulate the diffusion process of risk in the transportation network, and consider the dynamic influence of environmental factors;

[0062] Memory Enhanced Storage Formula:

[0063]

[0064] In the formula, Indicates time t Memory status at the time Indicates the memory update rate; Represents the memory enhancement coefficient, which is used to adjust the impact of historical patterns on new memory; Indicates the change in the risk propagation state; introduces memory-enhanced storage to dynamically update the historical pattern of risk propagation and improve the model's ability to model long-term risk evolution;

[0065] The multi-task learning framework formula of the meta-learning optimization submodule is:

[0066]

[0067] In the formula, represents the total loss function; Indicates The loss function of each task; Indicates The weight of each task; Represents the regularization coefficient, which is used to prevent overfitting; Represents the L2 regularization term of the weight matrix; builds a multi-task learning framework, takes the risk evolution in different time windows and spatial regions as independent tasks, and improves the generalization ability of the model;

[0068] The formula for knowledge transfer between tasks is:

[0069]

[0070] In the formula, Indicates the task and tasks The knowledge transfer matrix between them; Indicates the task The feature matrix of Indicates the task The feature matrix of represents the historical knowledge transfer coefficient; Represent the historical knowledge transfer matrix; design the knowledge transfer mechanism between tasks to realize the cross-temporal and cross-spatial migration of risk propagation models and improve the learning efficiency of the model;

[0071] Adaptive weight allocation formula:

[0072]

[0073] In the formula, Indicates The weight of each task; Indicates Adaptive weights for each task; Represents the historical weight influence coefficient; Represents historical weights; introduces an adaptive weight allocation strategy to dynamically adjust the contribution of different tasks to model training and improve the robustness of the model;

[0074] Probabilistic graph risk propagation network formula of the risk propagation simulation submodule:

[0075]

[0076] In the formula, Representation Node To Node The probability of risk transfer; Representation Node and nodes The strength of the association between represents the historical correlation strength influence coefficient; Indicates the historical correlation strength; establishes a risk propagation network based on probability graph, simulates the probability of risk transfer between nodes, and improves the prediction accuracy of the model;

[0077] Risk diffusion attenuation function formula:

[0078]

[0079] In the formula, Indicates time t and space s The risk value at represents the initial risk value; represents the time attenuation coefficient; represents the spatial attenuation coefficient; Represent the historical risk impact coefficient; design the attenuation function of risk diffusion, simulate the attenuation characteristics of risk over time and space, and improve the dynamic modeling ability of the model;

[0080] Risk superposition model formula:

[0081]

[0082] In the formula, represents the total risk value; Indicates The risk value of each risk source; Indicates The weight of each risk source; represents the historical risk impact coefficient; Represent historical risk values; construct a risk superposition model to simulate the superposition effect of multi-source risks in the transportation network and improve the comprehensive prediction ability of the model;

[0083] Real-time prediction mechanism formula:

[0084]

[0085] In the formula, Indicates the risk propagation status at the next moment; represents the learning rate; Indicates the risk change at the current moment; represents the feedback correction coefficient; Represents real-time feedback data; establishes a real-time prediction mechanism for risk propagation, dynamically updates the risk propagation path, and continuously optimizes the risk propagation model based on real-time data according to feedback correction.

[0086] In the above embodiment, the meta-knowledge extraction submodule establishes a multi-dimensional feature space, maps the spatiotemporal features, environmental features and event features in the traffic network to a unified high-dimensional space, and can comprehensively capture various influencing factors; adaptive feature selection dynamically identifies the feature dimensions that have a significant impact on risk propagation, improving the accuracy and efficiency of the model; nonlinear association network construction captures the potential coupling relationship between features and enhances the model's ability to understand complex relationships. Significance: By extracting and integrating multi-dimensional features, the meta-knowledge extraction submodule can provide a solid foundation for subsequent risk propagation analysis and ensure that the model can fully and accurately reflect the actual situation. The dynamic evolutionary learning submodule automatically captures the key spatiotemporal nodes of risk propagation and improves the model's ability to identify key events; simulates the diffusion process of risk in the traffic network and provides a dynamic risk propagation path; updates and stores historical patterns of risk propagation to enhance the model's memory and prediction capabilities. Significance: The dynamic evolutionary learning submodule can capture and simulate the dynamic process of risk propagation in real time, providing timely and accurate support for risk management and decision-making. The meta-learning optimization submodule treats the risk evolution of different time windows and spatial regions as independent tasks to improve the generalization ability of the model; realizes the cross-temporal and spatial migration of risk propagation patterns to enhance the adaptability of the model; dynamically adjusts the contribution of different tasks to model training to optimize the overall performance of the model. Significance: Through multi-task learning and knowledge transfer, the meta-learning optimization submodule can effectively improve the adaptability and prediction accuracy of the model in different scenarios, ensuring the stable performance of the model in various complex environments. The risk propagation simulation submodule simulates the probability of risk transfer between nodes and provides a detailed risk propagation path; simulates the attenuation characteristics of risk over time and space to enhance the reality fit of the model; simulates the superposition effect of multi-source risks in the traffic network to provide a comprehensive risk assessment; dynamically updates the risk propagation path, and continuously optimizes the risk propagation model based on real-time data according to feedback correction. Significance: The risk propagation simulation submodule can simulate and predict the risk propagation process in real time, provide a scientific basis for risk management and emergency response, and ensure the timeliness and accuracy of decision-making.

[0087] In summary, this embodiment constructs a comprehensive, dynamic, and adaptive risk propagation model; it not only improves the prediction accuracy and adaptability of the model, but also provides strong support for risk management and decision-making.

[0088] Embodiment 4:

[0089] like Figure 4 As shown, based on Example 1, the second modeling component provided by the embodiment of the present invention includes:

[0090] The weight allocation module is configured to adopt a dynamic weight allocation mechanism to automatically adjust the weight of each data source according to real-time changes in data (such as sudden weather changes and increased road congestion);

[0091] The decision-making hierarchical module is configured to use a spatiotemporal causal graph network to model the causal relationship between people, vehicles, roads and the environment through causal reasoning technology; identify the causal chain that causes slippery roads due to rainy days, which in turn affects the driver's braking distance; and use hierarchical deep reinforcement learning composed of macro, meso and micro layers to decompose the driver's decision-making behavior into multiple levels;

[0092] Among them, the macro layer simulates the driver's overall judgment of the global road conditions (such as choosing a driving route); the meso layer simulates the driver's response to the local environment (such as avoiding obstacles); the micro layer simulates the driver's response to instantaneous events (such as emergency braking);

[0093] The risk quantification module is configured to simulate the risk propagation process in the transportation network, capture the dynamic characteristics of risk propagation in real time, predict the risk transmission path and key nodes; and combine Bayesian reasoning and Monte Carlo sampling to quantify the probability of occurrence and impact range of potential risks.

