A Road Network High-Traffic Emergency Response System and Method Based on Artificial Intelligence Large Model

Abstract

This invention relates to the field of intelligent traffic control technology, and discloses a road network high-traffic emergency response system and method based on an artificial intelligence large-scale model. The system includes a data reliability fusion module, a traffic flow prediction module, a safe passage decision module, a cross-departmental collaborative scheduling module, and a passage decision optimization module. It performs spatiotemporal coupling analysis on pre-acquired traffic state matrix and road topology map data to obtain a congestion probability estimate, then performs passage decision analysis to obtain an initial traffic restriction plan. Combining emergency resource location data, it allocates permissions to multiple departments, generates multi-departmental collaborative instructions, and performs decision error analysis to obtain correction parameters. Based on the correction parameters, it generates an optimized traffic restriction plan. This invention can improve the efficiency of handling road congestion events.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control technology, and in particular to an emergency response system and method for high traffic volume on road networks based on a large artificial intelligence model. Background Technology

[0002] With the acceleration of urbanization, traffic emergency response in high-traffic road networks faces severe challenges. Existing technologies suffer from low data fusion reliability. Traditional systems rely on a single data source and lack spatiotemporal alignment capabilities for multi-source heterogeneous data such as vehicle trajectories, weather, and GPS drift correction, leading to significant errors in traffic state matrix generation. Furthermore, predictions of sudden congestion are lagging, and prediction models based on static graph convolution cannot capture abrupt changes in traffic density. Additionally, there is an imbalance in decision-making objectives. Traffic restriction scheme optimization often employs single-objective functions, neglecting the quantification of safety risks, resulting in secondary congestion or safety hazards during peak hours. Moreover, cross-departmental collaboration is inefficient, emergency resource scheduling relies on manual authorization allocation, and departmental command conflicts are high. Therefore, improving data fusion reliability, increasing the accuracy of sudden congestion prediction, and enhancing cross-departmental collaboration efficiency have become urgent problems to be solved. Summary of the Invention

[0003] This invention provides a road network high-traffic emergency response system and method based on an artificial intelligence large model. Its main purpose is to address the problems of low efficiency in handling traffic congestion events and inaccurate and untimely prediction of congestion events.

[0004] To achieve the above objectives, this invention provides a road network high-traffic emergency response system based on an artificial intelligence large-scale model. The system comprises: a data reliability fusion module, a traffic flow prediction module, a safe passage decision module, a cross-departmental collaborative scheduling module, and a passage decision optimization module, wherein:

[0005] The data trust fusion module is used to acquire a traffic state matrix that includes pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data.

[0006] The traffic flow prediction module is used to perform spatiotemporal coupling analysis on road condition data based on the traffic state matrix and road topology map data to obtain the estimated congestion probability of the road network.

[0007] The safe passage decision module is used to perform passage decision analysis on the road condition data based on the congestion probability estimate in order to obtain the initial traffic restriction plan for the road network.

[0008] The cross-departmental collaborative scheduling module is used to allocate permissions to multiple departments for the initial traffic restriction plan and emergency resource location data, so as to obtain multi-departmental collaborative instructions for the road network.

[0009] The traffic decision optimization module is used to perform decision error analysis on the initial traffic restriction plan based on the multi-department collaborative instructions and the road condition data, so as to obtain the correction parameters of the initial traffic restriction plan, and optimize the initial traffic restriction plan based on the correction parameters to obtain the optimized traffic restriction plan for the road network.

[0010] In a preferred embodiment, the step of acquiring the traffic state matrix, which includes pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data, is specifically used for:

[0011] A sliding time window function is used to align the pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data in the time dimension to obtain the time-aligned data stream of the road network.

[0012] The time-aligned data stream is spatially aligned using a geographic raster mapping function to obtain the spatiotemporally aligned data stream of the road network.

[0013] The spatiotemporally aligned data stream is dynamically weighted and fused to generate a traffic state matrix.

[0014] In a preferred embodiment, the step of performing spatiotemporal coupling analysis on road condition data based on the traffic state matrix and road topology map data to obtain a congestion probability estimate of the road network is specifically used for:

[0015] A three-stage spatiotemporal coupling analysis, including spatiotemporal graph convolution, dynamic graph attention mechanism, and congestion probability prediction, is performed on the traffic data. Specifically, the three-stage spatiotemporal coupling analysis involves:

[0016] In the first stage, the traffic state matrix and road topology map data are extracted using a spatiotemporal graph convolutional network to obtain the spatiotemporal feature tensor of the road network.

[0017] In the second stage, the pre-acquired traffic density change rate is compared with the traffic density threshold. Roads with traffic density change rates and traffic density thresholds are identified as roads with sudden traffic flow. Attention weights are calculated for the topological edges associated with the roads with sudden traffic flow to obtain the dynamic edge weight matrix of the road network.

[0018] The third stage involves performing congestion probability analysis on the spatiotemporal feature tensor and the dynamic edge weight matrix to obtain a congestion probability estimate of the road network. The congestion probability analysis is as follows:

[0019] Topological constraints are applied to the spatiotemporal feature tensor and the dynamic edge weight matrix to obtain the spatiotemporal evolution state tensor of the road network;

[0020] The spatiotemporal evolution state tensor is mapped using a fully connected layer to obtain the original scalar value of the road network's congestion risk.

[0021] The original scalar value of the congestion risk is probabilistically compressed to obtain the estimated congestion probability of the road network.

