Intelligent traffic evacuation method based on dynamic path optimization and multi-objective cooperation

Abstract

The invention discloses an intelligent traffic evacuation method based on dynamic path optimization and multi-objective cooperation, and relates to the technical field of intelligent traffic and emergency management, and the method comprises the steps: building an improved HD-PSO model based on a traffic violation risk space-time distribution diagram, calculating the risk index of a path node through a dynamic congestion impact factor matrix, and calculating the risk index of the path node; a four-layer dynamic early warning mechanism is constructed, and a real-time risk index and a hierarchical response instruction are output; on the basis of the optimized evacuation path weight parameters and resource allocation factors and in combination with a real-time environment state, through a DRL-LSTM hybrid model, predicting people flow density and dynamically updating the path weight parameters, and generating an optimized evacuation strategy; the improved HD-PSO model is used for outputting the Pareto optimal solution, the evacuation time is compressed through the fitness function, the accident rate is reduced, and the resource utilization rate is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation and emergency management, and in particular to an intelligent transportation evacuation method based on dynamic path optimization and multi-objective collaboration. Background Art

[0002] As an important part of urban transportation, the rapid growth in the number of electric bicycles has brought new challenges to traffic management. Illegal behaviors of electric bicycles, such as driving against traffic, running red lights, and occupying lanes, not only disrupt normal traffic order, but also significantly increase the risk of traffic accidents. According to statistics, the number of traffic accidents involving electric bicycles has increased year by year in recent years, especially during peak hours in the morning and evening and on complex roads, where congestion and safety problems caused by illegal behaviors are particularly prominent. Traditional traffic management methods have obvious limitations when dealing with dynamic illegal behaviors of electric bicycles, making it difficult to achieve accurate real-time monitoring and efficient management. Therefore, there is an urgent need for an intelligent traffic management method that can dynamically assess risks, optimize speed limit strategies, and coordinate the regulation of spatiotemporal resources to improve the overall safety and efficiency of urban transportation.

[0003] Existing evacuation methods have the following technical shortcomings: Real-time changes are difficult to detect: Traditional models (such as the classic Dijkstra algorithm) do not take into account real-time changes in passenger flow, resulting in increased path congestion; Single-objective optimization: Existing methods mostly focus on the shortest path, ignore the coordinated needs of safety and resource allocation, and do not clearly explain how to effectively balance the conflicts between multiple objectives such as safety and efficiency. They focus on evacuation path guidance and pay less attention to the integration and coordination of emergency resource scheduling; Failure in extreme scenarios: In scenarios such as fire and earthquake, the model does not embed a dynamic adjustment mechanism, and the evacuation success rate is less than 70%. At the same time, it has limitations such as high computational complexity and poor scenario adaptability. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration to solve the multi-objective collaborative optimization problem of evacuation efficiency and safety in extreme scenarios.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration, which includes:

[0008] Collect real-time pedestrian density data and electric bicycle traffic data, clean and standardize them, and construct a spatiotemporal distribution map of traffic violation risks;

[0009] Based on the spatiotemporal distribution map of traffic violation risks, an improved HD-PSO model is established. The risk index of path nodes is calculated using a dynamic congestion impact factor matrix. A four-layer dynamic early warning mechanism is constructed to output real-time risk indexes and hierarchical response instructions.

[0010] Based on the real-time risk index and hierarchical response instructions, the NSGA-III algorithm is used to optimize the evacuation path weight parameters and resource allocation factors. The constraints are applied to solve the multi-objective optimal solution and the optimized evacuation path weight parameters and resource allocation factors are output.

[0011] Based on the optimized evacuation path weight parameters and resource allocation factors, and combined with real-time environmental conditions, the DRL-LSTM hybrid model predicts crowd density and dynamically updates the path weight parameters to generate an optimized evacuation strategy.

[0012] Execute optimized evacuation strategies, implement evacuation guidance and emergency response in a real-time environment, verify evacuation results through the digital twin platform, and output evacuation execution reports.

[0013] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration described in the present invention, the electric bicycle traffic data includes the proportion of going against traffic, the proportion of running red lights, the proportion of occupying lanes, the number of traffic accidents, traffic flow and speed limit policy enforcement intensity parameters.