[0094] In the above embodiment, the weight distribution module adopts a dynamic weight distribution mechanism, which can automatically adjust the weight of each data source according to real-time data (such as sudden weather changes, increased road congestion); for example, on rainy days, the weight of the slippery road data source will be increased to ensure a more accurate reflection of the actual road conditions. Significance: The dynamic adjustment mechanism enhances flexibility and adaptability, enabling it to maintain high accuracy in a complex and changing traffic environment; by optimizing data weights in real time, it can more effectively capture key information, thereby improving the accuracy and reliability of the overall decision. The decision-making hierarchical module adopts a spatiotemporal causal graph network to model the complex relationship between people, vehicles, roads, and environments through causal reasoning technology; for example, it can identify the causal chain of "rainy day → slippery road → increased braking distance"; hierarchical deep reinforcement learning is used to decompose the driver's decision-making behavior into macro, meso, and micro layers. Significance: Through hierarchical modeling, the multi-level characteristics of driving behavior can be more comprehensively understood, thereby providing more accurate decision support; the hierarchical structure not only improves the interpretability of the model, but also enhances its adaptability in different scenarios. The risk quantification module captures the dynamic characteristics of risks in real time by simulating the process of risk propagation in the traffic network; combined with Bayesian reasoning and Monte Carlo sampling, the module can quantify the probability of occurrence and scope of impact of potential risks; for example, it can predict how risks will be transmitted to surrounding areas after congestion occurs on a certain road section, and identify key nodes. Significance: The risk quantification module provides a scientific risk assessment tool for traffic management departments, enabling them to take preventive measures in advance to reduce the possibility of accidents; at the same time, by identifying key nodes, the module can also help optimize the allocation of traffic resources and improve the operating efficiency of the overall traffic network.

[0095] In summary, the design of the second modeling component of this embodiment fully considers the complexity and dynamics of the traffic system, and realizes the comprehensive modeling and accurate prediction of the traffic environment through the synergy of the three modules of weight distribution, decision stratification and risk quantification. It can not only improve the intelligence level of the traffic system, but also effectively reduce the incidence of traffic accidents, optimize the utilization efficiency of traffic resources, and provide strong technical support for the construction of smart cities.

[0096] Embodiment 5:

[0097] like Figure 5 As shown, based on Example 4, the decision-making hierarchical module provided in this embodiment of the present invention includes:

[0098] The spatiotemporal data processing submodule is configured to extract spatiotemporal information from behavior data, vehicle status data, road condition data, and environmental data, unify timestamps and spatial coordinates through spatiotemporal alignment, and mine key features with spatiotemporal characteristics using high-order features;

[0099] The spatiotemporal causal graph construction submodule is configured to abstract elements such as people, vehicles, roads and the environment into nodes of a graph network; each node contains not only static attributes (such as vehicle type and road grade) but also dynamic attributes (such as vehicle speed and road slipperiness); the causal relationship between nodes is determined through causal reasoning, and the causal relationship is represented by directed edges and endowed with time delay attributes;

[0100] Among them, spatiotemporal information is extracted from behavioral data, vehicle status data, road condition data, and environmental data, and reassembled through spatiotemporal alignment to unify into a continuous spatiotemporal framework; under the unified spatiotemporal framework, the data is preliminarily abstracted to extract low-order spatiotemporal features, dynamic attributes such as speed and acceleration are extracted from vehicle status data, and static attributes such as road surface type and slope are extracted from road condition data; on the basis of low-order features, high-order features with spatiotemporal characteristics are constructed through multi-level feature combination and transformation:

[0101] Feature evolution in the time dimension: Capture the law of feature changes over time, analyze the fluctuation pattern of vehicle speed in different time periods, and identify the speed change trend during peak hours;

[0102] Feature correlation in spatial dimension: Analyze the distribution and correlation of features in space, and analyze how the traffic flow of a certain road section affects the traffic flow distribution of adjacent road sections;

[0103] Generation of spatiotemporal coupling features: The features in the time and space dimensions are coupled to generate high-order features with spatiotemporal characteristics. Combining the speed change in time and the traffic flow distribution in space, the "spatiotemporal density" feature is generated to describe the traffic pressure changes in a certain area in different time periods.

[0104] For example, rainy days (environmental node) cause the road surface to be slippery (road node), which in turn affects the driver's braking distance (human node).

[0105] The reasoning and hierarchical modeling submodule is configured to analyze the overall operating status of the traffic system at a global scale; for example, through a causal graph network, it can infer whether congestion in a certain section of road will lead to increased traffic pressure in the surrounding area and predict its impact on the driver's route choice; at a local scale, it can analyze the causal relationship in a specific environment; for example, it can identify whether the failure of a signal light at a certain intersection will lead to an increase in vehicle avoidance behavior, thereby affecting the stability of traffic flow; at an instantaneous scale, it can capture the driver's response to emergencies; for example, through a causal graph network, it can infer whether the emergency braking of a vehicle will cause a chain reaction of the vehicles behind it, thereby causing local traffic risks;

[0106] The causal chain identification and optimization submodule is configured to extract key causal chains from the spatiotemporal causal graph; for example, to identify the complete causal chain of "rainy day → slippery road → increased braking distance → increased risk of rear-end collision"; through reverse causal reasoning, identify the key nodes in the causal chain and propose optimization strategies.

[0107] In the above embodiment, the spatiotemporal data processing submodule extracts and unifies spatiotemporal information from behavioral data, vehicle status data, road condition data, and environmental data; through spatiotemporal alignment technology, the data from different sources (such as vehicle trajectories, road conditions, environmental parameters, etc.) are unified in terms of timestamps and spatial coordinates to ensure data consistency and comparability; at the same time, high-order feature extraction technology is used to mine key features with spatiotemporal characteristics from behavioral data, vehicle status data, road condition data, and environmental data, such as the spatiotemporal distribution of traffic flow, the spatiotemporal patterns of driving behavior, etc. Significance: It provides a high-quality data foundation for causal analysis and modeling; by unifying spatiotemporal information, it avoids analysis errors caused by data fragmentation; by extracting high-order features, it can more accurately capture key dynamic changes in the traffic system, providing strong support for causal reasoning and optimization; the spatiotemporal causal graph construction submodule abstracts various elements in the traffic system (people, vehicles, roads, environment, etc.) into nodes of the graph network, and determines the causal relationship between nodes through causal reasoning; each node contains not only static attributes (such as vehicle type, road grade), but also dynamic attributes (such as vehicle speed, road slipperiness); causal relationships are represented by directed edges and given time delay attributes, thereby constructing a dynamic spatiotemporal causal graph. Significance: It provides a structured framework for causal analysis of traffic systems; by abstracting complex systems into graph networks, it can intuitively display the interaction relationship between various elements; by introducing time delay attributes, it can more realistically reflect the dynamic characteristics of causal relationships; it lays the foundation for analysis at global, local and instantaneous scales. The reasoning and hierarchical modeling submodule analyzes the overall operation status of the traffic system at a global scale, such as inferring whether congestion in a certain section of road will lead to increased traffic pressure in the surrounding area and predicting its impact on the driver's route choice; the local scale analyzes the causal relationship in a specific environment, such as identifying whether the failure of a signal light at a certain intersection will lead to an increase in vehicle avoidance behavior, thereby affecting the stability of traffic flow; the instantaneous scale captures the driver's response to emergencies, such as inferring whether the emergency braking of a vehicle will cause a chain reaction of vehicles behind, thereby causing local traffic risks. Significance: It realizes multi-level modeling of the traffic system, which can comprehensively analyze the operation status and potential risks of the traffic system from macro to micro, from global to local, from long-term to instantaneous; the hierarchical analysis method can more accurately identify the root causes of traffic problems and provide a scientific basis for the formulation of optimization strategies. The causal chain identification and optimization submodule extracts key causal chains from the spatiotemporal causal graph, such as identifying the complete causal chain of "rainy day → slippery road → increased braking distance → increased risk of rear-end collision"; through reverse causal reasoning, it identifies the key nodes in the causal chain and proposes optimization strategies; for example, by improving the road drainage system or reminding the driver to slow down, the risk of rear-end collision can be reduced. Significance: It achieves accurate identification and optimization of key causal chains in the transportation system.By extracting the causal chain, the evolution process of traffic problems can be clearly displayed; through reverse causal reasoning, the root cause of the problem can be quickly located and targeted optimization strategies can be proposed; it not only improves the operating efficiency of the transportation system, but also significantly reduces the probability of traffic accidents.