[0022] In a preferred embodiment, the step of extracting spatiotemporal features from the traffic state matrix and road topology map data using a spatiotemporal graph convolutional network is specifically used for:

[0023] The standardized traffic state matrix is ​​fused with the road topology map data. The topology map is encoded into matrix form through graph Laplace transform to generate a normalized Laplace matrix. The normalized Laplace matrix is ​​then subjected to a three-dimensional convolution operation to obtain the spatiotemporal feature tensor of the road network. In the road topology map, nodes represent the entrances and exits of road sections, and edges represent the physical connection relationships of road segments. Each edge is attached with attributes such as length, slope, and number of lanes.

[0024] In a preferred embodiment, the step of calculating attention weights for the topological edges associated with the roads generated by the abrupt traffic flow to obtain the dynamic edge weight matrix of the road network is specifically used for:

[0025] Extract the spatiotemporal feature vectors of the road segments and their neighboring nodes generated by the abrupt traffic flow from the spatiotemporal feature tensor, and denote them as abrupt nodes;

[0026] The mutation nodes are subjected to similarity quantification processing to obtain the feature similarity of the mutation nodes, and the feature similarity is quantified to obtain the congestion risk association score of the road network;

[0027] The congestion risk association scores are normalized by edge weights to obtain the congestion propagation probability of the road network, and a dynamic edge weight matrix of the road network is constructed based on the congestion propagation probability.

[0028] In a preferred embodiment, the step of performing traffic decision analysis on the road condition data based on the congestion probability estimate to obtain the initial traffic restriction plan for the road network is specifically used for:

[0029] Based on the congestion probability estimate and the preset road safe traffic capacity threshold, a state space vector is constructed from the road condition data;

[0030] Substituting the state space vector into the bi-objective reward function yields the traffic efficiency reward and the safety risk reward of the road network, wherein the expression of the bi-objective reward function is as follows: In the formula, For the reward function, Incentives for improving traffic flow. As a reward for safety risks, It is a state-space vector. For limited actions;

[0031] The optimal strategy is solved based on the traffic efficiency reward function and the safety risk reward function to obtain the initial traffic restriction scheme for the road network. The calculation formula for the optimal strategy solution is as follows: In the formula, This is the initial traffic restriction plan. As candidate actions, For comparison of actions, To compare the action value functions of actions, Let the action value function be the candidate action. This is a Pareto dominance relationship. For comparison actions that are assumed to not exist, Let the action value function be... The final time step for solving the strategy. Number the time steps. For the current time step, For expected operators, As a discount factor, for Reward function at a given time point.

[0032] In a preferred embodiment, the step of allocating multi-department permissions to the initial traffic restriction plan and emergency resource location data to obtain multi-department collaborative instructions for the road network is specifically used for:

[0033] Based on the set of restricted area identifiers and the set of restricted actions in the initial traffic restriction scheme, a conflict scan is performed on the predefined multi-department permission rule table to generate the permission conflict matrix of the initial traffic restriction scheme.

[0034] Based on the permission conflict matrix and real-time location data of emergency resources, dynamic permission allocation is performed through blockchain smart contract functions to obtain the department-resource binding relationship of the initial traffic restriction plan;

[0035] Based on the department-resource binding relationship and the set of restricted actions, the instructions are encoded to prevent tampering through atomic encapsulation functions to obtain the multi-department collaborative instructions of the road network.

[0036] In a preferred embodiment, the step of performing decision error analysis on the initial traffic restriction plan based on the multi-department collaborative instructions and the traffic condition data to obtain the correction parameters of the initial traffic restriction plan is specifically used for:

[0037] Based on the departmental execution actions in the multi-departmental collaborative instructions and the expected actions in the initial traffic restriction plan, a difference analysis is performed using an action consistency function to obtain the instruction execution deviation matrix.

[0038] The traffic data and the congestion probability estimate are dynamically coupled using a state error function to obtain an environmental error vector.

[0039] Error fusion is performed on the instruction execution deviation matrix and the environmental error vector to obtain the correction parameter vector of the initial traffic restriction scheme.

[0040] In a preferred embodiment, when optimizing the initial traffic restriction scheme based on the correction parameters to obtain an optimized traffic restriction scheme for the road network, the specific steps are as follows:

[0041] Based on the correction parameters and the initial traffic restriction scheme, the set of traffic restriction actions is linearly adjusted to obtain the optimized action values ​​of the initial traffic restriction scheme;

[0042] Based on the optimized action value and the road safe traffic capacity threshold, the road condition data is filtered by dual thresholds to obtain a set of feasible solutions;

[0043] Based on real-time traffic density, the final solution is selected from the set of feasible solutions through density adaptation calculation to output the optimized traffic restriction solution.

[0044] To address the aforementioned problems, this invention also provides a method for handling high-traffic situations in road networks based on a large-scale artificial intelligence model, comprising:

[0045] S1, acquire a traffic state matrix containing pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data;

[0046] S2, Based on the traffic state matrix and road topology map data, perform spatiotemporal coupling analysis on the road condition data to obtain the estimated congestion probability of the road network;

[0047] S3, Based on the congestion probability estimate, perform traffic decision analysis on the road condition data to obtain the initial traffic restriction plan for the road network;

[0048] S4, perform multi-department permission allocation on the initial traffic restriction plan and emergency resource location data to obtain multi-department collaborative instructions for the road network;

[0049] S5. Based on the multi-department collaborative instructions and the road condition data, perform decision error analysis on the initial traffic restriction plan to obtain the correction parameters of the initial traffic restriction plan. Based on the correction parameters, optimize the initial traffic restriction plan to obtain the optimized traffic restriction plan for the road network.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. This invention achieves high-precision fusion of multi-source heterogeneous data by combining a sliding time window function with geographic raster mapping technology. This method can perform spatiotemporal dual alignment of vehicle trajectory, meteorological monitoring, and vehicle GPS drift correction data, effectively overcoming problems such as low reliability and error accumulation caused by inconsistent data sources and heterogeneous formats in traditional systems. Through a dynamic weighted fusion mechanism, the system further improves the accuracy and reliability of the traffic state matrix;

[0052] 2. This invention proposes a three-stage spatiotemporal coupling analysis method that integrates spatiotemporal graph convolution, dynamic graph attention mechanism, and congestion probability prediction. This method not only extracts spatiotemporal features from the road network but also identifies areas of sudden traffic flow changes through dynamic attention weights, thereby achieving early warning and high-precision prediction of sudden congestion. Compared with traditional methods relying on static graph structures, this invention possesses stronger real-time response capabilities and scenario adaptability, significantly improving the timeliness and accuracy of congestion prediction.

[0053] 3. This invention introduces a dynamic permission allocation mechanism based on blockchain smart contracts, which can automatically generate tamper-proof and traceable multi-department collaborative instructions based on the real-time location of emergency resources and departmental permission rules. This effectively solves the problems of instruction conflicts, response delays and unclear responsibilities in traditional manual dispatching, and greatly improves the execution efficiency of cross-departmental collaboration and the overall emergency response capability of the system.

[0054] Therefore, through a multi-level and multi-module systematic design, this invention achieves full-process optimization from data perception and predictive analysis to decision execution, comprehensively improving the intelligence level and operational efficiency of emergency response to high traffic volumes on the road network. Attached Figure Description

[0055] Figure 1 This invention provides a system architecture diagram of a road network high-traffic emergency response system based on an artificial intelligence large model, as an embodiment of the present invention.

[0056] Figure 2 This is a flowchart illustrating an emergency response method for high traffic volume on a road network based on a large artificial intelligence model, as provided in an embodiment of the present invention.

[0057] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0060] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0061] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0062] In practice, the server-side equipment deployed in a road network high-traffic emergency response system based on an AI-powered large-scale model may consist of one or more devices. This AI-powered large-scale model-based road network high-traffic emergency response system can be implemented as: a business instance, a virtual machine, and hardware devices. For example, this AI-powered large-scale model-based road network high-traffic emergency response system can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, this AI-powered large-scale model-based road network high-traffic emergency response system can be understood as software deployed on a cloud node, used to provide each user terminal with an AI-powered large-scale model-based road network high-traffic emergency response system. Alternatively, this AI-powered large-scale model-based road network high-traffic emergency response system can also be implemented as a virtual machine deployed on one or more devices in a cloud node. This virtual machine contains application software for managing each user terminal. Alternatively, this AI-powered large-scale model-based road network high-traffic emergency response system can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide each user terminal with an AI-powered large-scale model-based road network high-traffic emergency response system.

[0063] In terms of implementation, a road network high-traffic emergency response system based on an AI-powered large-scale model and the user terminal are mutually compatible. Specifically, if the AI-powered large-scale model is implemented as an application installed on a cloud service platform, the user terminal acts as a client establishing a communication connection with that application; or if the AI-powered large-scale model is implemented as a website, the user terminal acts as a webpage; or if the AI-powered large-scale model is implemented as a cloud service platform, the user terminal acts as a mini-program within an instant messaging application.

[0064] like Figure 1 The diagram shown is a system architecture diagram of a road network high-traffic emergency response system based on an artificial intelligence large model, provided by an embodiment of the present invention.

[0065] The road network high-traffic emergency response system 100 based on an artificial intelligence large-scale model described in this invention can be set up in a cloud server. In terms of implementation, it can be implemented as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed as a website. Depending on the functions implemented, the road network high-traffic emergency response system 100 based on an artificial intelligence large-scale model may include a data reliability fusion module 101, a traffic flow prediction module 102, a safe passage decision module 103, a cross-departmental collaborative scheduling module 104, and a passage decision optimization module 105. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0066] In this embodiment of the invention, in a road network high-traffic emergency response system based on an artificial intelligence large-scale model, each of the above-mentioned modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. The road network high-traffic emergency response system based on an artificial intelligence large-scale model provided by this embodiment of the invention allows for adjustment of the applicable scope of the system architecture without modifying the program code. This is achieved by adding modules and directly calling them, enabling cluster-based horizontal expansion and flexibly expanding the system. In practical applications, the above modules can be set in the same device or different devices, or in virtual devices, such as service instances in a cloud server.

[0067] The following describes, with reference to specific embodiments, each component and specific workflow of a road network high-traffic emergency response system based on a large artificial intelligence model:

[0068] The data trust fusion module is used to acquire a traffic state matrix that includes pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data.

[0069] In this embodiment of the invention, when the data trust fusion module executes the process of acquiring a traffic state matrix containing pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data, it is specifically used for:

[0070] A sliding time window function is used to align the pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data in the time dimension to obtain the time-aligned data stream of the road network.

[0071] The time-aligned data stream is spatially aligned using a geographic raster mapping function to obtain the spatiotemporally aligned data stream of the road network.