[0108] In summary, the various submodules of the decision-making hierarchical module of this embodiment cooperate with each other to jointly realize the multi-level and multi-dimensional analysis and optimization of complex traffic systems; from data extraction to causal reasoning, from global modeling to local optimization, it provides a scientific and accurate decision support tool for traffic management. Its significance lies in: improving the operating efficiency of the traffic system through accurate analysis and optimization, reducing traffic congestion and accidents, and improving overall operating efficiency; enhancing the safety of the traffic system by identifying and optimizing key causal chains, reducing the probability of traffic accidents; supporting intelligent traffic management to provide data-driven decision support for traffic management departments, and promote traffic management to develop in the direction of intelligence and refinement. It not only embodies the combination of technology and humanities, but also provides important support for the sustainable development of future transportation systems.

[0109] Embodiment 6:

[0110] like Figure 6 As shown, based on Example 5, the reasoning and hierarchical modeling submodule provided in this embodiment of the present invention includes:

[0111] The first association unit is configured to generate guidance suggestions for local areas through data analysis and model prediction; if the global data shows that the traffic pressure in a certain area exceeds the standard setting in a certain period of time, the local area will deploy measures such as traffic signal optimization and increase police force to guide traffic in advance;

[0112] The second association unit is configured as a meso-level between the global and instantaneous levels. Local-scale data (such as traffic flow at an intersection, signal light status) is transmitted to the instantaneous scale in real time to predict the driver's instantaneous behavior. If the traffic flow at a certain intersection suddenly increases, it is predicted that the driver may choose to change lanes or slow down.

[0113] The third association unit is configured to provide feedback to the global and local systems through real-time collection and transmission; the driver's instantaneous behavior data is used to update the global traffic pressure distribution map or optimize the signal light control strategy at the local intersection.

[0114] In the above embodiment, the first association unit provides guidance and suggestions from global to local, macro data analysis and prediction, and generates guidance and suggestions for local areas through global scale data analysis (such as traffic pressure distribution, historical flow data, etc.). The macro perspective provides a scientific basis for local management; dynamic strategy deployment When global data shows that the traffic pressure of a certain area exceeds the preset standard in a specific time period, the system can automatically trigger optimization measures for the local area, such as traffic signal optimization, police force diversion, etc.; resource optimization allocation Through the guidance of global data, the management resources of the local area (such as police force, traffic light control) can be accurately allocated to avoid waste or shortage of resources. The significance achieved: The guidance of global data makes the traffic management of local areas more efficient, avoiding the localized problem of "treating the symptoms rather than the root cause"; by deploying measures in advance, it can effectively prevent the occurrence of traffic congestion and reduce the negative impact of emergencies; the support of global data makes decision-making more scientific and data-based, reducing the subjectivity of human judgment. The second association unit supports local to instantaneous meso-information. Local-scale data (such as traffic flow at intersections and signal light status) are collected in real time and transmitted to the instantaneous-scale model to support instantaneous behavior prediction. Based on the meso-information of local intersections, the system can predict the driver's instantaneous behavior (such as lane change, deceleration, acceleration, etc.), thereby providing a more fine-grained decision-making basis for traffic management. When the traffic flow at a certain intersection suddenly increases, it can respond quickly, predict the driver's possible behavior, and dynamically adjust the traffic management strategy for the local area. Significance achieved: The support of meso-information enables traffic management to go deep into the driver's instantaneous behavior level and achieve more refined management. The real-time transmission and dynamic analysis of local data improves the accuracy of predicting driver behavior and provides a more reliable basis for traffic management. It can quickly adjust strategies according to the real-time status of local intersections, enhancing the flexibility and adaptability of traffic management. The third association unit provides instantaneous feedback to the global and local areas; the driver's instantaneous behavior data (such as acceleration, deceleration, lane change, etc.) is collected in real time and transmitted to the global and local models to update the traffic pressure distribution map and optimize the signal light control strategy at the local intersection; the instantaneous scale data can be fed back to the global and local models in real time, so that the traffic pressure distribution map and signal light control strategy can be dynamically updated to adapt to the changing traffic conditions; through the feedback of instantaneous data, the global and local models can form a closed-loop optimization mechanism to continuously adjust and optimize the traffic management strategy. The significance achieved: the feedback of instantaneous data enables the global and local models to be updated in real time, realizing data-driven dynamic optimization; the real-time transmission and feedback of instantaneous data enables the system to quickly respond to changes in traffic conditions and improve the response speed of traffic management; through the closed-loop optimization mechanism, the system can continuously learn and adjust, enhancing the adaptability and intelligence level of traffic management.

[0115] In summary, the three units of this embodiment realize multi-level collaborative management from the global, local and instantaneous scales respectively, and build a complete traffic management ecosystem; through the flow and feedback of data, it can realize intelligent management from macro to micro, and improve the scientificity and accuracy of traffic management; the dynamic feedback mechanism of the three units enables the system to continuously optimize and adjust, adapt to the complex and changing traffic environment, and finally realize efficient, intelligent and humanized traffic management. Through this multi-level and multi-dimensional synergy, the reasoning and hierarchical modeling submodule not only improves the technical level of traffic management, but also provides important technical support for the sustainable development of urban transportation.

[0116] Embodiment 7:

[0117] like Figure 7 As shown, based on Example 4, the risk quantification module provided by the embodiment of the present invention includes:

[0118] The risk source identification submodule is configured to initialize the risk source as a node in the network and establish connections with other nodes through a causal network (such as a spatiotemporal causal graph); dynamically expand the causal chain of risks to capture the chain reaction between multiple events; continuously update the causal chain through real-time data (such as traffic flow, weather changes, driver behavior, etc.) to capture the dynamic characteristics of risk propagation;

[0119] The Bayesian reasoning submodule is configured to construct a priori probabilities for each risk source and causal chain. Through the input of real-time data, the Bayesian reasoning is used to dynamically update the posterior probability of the risk. The macro level evaluates global risks (such as the probability of congestion in the entire road network), the meso level evaluates local risks (such as the probability of rear-end collision at a certain intersection), and the micro level evaluates instantaneous risks (such as the probability of insufficient braking distance for a certain vehicle).