[0072] The spatiotemporally aligned data stream is dynamically weighted and fused to generate a traffic state matrix.

[0073] It should be noted that the specific process of the time dimension alignment is as follows: an exponential decay time window function with a 15-minute period is used for processing, and the time decay factor controls the decay rate of historical data weights. The larger the time decay factor value, the higher the weight of recent data.

[0074] It should be noted that the spatial dimension alignment is achieved by matching latitude and longitude coordinates to road network grid cells through a geographic raster mapping function, with each grid cell measuring 50m x 50m. Spatial aggregation weights are assigned to road network data falling within the same grid cell, generating a spatiotemporally aligned data stream.

[0075] It should be noted that the dynamic weighted fusion process is as follows: First, the ratio of the actual number of traffic congestion events in the historical traffic data to the total number of predicted traffic congestion events is calculated to obtain the historical error rate;

[0076] Based on the current time and the collection time, calculate the ratio of 1 minus the difference between the current time and the collection time to the collection period to obtain the data freshness.

[0077] The confidence weight is obtained by multiplying the reciprocal of the historical error rate and the data freshness by a corresponding adjustment coefficient and summing them.

[0078] The traffic state matrix is ​​obtained by weighted summation of the product of the confidence weight and the spatiotemporally aligned data stream.

[0079] The traffic flow prediction module is used to perform spatiotemporal coupling analysis on road condition data based on the traffic state matrix and road topology map data to obtain the estimated congestion probability of the road network.

[0080] In this embodiment of the invention, the step of performing spatiotemporal coupling analysis on road condition data based on the traffic state matrix and road topology map data to obtain the estimated congestion probability of the road network is specifically used for:

[0081] A three-stage spatiotemporal coupling analysis, including spatiotemporal graph convolution, dynamic graph attention mechanism, and congestion probability prediction, is performed on the traffic data. Specifically, the three-stage spatiotemporal coupling analysis involves:

[0082] In the first stage, the traffic state matrix and road topology map data are extracted using a spatiotemporal graph convolutional network to obtain the spatiotemporal feature tensor of the road network.

[0083] In the second stage, the pre-acquired traffic density change rate is compared with the traffic density threshold. Roads with traffic density change rates and traffic density thresholds are identified as roads with sudden traffic flow. Attention weights are calculated for the topological edges associated with the roads with sudden traffic flow to obtain the dynamic edge weight matrix of the road network.

[0084] The third stage involves performing congestion probability analysis on the spatiotemporal feature tensor and the dynamic edge weight matrix to obtain a congestion probability estimate of the road network. The congestion probability analysis is as follows:

[0085] Topological constraints are applied to the spatiotemporal feature tensor and the dynamic edge weight matrix to obtain the spatiotemporal evolution state tensor of the road network;

[0086] The spatiotemporal evolution state tensor is mapped using a fully connected layer to obtain the original scalar value of the road network's congestion risk.

[0087] The original scalar value of the congestion risk is probabilistically compressed to obtain the estimated congestion probability of the road network.

[0088] It should be noted that the traffic data includes: historical traffic flow, carbon monoxide concentration, and airflow disturbance characteristics.

[0089] It should be noted that the state space vector construction process is as follows: Based on the congestion probability estimate, real-time traffic density, and real-time carbon monoxide concentration, multi-dimensional vector fusion is performed on the monitoring data and prediction data of the same road number to obtain a three-dimensional decision state matrix that includes future risks, current load, and environmental safety. The multi-dimensional vector fusion refers to the process of spatially binding and dimensionally sorting the monitoring data and prediction data of the same road number based on the congestion probability estimate, real-time traffic density, and real-time carbon monoxide concentration.

[0090] It should be noted that the candidate action refers to the traffic restriction scheme to be evaluated, and the comparison action is an alternative traffic restriction scheme used to compare its merits with the candidate action.

[0091] It should be noted that the priority relationship of the Pareto dominance relationship is characterized as follows: This indicates that the comparison action function is superior to the candidate action function in both efficiency and safety objectives.

[0092] In this embodiment of the invention, the step of extracting spatiotemporal features from the traffic state matrix and road topology map data using a spatiotemporal graph convolutional network is specifically used for:

[0093] The standardized traffic state matrix is ​​fused with the road topology map data. The topology map is encoded into matrix form through graph Laplace transform to generate a normalized Laplace matrix. The normalized Laplace matrix is ​​then subjected to a three-dimensional convolution operation to obtain the spatiotemporal feature tensor of the road network. In the road topology map, nodes represent the entrances and exits of road sections, and edges represent the physical connection relationships of road segments. Each edge is attached with attributes such as length, slope, and number of lanes.

[0094] It should be noted that the first stage of the spatiotemporal coupling analysis, the spatiotemporal graph convolution stage, is to fuse the traffic state matrix and road topology map data, and encode the topology map into matrix form through graph Laplace transform to generate a normalized Laplace matrix.

[0095] Furthermore, the nodes in the road topology graph represent the entrances and exits of road segments, and the edges represent the physical connections between roads.

[0096] Furthermore, the process of the graph Laplace transform is as follows: An adjacency matrix A is constructed using the graph Laplace operator. If road segments i and j are directly connected, then... Otherwise, it is 0. Then, the degree matrix is ​​calculated using the diagonal elements as the number of node connections, and the Laplace matrix is ​​normalized.

[0097] Furthermore, by propagating features along the edges of the topological graph, the physical connection effects of road segment groups are captured, and spatiotemporal dimension features are extracted. By sliding along continuous time slices, the traffic flow evolution law is learned, and spatial dimension features are extracted. Finally, a spatiotemporal feature tensor is output, wherein the spatiotemporal feature tensor represents the dynamic evolution law of the road network.

[0098] In this embodiment of the invention, the step of calculating attention weights for the topological edges associated with the roads generated by the abrupt traffic flow to obtain the dynamic edge weight matrix of the road network is specifically used for:

[0099] Extract the spatiotemporal feature vectors of the road segments and their neighboring nodes generated by the abrupt traffic flow from the spatiotemporal feature tensor, and denote them as abrupt nodes;

[0100] The mutation nodes are subjected to similarity quantification processing to obtain the feature similarity of the mutation nodes, and the feature similarity is quantified to obtain the congestion risk association score of the road network;

[0101] The congestion risk association scores are normalized by edge weights to obtain the congestion propagation probability of the road network, and a dynamic edge weight matrix of the road network is constructed based on the congestion propagation probability.

[0102] It should be noted that the principle of the dynamic graph attention mechanism in the second stage of the spatiotemporal coupling analysis is to automatically adjust the weight of the tunnel connection edge according to the rate of change of traffic flow density, and to enhance the attention of the critical path during abrupt changes.

[0103] Furthermore, the traffic flow density change rate is obtained by calculating the ratio of the difference in traffic flow density between two collection time points to the corresponding time interval between the collection time points.

[0104] Furthermore, the feature similarity in the dynamic graph attention mechanism is obtained by concatenating vectors after weight matrix transformation.

[0105] Furthermore, while traditional graph convolution relies on a fixed topology, the dynamic graph attention mechanism triggers real-time weight updates through the rate of change in traffic density, enabling the model to adapt to sudden congestion scenarios, such as local traffic stagnation caused by accidents.

[0106] It should be noted that the congestion risk association score represents the intensity of congestion risk propagation between two nodes following the direction of traffic flow.

[0107] It should be noted that the congestion propagation probability is a dynamic edge weight scalar, and the higher the congestion propagation probability value, the higher the risk of congestion propagation.

[0108] It should be noted that the construction method of the dynamic edge weight matrix is ​​as follows: First, initialize the global matrix, construct a probability distribution based on the mutation nodes, inject the probability distribution locally into the initialized global matrix, mark the identified mutation nodes, and leave the non-mutation nodes unchanged, so as to obtain the dynamic edge weight matrix, wherein the dynamic edge weight matrix is ​​a set of dynamically changing weights.

[0109] The safe passage decision module is used to perform passage decision analysis on the road condition data based on the congestion probability estimate in order to obtain the initial traffic restriction plan for the road network.

[0110] In this embodiment of the invention, the step of performing traffic decision analysis on the road condition data based on the congestion probability estimate to obtain the initial traffic restriction scheme for the road network is specifically used for:

[0111] Based on the congestion probability estimate and the preset road safe traffic capacity threshold, a state space vector is constructed from the road condition data;

[0112] Substituting the state space vector into the bi-objective reward function yields the traffic efficiency reward and the safety risk reward of the road network, wherein the expression of the bi-objective reward function is as follows: In the formula, For the reward function, Incentives for improving traffic flow. As a reward for safety risks, It is a state-space vector. For limited actions;

[0113] The optimal strategy is solved based on the traffic efficiency reward function and the safety risk reward function to obtain the initial traffic restriction scheme for the road network. The calculation formula for the optimal strategy solution is as follows: In the formula, This is the initial traffic restriction plan. As candidate actions, For comparison of actions, To compare the action value functions of actions, Let the action value function be the candidate action. This is a Pareto dominance relationship. For comparison actions that are assumed to not exist, Let the action value function be... The final time step for solving the strategy. Number the time steps. For the current time step, For expected operators, As a discount factor, for Reward function at a given time point.

[0114] It should be noted that the topological constraints are based on the spatiotemporal feature tensor and dynamic edge weight matrix to correct the relationships between nodes in the road topology graph data. The direct connected nodes in the road topology graph are traversed. If two nodes are directly connected, the product of the propagation probability value in the dynamic edge weight matrix and the spatiotemporal feature tensor of the previous node is calculated, which represents the degree to which the state of the next node is affected by the previous node, and is denoted as the influence quantity. For road nodes that are not directly connected, their spatiotemporal feature tensors are kept unchanged, and the corrected set of road states is output as the spatiotemporal evolution state tensor.

[0115] Furthermore, the spatiotemporal evolution state tensor is used to simulate the physical propagation process of congestion in the road network, ensuring that the prediction conforms to the actual road connectivity.

[0116] It should be noted that the fully connected layer mapping refers to assigning weight parameters to the spatiotemporal feature tensors of each road node and performing weighted summation to obtain the original risk value, adding a preset bias constant to the original risk value, and generating the original scalar of the congestion risk value.

[0117] Furthermore, the weights assigned to the spatiotemporal feature tensor are obtained based on training with historical data, with higher weights for important nodes. The preset bias constant is used to calibrate the basic risk level, and the original scalar of the congestion risk value is used to aggregate the dispersed road states into a global risk indicator, providing input for probability transformation.

[0118] It should be noted that the probabilistic compression is a valuation that converts the original value of congestion risk into a probabilistic form, obtained by calculating the reciprocal of the original scalar power of the negative congestion risk value of the natural constant e.