[0120] The risk impact quantitative analysis submodule is configured to capture the uncertainty and complexity of risk propagation through a large number of random simulations; generate a large number of possible risk propagation paths based on the results of causal chains and Bayesian reasoning; simulate global risk propagation at the macro level (such as the impact of rainy days on the entire road network), local risk propagation at the meso level (such as the impact of a rear-end collision at a certain intersection on surrounding roads), and instantaneous risk propagation at the micro level (such as the impact of a vehicle's braking behavior on subsequent vehicles); and quantify the scope of risk impact at the same time;

[0121] The path and key node identification submodule is configured to dynamically identify the key paths of risk transmission based on the results of Bayesian reasoning and Monte Carlo sampling; locate key nodes (such as high-risk intersections and sections of road prone to accidents) by analyzing the risk propagation paths; adjust the global traffic signal control strategy at the macro level, deploy local police forces for guidance at the meso level, and send real-time warning information to drivers at the micro level.

[0122] In the above embodiment, the risk source identification submodule initializes the risk source (such as rainy days, traffic accidents) as a node of the network, and establishes connections with other nodes through a causal network (such as a spatiotemporal causal graph) to form a basic framework for risk propagation; through causal reasoning technology, the causal chain of risks is dynamically expanded to capture the chain reaction between multiple events; for example, rainy days cause the road surface to be slippery, which in turn affects the driver's braking distance, and may eventually lead to rear-end collisions; through real-time data (such as traffic flow, weather changes, driver behavior, etc.), the causal chain is continuously updated to capture the dynamic characteristics of risk propagation and ensure the timeliness and accuracy of risk identification. Significance: By initializing the risk source and establishing a causal network, the starting point of the risk can be accurately located; through the dynamic expansion of the causal chain and the update of real-time data, the dynamic characteristics of risk propagation can be captured, providing a basis for risk quantification; by capturing the chain reaction between multiple events, the complexity of the risk can be comprehensively assessed, providing a scientific basis for traffic management. The Bayesian reasoning submodule builds a priori probabilities for each risk source and causal chain based on historical data and expert knowledge to ensure the basic reliability of risk quantification; through the input of real-time data, the posterior probability of risk is dynamically updated using Bayesian reasoning to ensure the timeliness and accuracy of risk quantification. Significance: Through Bayesian reasoning, the system can dynamically update the probability of risk occurrence and enhance the timeliness of risk quantification; through macro, meso and micro level evaluation, it can comprehensively capture the characteristics of risk propagation from global to local, from static to dynamic; through accurate risk probability evaluation, it provides a scientific basis for traffic management and enhances the pertinence and effectiveness of risk management. The risk impact quantification analysis submodule captures the uncertainty and complexity of risk propagation through a large number of random simulations (Monte Carlo sampling) to ensure the comprehensiveness of risk quantification; based on the results of causal chains and Bayesian reasoning, it generates a large number of possible risk propagation paths to ensure the diversity and comprehensiveness of risk quantification; through Monte Carlo sampling, it quantifies the impact range of risks (for example, the probability that a rear-end collision at a certain intersection may cause congestion on surrounding roads is 70%, and the impact range is 500 meters). Significance: Comprehensive risk propagation analysis, through Monte Carlo sampling, can capture the uncertainty and complexity of risk propagation and enhance the comprehensiveness of risk quantification; hierarchical risk propagation simulation can capture the characteristics of risk propagation from global to local, from static to dynamic through simulation at three levels: macro, meso and micro; scientific risk impact assessment provides a scientific basis for traffic management by quantifying the scope of risk impact, and enhances the pertinence and effectiveness of risk management. The path and key node identification submodule dynamically identifies the key path of risk transmission based on the results of Bayesian reasoning and Monte Carlo sampling to ensure the accuracy of risk management; by analyzing the risk propagation path, key nodes (such as high-risk intersections and sections prone to accidents) are located to ensure the pertinence of risk management.Significance: Accurate identification of risk transmission paths: By dynamically identifying the key paths of risk transmission, the key areas of risk transmission can be accurately located; by locating key nodes, targeted management strategies can be formulated for high-risk areas; through management strategies at the macro, meso and micro levels, the spread of risks can be comprehensively controlled from the global to the local, from static to dynamic.

[0123] In summary, this embodiment can accurately identify the risk source and quantify its occurrence probability through the collaborative work of the risk source identification submodule and the Bayesian reasoning submodule; through the collaborative work of the risk impact quantification analysis submodule and the path and key node identification submodule, it can comprehensively analyze the characteristics of risk propagation and quantify its impact range; through the collaborative work of each submodule, it can dynamically adjust the risk management strategy, from global to local, from static to dynamic, and comprehensively control the spread of risk; through the collaborative work of each submodule, it provides a scientific basis for traffic management, enhances the pertinence and effectiveness of risk management, and provides important technical support for the sustainable development of urban transportation. This hierarchical and dynamic risk quantification method not only improves the technical level of traffic management, but also lays a solid foundation for the intelligent development of urban transportation.

[0124] Embodiment 8:

[0125] like Figure 8 As shown, based on Example 7, the path and key node identification submodule provided in this embodiment of the present invention includes:

[0126] A dynamic identification unit is configured to identify global risk transmission paths at the macro level, local risk transmission paths at the meso level, and instantaneous risk transmission paths at the micro level;

[0127] Among them, the identification of global risk transmission paths at the macro level:

[0128] Data-driven causal chain expansion dynamically expands the risk causal chain based on real-time data (such as traffic flow, weather changes, driver behavior, etc.) to capture potential paths of global risk transmission; dynamic screening of key paths screens out the key paths that have the greatest impact on global risk transmission (such as the transmission path of congestion on major roads caused by rainy days) by analyzing the weight and transmission speed of the causal chain; dynamic update of paths continuously adjusts the weight and priority of key paths with the input of real-time data to ensure the timeliness of path identification;

[0129] Identification of local risk transmission paths at the meso level:

[0130] Refined modeling of local causal chains: Based on the global path, the causal chains in the local area are further refined to capture the characteristics of local risk transmission (such as the impact of a rear-end collision at a certain intersection on surrounding roads); dynamic positioning of local key paths: By analyzing the propagation speed and impact range of the local causal chain, the key paths of local risk transmission are located (such as the path where a rear-end collision at a certain intersection may cause congestion on surrounding roads); dynamic adjustment of paths: According to the real-time data of the local area (such as traffic flow, accident rate, etc.), the weights and priorities of local key paths are dynamically adjusted;

[0131] Identification of instantaneous risk transmission paths at the micro level:

[0132] Real-time capture of instantaneous causal chains is based on micro data (such as the braking behavior of a certain vehicle, speed changes, etc.), capturing the causal chain of instantaneous risk transmission in real time; dynamic identification of instantaneous critical paths identifies the critical paths of instantaneous risk transmission (such as the path where the braking behavior of a certain vehicle causes the subsequent vehicle to rear-end) by analyzing the propagation speed and impact range of the instantaneous causal chain; real-time update of paths updates the weights and priorities of instantaneous critical paths in real time with the input of micro data;

[0133] A dynamic positioning unit is configured to position global key nodes at the macro level, local key nodes at the meso level, and instantaneous key nodes at the micro level;

[0134] Among them, the positioning of global key nodes at the macro level:

[0135] The weight analysis of the global causal chain locates the key nodes (such as intersections of major roads) that have the greatest impact on the global risk propagation by analyzing the weight and propagation speed of the global causal chain; the dynamic screening of key nodes dynamically screens out the key nodes that have the greatest impact on the global risk propagation based on real-time data (such as traffic flow, weather changes, etc.); the dynamic adjustment of nodes dynamically adjusts the weights and priorities of key nodes as global data is input;