[0119] The cross-departmental collaborative scheduling module is used to allocate permissions to multiple departments for the initial traffic restriction plan and emergency resource location data, so as to obtain multi-departmental collaborative instructions for the road network.

[0120] In this embodiment of the invention, the step of allocating multi-department permissions to the initial traffic restriction plan and emergency resource location data to obtain multi-department collaborative instructions for the road network is specifically used for:

[0121] Based on the set of restricted area identifiers and the set of restricted actions in the initial traffic restriction scheme, a conflict scan is performed on the predefined multi-department permission rule table to generate the permission conflict matrix of the initial traffic restriction scheme.

[0122] Based on the permission conflict matrix and real-time location data of emergency resources, dynamic permission allocation is performed through blockchain smart contract functions to obtain the department-resource binding relationship of the initial traffic restriction plan;

[0123] Based on the department-resource binding relationship and the set of restricted actions, the instructions are encoded to prevent tampering through atomic encapsulation functions to obtain the multi-department collaborative instructions of the road network.

[0124] It should be noted that the permission conflict scan is based on the set of restricted area identifiers and restricted action sets in the initial access restriction scheme, and performs conflict detection on the predefined multi-department permission rule table.

[0125] The blockchain smart contract dynamic permission allocation is based on the conflict matrix and emergency resource location. The optimal department is allocated through the smart contract. The specific process is as follows: First, the reciprocal of the conflict flag is calculated to obtain the permission compliance score. The minimum value of the conflict flag is 0 when there is no conflict, and the corresponding permission compliance score is 100. Otherwise, the minimum is 0.01.

[0126] Then, the negative linear distance of e is multiplied by 0.2 to obtain the distance attenuation fraction;

[0127] Finally, the ratio of current resource usage to total resource usage is calculated to obtain the resource load score.

[0128] Multiply the permission compliance score, the distance attenuation score, and the resource load score to obtain the department's total score, and bind the department with the highest total score to the corresponding region with the highest score.

[0129] The traffic decision optimization module is used to perform decision error analysis on the initial traffic restriction plan based on the multi-department collaborative instructions and the road condition data, so as to obtain the correction parameters of the initial traffic restriction plan, and optimize the initial traffic restriction plan based on the correction parameters to obtain the optimized traffic restriction plan for the road network.

[0130] In this embodiment of the invention, the step of performing decision error analysis on the initial traffic restriction plan based on the multi-department collaborative instructions and the road condition data to obtain the correction parameters of the initial traffic restriction plan is specifically used for:

[0131] Based on the departmental execution actions in the multi-departmental collaborative instructions and the expected actions in the initial traffic restriction plan, a difference analysis is performed using an action consistency function to obtain the instruction execution deviation matrix.

[0132] The traffic data and the congestion probability estimate are dynamically coupled using a state error function to obtain an environmental error vector.

[0133] Error fusion is performed on the instruction execution deviation matrix and the environmental error vector to obtain the correction parameter vector of the initial traffic restriction scheme.

[0134] It should be noted that the difference analysis is based on the departmental execution actions in the multi-departmental collaborative instructions and the expected actions in the initial traffic restriction plan. The execution deviation is quantified, the expected actions and actual execution actions in the same traffic restriction area are extracted, the deviation rate is calculated, and a deviation rate list is generated based on the deviation rate to form a deviation matrix.

[0135] The dynamic coupling refers to calculating the environmental prediction error based on the difference between real-time traffic conditions and predicted values ​​through weighted error calculation.

[0136] Error fusion refers to the comprehensive weighting of execution deviation and environmental error to generate correction parameters.

[0137] In this embodiment of the invention, when optimizing the initial traffic restriction scheme based on the correction parameters to obtain an optimized traffic restriction scheme for the road network, the specific steps are as follows:

[0138] Based on the correction parameters and the initial traffic restriction scheme, the set of traffic restriction actions is linearly adjusted to obtain the optimized action values ​​of the initial traffic restriction scheme;

[0139] Based on the optimized action value and the road safe traffic capacity threshold, the road condition data is filtered by dual thresholds to obtain a set of feasible solutions;

[0140] Based on real-time traffic density, the final solution is selected from the set of feasible solutions through density adaptation calculation to output the optimized traffic restriction solution.

[0141] It should be noted that the linear adjustment is based on the correction parameter and the action value in the initial traffic restriction scheme. The product of the correction parameter and the adjustment coefficient is calculated to obtain the adjustment amount. The adjustment amount is added to the original action value to optimize the action value.

[0142] It should be noted that the dual-threshold filtering is based on optimizing action values ​​and road safety thresholds to screen feasible solutions, and the solution selection is based on preset safety thresholds and preset efficiency thresholds.

[0143] It should be noted that the density adaptation calculation process is as follows: First, the theoretical density of the scheme is obtained by calculating an optimization action value of 0.6 times. Then, the density deviation is obtained by subtracting the theoretical density of the scheme from the real-time density. Finally, the scheme with the smallest density deviation is selected.

[0144] Reference Figure 2 The diagram shown is a flowchart illustrating a method for handling high traffic volume on a road network based on a large-scale artificial intelligence model, according to an embodiment of the present invention. In this embodiment, the method includes:

[0145] S1, acquire a traffic state matrix containing pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle GPS drift correction data;

[0146] S2, Based on the traffic state matrix and road topology map data, perform spatiotemporal coupling analysis on the road condition data to obtain the estimated congestion probability of the road network;

[0147] S3, Based on the congestion probability estimate, perform traffic decision analysis on the road condition data to obtain the initial traffic restriction plan for the road network;

[0148] S4, perform multi-department permission allocation on the initial traffic restriction plan and emergency resource location data to obtain multi-department collaborative instructions for the road network;

[0149] S5. Based on the multi-department collaborative instructions and the road condition data, perform decision error analysis on the initial traffic restriction plan to obtain the correction parameters of the initial traffic restriction plan. Based on the correction parameters, optimize the initial traffic restriction plan to obtain the optimized traffic restriction plan for the road network.

[0150] This invention achieves high-precision fusion of multi-source heterogeneous data by combining a sliding time window function with geographic raster mapping technology. This method can perform spatiotemporal dual alignment of vehicle trajectory, meteorological monitoring, and vehicle-mounted GPS drift correction data, effectively overcoming problems such as low reliability and error accumulation caused by inconsistent data sources and heterogeneous formats in traditional systems. Through a dynamic weighted fusion mechanism, it further improves the accuracy and reliability of the traffic state matrix.

[0151] This invention proposes a three-stage spatiotemporal coupling analysis method that integrates spatiotemporal graph convolution, dynamic graph attention mechanism, and congestion probability prediction. This method not only extracts spatiotemporal features from the road network but also identifies areas of sudden traffic flow changes through dynamic attention weights, thereby achieving early warning and high-precision prediction of sudden congestion. Compared with traditional methods relying on static graph structures, this method possesses stronger real-time response capabilities and scenario adaptability, significantly improving the timeliness and accuracy of congestion prediction.

[0152] This invention introduces a dynamic permission allocation mechanism based on blockchain smart contracts. It can automatically generate tamper-proof, traceable multi-department collaborative instructions based on real-time emergency resource locations and departmental permission rules. This effectively solves problems such as instruction conflicts, response delays, and unclear responsibilities inherent in traditional manual dispatching, significantly improving the execution efficiency of cross-departmental collaboration and the overall emergency response capability of the system. Therefore, through a multi-layered, multi-module systematic design, this invention achieves end-to-end optimization from data perception and predictive analysis to decision execution, comprehensively enhancing the intelligence level and operational efficiency of high-traffic emergency response in road networks.

[0153] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0154] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An artificial intelligence large model-based road network large flow emergency disposal system, characterized in that, The system comprises a data credible fusion module, a traffic flow prediction module, a safe passing decision module, a cross-department collaborative scheduling module and a passing decision optimization module, wherein: The data credible fusion module is configured to obtain a traffic state matrix comprising pre-acquired vehicle trajectory data, meteorological monitoring data and vehicle-mounted GPS drift correction data; The traffic flow prediction module is configured to perform spatio-temporal coupling analysis on road condition data based on the traffic state matrix and road topology graph data to obtain congestion probability estimates of the road network; When performing spatio-temporal coupling analysis on road condition data based on the traffic state matrix and road topology graph data to obtain congestion probability estimates of the road network, the traffic flow prediction module is specifically configured to: perform three-stage spatio-temporal coupling analysis on the road condition data comprising spatio-temporal graph convolution, dynamic graph attention mechanism and congestion probability prediction, wherein the three-stage spatio-temporal coupling analysis specifically comprises: In the first stage, spatio-temporal feature extraction is performed on the traffic state matrix and road topology graph data by a spatio-temporal graph convolution network to obtain a spatio-temporal feature tensor of the road network; In the second stage, the pre-acquired vehicle flow density change rate is compared with the vehicle flow density threshold, and the road on which the vehicle flow density change rate is greater than the vehicle flow density threshold is determined as a road on which a sudden vehicle flow occurs, and attention weight calculation is performed on the topology edges associated with the road on which the sudden vehicle flow occurs to obtain a dynamic edge weight matrix of the road network; In the third stage, congestion probability analysis is performed on the spatio-temporal feature tensor and the dynamic edge weight matrix to obtain congestion probability estimates of the road network, wherein the congestion probability analysis comprises: topology constraint on the spatio-temporal feature tensor and the dynamic edge weight matrix to obtain a spatio-temporal evolution state tensor of the road network; full connection layer mapping on the spatio-temporal evolution state tensor to obtain a congestion risk value original scalar of the road network; probability compression on the congestion risk value original scalar to obtain congestion probability estimates of the road network; The safe passing decision module is configured to perform passing decision analysis on the road condition data based on the congestion probability estimates to obtain an initial traffic restriction scheme of the road network; The cross-department collaborative scheduling module is configured to perform multi-department authority allocation on the initial traffic restriction scheme and emergency resource location data to obtain multi-department collaborative instructions of the road network; The passing decision optimization module is configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instructions and the road condition data to obtain correction parameters of the initial traffic restriction scheme, and optimize the initial traffic restriction scheme based on the correction parameters to obtain an optimized traffic restriction scheme of the road network.

2. The road network large flow emergency disposal system based on an artificial intelligence large model according to claim 1, wherein, When performing spatio-temporal coupling analysis on road condition data based on the traffic state matrix and road topology graph data to obtain congestion probability estimates of the road network, the data credible fusion module is specifically configured to: align the pre-acquired vehicle trajectory data, meteorological monitoring data and vehicle-mounted GPS drift correction data in the time dimension by using a sliding time window function to obtain time-aligned data streams of the road network; aligning the time-aligned data stream in a spatial dimension by a geographic grid mapping function to obtain a space-time aligned data stream of the road network; performing dynamic weighted fusion on the space-time aligned data stream to generate a traffic state matrix.