[0136] Positioning of local key nodes at the meso-level:

[0137] Among them, the weight analysis of the local causal chain locates the key nodes (such as the intersection of a certain road) that have the greatest impact on the local risk propagation by analyzing the weight and propagation speed of the local causal chain; the dynamic screening of key nodes dynamically screens out the key nodes that have the greatest impact on the local risk propagation based on the real-time data of the local area (such as traffic flow, accident rate, etc.); the dynamic adjustment of nodes dynamically adjusts the weights and priorities of key nodes as the local data is input; the positioning of instantaneous key nodes at the micro level, the weight analysis of the instantaneous causal chain locates the key nodes (such as the braking point of a certain vehicle) that have the greatest impact on the instantaneous risk propagation by analyzing the weight and propagation speed of the instantaneous causal chain; the dynamic screening of key nodes dynamically screens out the key nodes that have the greatest impact on the instantaneous risk propagation based on micro data (such as vehicle speed, braking distance, etc.); the real-time update of nodes updates the weights and priorities of key nodes in real time as the micro data is input;

[0138] The quantitative output unit is configured to quantify potential risks through multi-level causal chain decomposition and dynamic probability propagation.

[0139] In the above embodiments, the dynamic identification unit realizes comprehensive coverage and dynamic capture of risk transmission through risk transmission path identification at the macro, meso and micro levels; the global risk transmission path identification at the macro level dynamically captures the potential path of global risk transmission through data-driven causal chain expansion; the key path with the greatest impact on global risk transmission is identified through dynamic screening of key paths; the timeliness and accuracy of the global risk transmission path are ensured through dynamic updating of the path; the local risk transmission path identification at the meso level captures the characteristics of local risk transmission through refined modeling of local causal chains; the key path with the greatest impact on local risk transmission is identified through dynamic positioning of local key paths; the timeliness and accuracy of the local risk transmission path are ensured through dynamic adjustment of the path; the instantaneous risk transmission path identification at the micro level dynamically identifies the causal chain of instantaneous risk transmission through real-time capture of the instantaneous causal chain, captures the key path of instantaneous risk transmission through dynamic identification of instantaneous key paths, and ensures the timeliness and accuracy of the instantaneous risk transmission path through real-time updating of the path. Significance achieved: The dynamic identification unit achieves comprehensive coverage and dynamic capture of risk transmission through multi-level risk transmission path identification; it not only improves the accuracy of risk identification, but also provides a solid foundation for risk quantification and prevention and control. The dynamic positioning unit realizes the dynamic capture and optimization of key nodes of risk propagation by locating key nodes at the macro, meso and micro levels; the global key node positioning at the macro level locates the key nodes with the greatest impact on global risk propagation through the weight analysis of the global causal chain; the timeliness and accuracy of the global key nodes are ensured through the dynamic screening of key nodes; the weights and priorities of the global key nodes are optimized through the dynamic adjustment of nodes; the local key node positioning at the meso level locates the key nodes with the greatest impact on local risk propagation through the weight analysis of the local causal chain; the timeliness and accuracy of the local key nodes are ensured through the dynamic screening of key nodes; the weights and priorities of the local key nodes are optimized through the dynamic adjustment of nodes; the instantaneous key node positioning at the micro level locates the key nodes with the greatest impact on instantaneous risk propagation through the weight analysis of the instantaneous causal chain; the timeliness and accuracy of the instantaneous key nodes are ensured through the dynamic screening of key nodes; the weights and priorities of the instantaneous key nodes are optimized through the real-time updating of nodes. Significance achieved: The dynamic positioning unit realizes the dynamic capture and optimization of key nodes of risk propagation through multi-level key node positioning; it not only improves the accuracy of risk prevention and control, but also provides a scientific basis for risk management decision-making.The quantitative output unit achieves accurate quantification of potential risks through multi-level causal chain decomposition and dynamic probability propagation; macro-level risk quantification, through the decomposition of the global causal chain and dynamic probability propagation, quantifies the risk probability distribution of the entire road network, outputs the probability distribution map of high-risk areas, and provides data support for global risk management; meso-level risk quantification, through the decomposition of local causal chains and dynamic probability propagation, quantifies the risk transmission path of local sections, outputs the risk accumulation probability of local key nodes, and provides data support for local risk management; micro-level risk quantification, through the decomposition of instantaneous causal chains and dynamic probability propagation, quantifies the risk occurrence probability of instantaneous events, outputs the probability of the impact range of instantaneous events, and provides data support for micro-risk management. Significance achieved: The quantitative output unit achieves accurate assessment of potential risks through multi-level risk quantification; it not only improves the scientific nature of risk management, but also provides reliable data support for the formulation and optimization of risk prevention and control measures.

[0140] In summary, the dynamic identification unit, dynamic positioning unit and quantitative output unit of this embodiment achieve comprehensive coverage, dynamic capture and accurate quantification of potential risks through multi-level risk transmission path identification, key node positioning and risk quantification. The multi-level and dynamic risk identification and quantification mechanism not only improves the accuracy and scientificity of risk management, but also provides a strong technical guarantee for the safe operation of the transportation network.

[0141] The difference between this embodiment and the prior art: hierarchical path identification and node positioning. The prior art usually adopts a single-level risk transmission path identification and node positioning, while this embodiment realizes comprehensive capture from global to local, from static to dynamic through path identification and node positioning at three levels: macro, meso and micro. Dynamic path and node adjustment. The prior art usually adopts a static path identification and node positioning method, while this embodiment dynamically adjusts the weights and priorities of paths and nodes through the input of real-time data, thereby enhancing the timeliness and accuracy of risk management. Data-driven causal chain expansion. The prior art usually adopts a fixed causal chain expansion method, while this embodiment dynamically expands the causal chain through real-time data, captures the dynamic characteristics of risk transmission, and enhances the comprehensiveness and accuracy of path identification. Refined modeling of weight analysis. The prior art usually adopts a simple weight analysis method, while this embodiment accurately locates key paths and key nodes by analyzing the weight and propagation speed of the causal chain, thereby enhancing the pertinence and effectiveness of risk management.

[0142] Embodiment 9:

[0143] like Fig. 9 As shown, based on Embodiment 8, the quantization output unit provided in the embodiment of the present invention includes:

[0144] The area identification subunit is configured to provide the causal structure and initial probability distribution of risk propagation for Bayesian reasoning, while Monte Carlo sampling verifies and optimizes the probability distribution through random simulation. The macro layer outputs the risk probability distribution map of the entire road network and identifies high-risk areas (such as sections prone to congestion or accidents).

[0145] The risk prediction subunit is configured to provide the causal structure and initial probability distribution of local risk transmission through Bayesian reasoning, while Monte Carlo sampling verifies and optimizes these probability distributions through random simulation. Finally, the meso-level outputs the risk transmission path diagram of the local road section and predicts the risk accumulation of key nodes (such as congestion or accident-prone points on a specific road section).

[0146] The time output subunit is configured to provide the causal structure and initial probability distribution of the risk propagation of instantaneous events for Bayesian reasoning, while Monte Carlo sampling verifies and optimizes these probability distributions through random simulation; finally, the micro layer outputs the risk probability of instantaneous events and their local and global impact range (such as the chain reaction of accidents on traffic flow).