3. The road network large flow emergency disposal system based on artificial intelligence large model of claim 1, wherein, The space-time feature extraction on the traffic state matrix and the road topology graph data by the space-time graph convolution network is specifically used for: fusing the normalized traffic state matrix and the road topology graph data, encoding the topology graph into a matrix form through graph Laplace transformation to generate a normalized Laplace matrix, and performing three-dimensional convolution operation on the normalized Laplace matrix to obtain a space-time feature tensor of the road network, wherein the road topology graph represents road section entrances and exits as nodes and represents road segment physical connection relationships as edges, and each edge is attached with length, slope, and lane number attributes.

4. The road network large flow emergency disposal system based on an artificial intelligence large model of claim 1, wherein, The attention weight calculation on the topology edges related to the road of the mutant traffic flow is specifically used for: extracting the space-time feature vectors of the nodes and neighbor nodes of the road segments of the mutant traffic flow from the space-time feature tensor, denoted as mutant nodes; performing similarity quantization processing on the mutant nodes to obtain a mutant node feature similarity, and quantizing the feature similarity to obtain a congestion risk correlation score of the road network; performing edge weight normalization processing on the congestion risk correlation score to obtain a congestion propagation probability of the road network, and constructing a dynamic edge weight matrix of the road network based on the congestion propagation probability.

5. The road network large flow emergency disposal system based on an artificial intelligence large model of claim 1, wherein, The safe passage decision module, when performing passage decision analysis on the road condition data based on the congestion probability estimate to obtain an initial traffic restriction scheme of the road network, is specifically used for: constructing a state space vector based on the congestion probability estimate and a preset road safe passage capacity threshold based on the congestion probability estimate and a preset road safe passage capacity threshold; substituting the state space vector into a double-objective reward function to obtain a passage efficiency reward of the road network and a safety risk reward of the road network, wherein the expression of the double-objective reward function is as follows: wherein, is a reward function, is a passing efficiency reward, is a safety risk reward, is a state space vector, is a limiting action; solving an optimal strategy of the road condition data based on the function of the passage efficiency reward and the function of the safety risk reward to obtain an initial traffic restriction scheme of the road network, wherein the calculation formula of the optimal strategy solving is as follows: where, is the initial routing scheme, is a candidate action, is a contrast action, is the action value function for the contrast action, is the action value function for the candidate action, is the Pareto dominance relation, is the contrast action assuming non-existence, is the action value function, is the termination time step for the policy solution, is the time step number, is the current time step, is the expectation operator symbol, is the discount factor, is the reward function at the time point.

6. The road network large flow emergency disposal system based on an artificial intelligence large model of claim 5, wherein, The cross-department collaborative scheduling module, when performing multi-department authority allocation on the initial traffic restriction scheme and emergency resource location data to obtain a multi-department collaborative instruction of the road network, is specifically used for: performing conflict scanning on a pre-defined multi-department authority rule table based on a traffic restriction area identifier set and a traffic restriction action set in the initial traffic restriction scheme to generate an authority conflict matrix of the initial traffic restriction scheme; performing dynamic authority allocation through a blockchain smart contract function based on the authority conflict matrix and real-time emergency resource location data to obtain a department-resource binding relationship of the initial traffic restriction scheme; performing instruction tamper-proof encoding through an atomic encapsulation function based on the department-resource binding relationship and the traffic restriction action set to obtain a multi-department collaborative instruction of the road network.

7. The road network large flow emergency disposal system based on artificial intelligence large model of claim 6, wherein, The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis.

8. The road network large-flow emergency disposal system based on artificial intelligence large model of claim 7, wherein, The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The method comprises: S1, obtaining a traffic state matrix containing pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle-mounted GPS drift correction data; S2, performing spatio-temporal coupling analysis on road condition data based on the traffic state matrix and road topology graph data to obtain a congestion probability estimate of the road network; 9. A road network large flow emergency disposal method based on an artificial intelligence large model, characterized in that, The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The method comprises: S1, obtaining a traffic state matrix containing pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle-mounted GPS drift correction data; S2, performing spatio-temporal coupling analysis on road condition data based on the traffic state matrix and road topology graph data to obtain a congestion probability estimate of the road network; The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The method comprises: S1, obtaining a traffic state matrix containing pre-acquired vehicle trajectory data, meteorological monitoring data, and vehicle-mounted GPS drift correction data; S2, performing spatio-temporal coupling analysis on road condition data based on the traffic state matrix and road topology graph data to obtain a congestion probability estimate of the road network; The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. The traffic decision optimization module is specifically configured to perform decision error analysis on the initial traffic restriction scheme based on the multi-department collaborative instruction and the road condition data to obtain a correction parameter of the initial traffic restriction scheme when performing the decision error analysis. S3, performing traffic decision analysis on the traffic data based on the congestion probability estimation to obtain an initial traffic restriction scheme of the road network; S4, performing multi-department authority allocation on the initial traffic restriction scheme and emergency resource location data to obtain a multi-department coordination instruction of the road network; S5, performing decision error analysis on the initial traffic restriction scheme based on the multi-department coordination instruction and the traffic data to obtain a correction parameter of the initial traffic restriction scheme, and performing optimization on the initial traffic restriction scheme based on the correction parameter to obtain an optimized traffic restriction scheme of the road network.

Images