[0147] In the above embodiment, the area identification subunit constructs the causal structure of risk propagation through Bayesian reasoning, and optimizes the initial probability distribution by Monte Carlo sampling, and finally outputs the risk probability distribution map of the entire road network; the subunit can identify high-risk areas (such as sections prone to congestion or accidents) and quantify the risk levels of these areas. Significance achieved: Macro risk perception provides the traffic management department with the risk distribution of the entire road network, helping it to quickly locate high-risk areas and formulate targeted management strategies; resource optimization allocation optimizes the allocation of traffic resources (such as police deployment and emergency resource allocation) by identifying high-risk areas, and improves the efficiency and effectiveness of traffic management; preventive management identifies high-risk areas in advance and takes preventive measures (such as speed limit, traffic light optimization) to reduce the probability of accidents. The risk prediction subunit constructs the causal structure of local risk transmission through Bayesian reasoning, and optimizes the initial probability distribution by Monte Carlo sampling, and finally outputs the risk transmission path map of the local section; the subunit can predict the risk accumulation of key nodes (such as congestion or accident-prone points in a specific section) and quantify the risk levels of these nodes. Significance achieved: Meso-risk prediction provides traffic management departments with risk transmission paths for local sections, helping them predict the risk accumulation of key nodes and take intervention measures in advance; precision intervention predicts the risk accumulation of key nodes and formulates precise intervention strategies (such as dynamic speed limits and traffic light optimization) to reduce the probability of accidents; traffic flow optimization optimizes the risk transmission paths of local sections to improve the efficiency of traffic flow and alleviate congestion. The time output subunit constructs the causal structure of instantaneous event risk propagation through Bayesian reasoning, and optimizes the initial probability distribution using Monte Carlo sampling, and finally outputs the risk probability of instantaneous events and their local and global impact range (such as the chain reaction of accidents on traffic flow); this subunit can quantify the risk level of instantaneous events and their impact range. Significance achieved: Micro-risk perception provides traffic management departments with the risk probability of instantaneous events and their impact range, helping them to respond quickly to emergencies and reduce the impact of accidents on traffic flow; emergency management quantifies the risk level of instantaneous events and their impact range, formulates emergency management strategies (such as path guidance and traffic light optimization), and reduces the chain reaction of accidents on traffic flow; traffic flow recovery optimizes the impact range of instantaneous events, improves the recovery speed of traffic flow, and reduces the long-term impact of accidents on traffic flow.

[0148] In summary, the region identification subunit, risk prediction subunit and time output subunit of this embodiment respectively realize the quantitative output of traffic risks from the macro, meso and micro levels; the region identification subunit helps the traffic management department to quickly locate high-risk areas through macro risk perception; the risk prediction subunit helps the traffic management department to predict the risk accumulation of key nodes through meso risk prediction; the time output subunit helps the traffic management department to quickly respond to emergencies through micro risk perception. The combination of the three subunits not only improves the accuracy and timeliness of traffic risk management, but also provides a strong technical guarantee for the safe operation of the traffic network.

[0149] Embodiment 10:

[0150] like Fig.10 As shown, based on Example 1, the active intervention component provided by the embodiment of the present invention includes:

[0151] The dynamic stratification module is configured to classify the risks in the traffic network into three levels: low, medium and high, based on the real-time risk assessment results, and further refine them into three levels: region, road section and node. The risk level of each level is calculated by dynamic weighting of multi-dimensional data (such as traffic flow, accident probability, environmental conditions, etc.);

[0152] The logic building module is configured to establish a multi-level strategy matching mechanism for different risk levels and layers; low-risk areas use suggestive interventions, medium-risk areas use preventive interventions, and high-risk areas use emergency interventions; the strategy matching logic is based on the dynamic analysis of risk transmission paths;

[0153] Among them, the generation of dynamic speed limit strategy:

[0154] Regional dynamic speed limit: By analyzing the traffic flow, accident probability and environmental conditions in the area, the regional speed limit is dynamically adjusted; for example, in rainy days or low visibility conditions, the regional speed limit is reduced to prevent accidents; during peak traffic flow, the speed limit is optimized to alleviate congestion;

[0155] Dynamic speed limit on road sections: adjust the speed limit of road sections in real time according to the risk transmission characteristics of the road sections (such as sections with high incidence of rear-end collisions and curves, etc.); generate dynamic speed limit strategies by analyzing the historical accident data and real-time traffic status of the road sections to ensure the safety of road section traffic;

[0156] Dynamic speed limit at nodes: for key nodes (such as intersections, ramps, etc.), the node speed limit is adjusted dynamically by analyzing the traffic flow and risk transmission path of the node; for example, when the intersection is congested or accidents occur frequently, the node speed limit is reduced to prevent secondary accidents;

[0157] Generation of traffic light optimization strategy:

[0158] Regional traffic light collaborative optimization: By analyzing the traffic flow and risk transmission paths in the region, the timing of regional traffic lights is dynamically adjusted; for example, during peak traffic flow, the green light time on the main roads is extended to ease congestion; in accident-prone areas, the timing of traffic lights is optimized to reduce accident risks;

[0159] Dynamic adjustment of traffic lights on sections. According to the risk transmission characteristics of the sections (such as sections with high incidence of rear-end collisions and sections with pedestrian crossings, etc.), the timing plan of traffic lights on sections is adjusted in real time. By analyzing the real-time traffic status and historical accident data of the sections, dynamic traffic light optimization strategies are generated to ensure the safety of traffic on the sections.

[0160] Precise control of node traffic lights. For key nodes (such as intersections, ramps, etc.), the timing of node traffic lights is dynamically adjusted by analyzing the traffic flow and risk transmission path of the nodes. For example, when the intersection is congested or accidents occur frequently, the timing of traffic lights is optimized to prevent secondary accidents.

[0161] Generation of path induction strategy:

[0162] Regional route guidance: By analyzing the traffic flow and risk transmission paths in the region, regional route guidance schemes are dynamically generated; for example, vehicles are guided to bypass congested areas during peak traffic flow periods; and vehicles are guided to avoid high-risk sections in accident-prone areas.

[0163] Road section path guidance: based on the risk transmission characteristics of the road section (such as sections with high incidence of rear-end collision accidents, construction sections, etc.), the road section path guidance plan is generated in real time; by analyzing the real-time traffic status and historical accident data of the road section, a dynamic path guidance strategy is generated to ensure the safety of road section traffic;

[0164] Node path induction: for key nodes (such as intersections, ramps, etc.), by analyzing the traffic flow and risk transmission path of the nodes, dynamically generate node path induction plans; for example, when the intersection is congested or accidents are frequent, instruct vehicles to bypass high-risk nodes;

[0165] The dynamic update optimization module is configured to dynamically update the intervention strategy based on real-time risk assessment results and traffic status data; for example, when traffic flow changes or environmental conditions change, the speed limit, signal light timing and path guidance plan are adjusted in real time; by analyzing the implementation effect of the intervention strategy (such as reduced accident rate, improved traffic efficiency, etc.), the strategy generation logic is dynamically optimized.

[0166] In the above embodiment, the dynamic stratification module accurately identifies the risk level and level in the traffic network through dynamic weighted calculation of multi-dimensional data, providing a scientific basis for the formulation of intervention strategies; dynamically adjusts the risk level and level according to the real-time risk assessment results to ensure the timeliness and pertinence of the intervention strategy; combines multi-dimensional data such as traffic volume, accident probability, and environmental conditions to fully perceive the risk status of the traffic network and improve the accuracy of risk identification. The logic construction module formulates differentiated intervention strategies according to the risk level and level to ensure the accuracy and effectiveness of intervention measures; prevents accidents and alleviates congestion through dynamic speed limit strategies for regions, sections, and nodes, and improves the safety and efficiency of traffic networks; optimizes the traffic flow efficiency and reduces the risk of accidents through signal light optimization strategies for regions, sections, and nodes; guides vehicles to avoid high-risk areas and sections through path induction strategies for regions, sections, and nodes, and improves the overall traffic efficiency of the traffic network. The dynamic update optimization module dynamically updates the intervention strategy according to the real-time traffic status data to ensure the timeliness and adaptability of the intervention measures. It dynamically optimizes the strategy generation logic by analyzing the implementation effect of the intervention strategy to improve the scientificity and effectiveness of the intervention strategy. Through dynamic updating and optimization, it continuously improves the traffic safety and efficiency of the traffic network and realizes the intelligent and refined traffic management.

[0167] In summary, the dynamic stratification module of this embodiment provides a scientific basis for the formulation of intervention strategies through accurate risk stratification and dynamic response capabilities; the logic construction module realizes accurate intervention of traffic risks through multi-level strategy matching and the generation of dynamic speed limit, signal light optimization, and path induction strategies; the dynamic update optimization module continuously optimizes the intervention strategy through real-time dynamic update and strategy effect evaluation. The combination of the three modules not only improves the accuracy and timeliness of traffic risk management, but also provides a strong technical guarantee for the safe operation and efficient passage of the traffic network. The generation process of active intervention strategy realizes accurate intervention of traffic risks through dynamic stratification of risk levels, logical construction of strategy matching, generation of dynamic speed limit strategy, generation of signal light optimization strategy, generation of path induction strategy, multi-level warning and information push, and dynamic update and optimization of strategy. This hierarchical, creative and adaptive active intervention strategy generation method not only improves the accuracy and timeliness of traffic risk management, but also provides a strong technical guarantee for the safe operation of the traffic network.

[0168] Embodiment 11:

[0169] Fig.11A block diagram of an exemplary electronic device suitable for implementing an embodiment of the present invention is shown. The electronic device may include a central processing unit / microprocessor / main control chip, etc. 1; a storage medium 2, coupled to the central processing unit / microprocessor / main control chip, etc. 1, and storing computer executable instructions therein, for performing the steps of each method of the embodiment of the present invention when executed by the processor. The central processing unit / microprocessor / main control chip, etc. 1 may include, but is not limited to, for example, one or more processors or microprocessors, etc. The storage medium 2 may include, but is not limited to, for example, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, register, computer storage medium (such as hard disk, floppy disk, solid state hard disk, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.). In addition, the electronic device may also include (but is not limited to) a data bus 3, an input / output bus / external bus / device bus, etc. 4, a display 5, and an input / output device 6 (such as a keyboard, a mouse, a speaker, etc.). The central processing unit / microprocessor / main control chip, etc. 1 may communicate with external devices (5, 6, etc.) through the I / O bus 4 via a wired or wireless network (not shown). The storage medium 2 may also store at least one computer executable instruction for executing the various functions and / or method steps in the embodiments described in the present technology when the central processing unit / microprocessor / main control chip 1 is executed. In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product, wherein one or more computer executable instructions are executed by the processor to execute the various functions and / or method steps in the embodiments described in the present technology.

[0170] Fig.12 A schematic diagram of a computer-readable storage medium according to an embodiment of the present invention is shown. Fig.12As shown, instructions are stored on the non-transitory computer-readable storage medium 8, and the instructions are, for example, computer-readable instructions 7. When the computer-readable instructions 7 are executed by the processor, the various methods described above can be executed. The non-transitory computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-transitory non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium 8 can be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions 7 stored on the computer-readable storage medium 8, the various methods described above can be performed. In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. Another point is that the coupling or direct coupling or communication connection between each other shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional units in various embodiments of the present invention may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for executing all or part of the steps of the methods of various embodiments of the present invention through a computer device (which may be a personal computer, a server, or a network device, etc.). The aforementioned storage media include: U disk, mobile hard disk, read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), random access memory (English full name: Random Access Memory, English abbreviation: RAM), disk or CD and other media that can store program codes.

[0171] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A traffic risk dynamic assessment model construction system, characterized in that: Include: The first modeling component is configured to use spatiotemporal graph neural networks and spatiotemporal attention mechanisms to capture the coupling relationship of traffic risks in time and space and perceive the real-time risks of complex traffic networks; Construct a risk propagation model using meta-learning to simulate the process of risk propagation and diffusion in the transportation network; Generate dynamic risk maps to display the risk level and changing trends of each area in the transportation network in real time; a second modeling component configured to collect driver behavior data, vehicle status data, road condition data, and environmental data; Construct a multi-agent collaborative risk assessment model to analyze the interaction between people, vehicles, roads and the environment; use deep reinforcement learning to simulate drivers’ decision-making behaviors in different environments and predict potential risks; Analyze the transmission paths and key nodes of risks in the transportation system; Active intervention component, configured to establish a multi-level early warning mechanism based on risk level; Automatically generate early warning information using natural language and push it in real time through multiple channels; generate proactive intervention strategies; The second modeling component includes: The weight allocation module is configured to adopt a dynamic weight allocation mechanism to automatically adjust the weight of each data source according to the real-time changes of the data; The decision-making hierarchical module is configured to use a spatiotemporal causal graph network to model the causal relationship between people, vehicles, roads and the environment through causal reasoning technology; identify the causal chain that causes slippery roads due to rainy days, which in turn affects the driver's braking distance; and use hierarchical deep reinforcement learning composed of macro, meso and micro layers to decompose the driver's decision-making behavior into multiple levels; The risk quantification module is configured to simulate the risk propagation process in the transportation network, capture the dynamic characteristics of risk propagation in real time, and predict the risk transmission path and key nodes; Combining Bayesian reasoning and Monte Carlo sampling to quantify the probability of occurrence and impact of potential risks.

2. The traffic risk dynamic assessment model construction system according to claim 1 is characterized in that: The first modeling component includes: The relationship capture module is configured to collect real-time data from multiple data sources, clean the collected real-time data, remove noise and outliers, and perform standardization processing; define each intersection or road section in the traffic network as a node in the graph, and define the connection relationship between nodes as edges; use the spatiotemporal graph neural network to extract the spatiotemporal characteristics of the traffic network and capture the coupling relationship of traffic risks in time and space; The model training module is configured to define the traffic risk evolution problem as a series of related tasks, each task corresponding to a time window or spatial region; The model is trained through a meta-learner to capture the dynamic characteristics of risk evolution; The real-time display module is configured to build a risk propagation model based on the meta-learner to simulate the risk propagation and diffusion process in the transportation network; Based on the output of the risk propagation model, a dynamic risk map is generated to display the risk level and changing trend of each area in the transportation network in real time.

3. The traffic risk dynamic assessment model construction system according to claim 2 is characterized in that: Real-time display module, including: The meta-knowledge extraction submodule is configured to establish a multi-dimensional feature space, map the spatiotemporal features, environmental features, and event features in the traffic network into a unified high-dimensional space, and dynamically identify the feature dimensions that have a significant impact on risk propagation through adaptive feature selection; Construct a nonlinear correlation network between features to capture the potential coupling relationship between features; The dynamic evolutionary learning submodule is configured to design a spatiotemporal attention mechanism to automatically capture the key spatiotemporal nodes of risk propagation; Establish a differential equation model of risk propagation to simulate the diffusion process of risk in the transportation network; Introducing memory enhancement to store and update historical patterns of risk propagation; The meta-learning optimization submodule is configured to build a multi-task learning framework that treats risk evolution in different time windows and spatial regions as independent tasks; Design a knowledge transfer mechanism between tasks to achieve cross-temporal and cross-spatial migration of risk propagation models; Introducing an adaptive weight allocation strategy to dynamically adjust the contribution of different tasks to model training; The risk propagation simulation submodule is configured to establish a risk propagation network based on a probability graph and simulate the probability of risk transfer between nodes; Design a risk diffusion attenuation function to simulate the risk attenuation characteristics over time and space; construct a risk superposition model to simulate the superposition effect of multi-source risks in the transportation network; Establish a real-time prediction mechanism for risk propagation, dynamically update the risk propagation path, and continuously optimize the risk propagation model based on real-time data according to feedback correction.

4. The traffic risk dynamic assessment model construction system according to claim 1 is characterized in that: in, The macro layer simulates the driver's overall judgment of the global road conditions; the meso layer simulates the driver's response to the local environment; and the micro layer simulates the driver's response to instantaneous events.

5. The traffic risk dynamic assessment model construction system according to claim 1 is characterized in that: Decision-making hierarchical module, including: The spatiotemporal data processing submodule is configured to extract spatiotemporal information from behavior data, vehicle status data, road condition data, and environmental data, unify timestamps and spatial coordinates through spatiotemporal alignment, and mine key features with spatiotemporal characteristics using high-order features; Among them, the process of using high-order features to mine key features with spatiotemporal characteristics includes extracting spatiotemporal information from behavior data, vehicle status data, road condition data, and environmental data, re-stitching them through spatiotemporal alignment, and unifying them into a continuous spatiotemporal framework; in a unified spatiotemporal framework, the data is preliminarily abstracted to extract low-order spatiotemporal features, extracting speed and acceleration dynamic attributes from vehicle status data, and extracting road surface type and slope static attributes from road condition data; on the basis of low-order features, high-order features with spatiotemporal characteristics are constructed through multi-level feature combination and transformation: Feature evolution in the time dimension: Capture the law of feature changes over time, analyze the fluctuation pattern of vehicle speed in different time periods, and identify the speed change trend during peak hours; Feature correlation in spatial dimension: Analyze the distribution and correlation of features in space, and analyze how the traffic flow of a certain road section affects the traffic flow distribution of adjacent road sections; Generation of spatiotemporal coupling features: The features in the time and space dimensions are coupled to generate high-order features with spatiotemporal characteristics. The spatiotemporal density features are generated by combining the speed changes in time and the traffic flow distribution in space to describe the changes in traffic pressure in a certain area in different time periods. The spatiotemporal causal graph construction submodule is configured to abstract people, vehicles, roads and environmental elements into nodes of a graph network; each node contains not only static attributes but also dynamic attributes; the causal relationship between nodes is determined through causal reasoning, and the causal relationship is represented by directed edges and endowed with time delay attributes; The reasoning and hierarchical modeling submodule is configured to analyze the overall operation status of the traffic system at the global scale, analyze the cause-effect relationship in the environment at the local scale, and capture the driver's response to unexpected events at the instantaneous scale; The causal chain identification and optimization submodule is configured to extract key causal chains from the spatiotemporal causal graph; identify key nodes in the causal chain through reverse causal reasoning, and propose optimization strategies.

6. The traffic risk dynamic assessment model construction system according to claim 5 is characterized in that: Reasoning and hierarchical modeling submodule, including: The first association unit is configured to generate guidance suggestions for local areas through data analysis and model prediction; if the global data shows that the traffic pressure in a certain area exceeds the standard setting in a certain period of time, the local area will deploy traffic signal optimization and increase police force diversion measures in advance; The second association unit is configured as a mesoscopic level between the global and instantaneous levels. The data at the local scale is transmitted to the instantaneous scale in real time to predict the driver's instantaneous behavior. If the traffic volume at a certain intersection suddenly increases, it is predicted that the driver will choose to change lanes or slow down. The third association unit is configured to provide feedback to the global and local systems through real-time collection and transmission; the driver's instantaneous behavior data is used to update the global traffic pressure distribution map or optimize the signal light control strategy at the local intersection.

7. The traffic risk dynamic assessment model construction system according to claim 1 is characterized in that: Risk quantification module, including: The risk source identification submodule is configured to initialize the risk source as a node in the network and establish connections with other nodes through the causal network; dynamically expand the causal chain of risks to capture the chain reaction between multiple events; continuously update the causal chain through real-time data to capture the dynamic characteristics of risk propagation; The Bayesian reasoning submodule is configured to construct a priori probabilities for each risk source and causal chain. Through the input of real-time data, the Bayesian reasoning is used to dynamically update the posterior probability of risks. The macro level evaluates global risks, the meso level evaluates local risks, and the micro level evaluates instantaneous risks. The risk impact quantification analysis submodule is configured to capture the uncertainty and complexity of risk propagation through a large number of stochastic simulations; Generate risk propagation paths based on the results of causal chains and Bayesian reasoning; The macro level simulates global risk propagation, the meso level simulates local risk propagation, and the micro level simulates instantaneous risk propagation; at the same time, the scope of risk impact is quantified; The path and key node identification submodule is configured to dynamically identify the key paths of risk transmission based on the results of Bayesian reasoning and Monte Carlo sampling; locate key nodes by analyzing the risk propagation path; adjust the global traffic signal control strategy at the macro level, deploy local police force for guidance at the meso level, and send real-time warning information to drivers at the micro level.

8. The traffic risk dynamic assessment model construction system according to claim 7 is characterized in that: The path and key node identification submodule includes: A dynamic identification unit is configured to identify global risk transmission paths at the macro level, local risk transmission paths at the meso level, and instantaneous risk transmission paths at the micro level; A dynamic positioning unit is configured to position global key nodes at the macro level, local key nodes at the meso level, and instantaneous key nodes at the micro level; The quantitative output unit is configured to quantify potential risks through multi-level causal chain decomposition and dynamic probability propagation.

9. The traffic risk dynamic assessment model construction system according to claim 1 is characterized in that: Active intervention components, including: The dynamic stratification module is configured to classify the risks in the traffic network into three levels: low, medium and high, based on the real-time risk assessment results, and further refine them into three levels: region, road section and node. The risk level of each level is calculated by dynamic weighting of multi-dimensional data; The logic building module is configured to establish a multi-level strategy matching mechanism for different risk levels and layers; In low-risk areas, suggestive interventions are used; in medium-risk areas, preventive interventions are used; and in high-risk areas, emergency interventions are used; The strategy matching logic is based on the dynamic analysis of risk transmission paths; The dynamic update optimization module is configured to dynamically update the intervention strategy based on the real-time risk assessment results and traffic status data; and dynamically optimize the strategy generation logic by analyzing the implementation effect of the intervention strategy.

Images