[0014] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration described in the present invention, the construction of the spatiotemporal distribution map of traffic violation risks includes the following steps:

[0015] Obtain pedestrian density data and electric bicycle traffic data, clean outliers, fill in missing fields, supplement historical parameters, and perform normalized mapping;

[0016] Construct an urban violation intensity index and integrate the mapped data to generate a spatiotemporal fusion dataset. Through time period density analysis, high-conflict area labeling and violation ratio aggregation, the spatiotemporal distribution map of traffic violation risks is output.

[0017] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration of the present invention, the output of real-time risk index and graded response instructions includes the following steps:

[0018] Analyze the spatiotemporal distribution map of traffic violation risks, extract the coordinates and density ratio data of high-conflict areas, and generate a gridded data table after gridding transformation and principal component analysis dimensionality reduction;

[0019] Based on the violation ratio and real-time normal intervention intensity in the grid data table, the violation ratio of grid units is calculated. Combined with the traffic infrastructure data and real-time traffic flow, the conflict probability value is calculated for each grid.

[0020] An improved HD-PSO model was established to calculate the path congestion risk index based on the path node parameters in historical evacuation event data. The conflict probability values ​​were associated to generate a dynamic congestion impact factor matrix, and the dynamic risk index was calculated based on real-time traffic flow.

[0021] A comprehensive node risk table is generated based on the path congestion risk index and the dynamic risk index, and the four-layer warning rules are applied to mark the risk status and trigger corresponding response instructions.

[0022] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration described in the present invention, the risk status includes low risk, medium risk, high risk and critical risk.

[0023] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration of the present invention, wherein: the output of the optimized evacuation path weight parameters and resource allocation factors includes the following steps:

[0024] Filter high-risk nodes from the warning status table, associate hierarchical response instructions, generate a high-risk node historical data table, initialize the evacuation path weight parameters and resource allocation factor optimization parameters of the NSGA-III algorithm, and define a multi-objective optimization function;

[0025] According to the defined multi-objective optimization function and initialization parameters, NSGA-III optimization is performed to output the optimized evacuation path weight parameters and resource allocation factors.

[0026] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration of the present invention, wherein: the generating of the optimized evacuation strategy includes the following steps:

[0027] Construct an environmental state vector, apply the LSTM gating mechanism to output the crowd density prediction value, combine it with the actual number of accidents, and use DRL to update the optimized evacuation path weight parameters;

[0028] Based on the updated path weight parameters, combined with the safety factor and efficiency coefficient, the priority weight of each path is calculated, and the resource allocation factor and the predicted value of crowd density are integrated to generate an optimized evacuation strategy.

[0029] As a preferred solution of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration of the present invention, the output of the evacuation execution report includes the following steps:

[0030] Execute optimized evacuation strategies and resource allocation plans, combine with the warning status table, continuously update path weight parameters, and dynamically optimize evacuation strategies;

[0031] On the digital twin platform, a virtual scene is reconstructed based on the real-time environmental status and updated evacuation strategy parameters. The evacuation time interval and accident rate distribution are output through Monte Carlo simulation. The effectiveness of the strategy is verified by combining the path overlap rate, and an evacuation execution report is generated.

[0032] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration as described in the first aspect of the present invention is implemented.

[0033] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration as described in the first aspect of the present invention.

[0034] The beneficial effects of the present invention are as follows: based on the conflict probability matrix, the spatiotemporal distribution of behaviors such as driving against traffic and running red lights is integrated in real time for dynamic violation ratios, thereby reducing the risk identification delay; combining the path congestion index with the dynamic risk index to trigger a graded response; using the improved HD-PSO model to output the Pareto optimal solution, and using the fitness function to compress evacuation time, reduce the accident rate, and improve resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 This is a flow chart of an intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration.

[0037] Figure 2 This is the spatiotemporal distribution map of traffic violation risks in Beijing.

[0038] Figure 3 This is a radar chart showing the proportion of electric bicycle traffic violations from 2020 to 2024.

[0039] Figure 4 This is a trend chart showing the number of traffic accidents and the proportion of violations from 2020 to 2024.

[0040] Figure 5 This is a multi-dimensional analysis chart of the speed limit strategy effect.

[0041] Figure 6 The table shows the impact forecast data of key parameters across the country.

[0042] Figure 7 This is a table showing the optimal strategy effects for different city types.

[0043] Figure 8 This is a table showing the predicted data for traffic efficiency and speed limit strategy coverage from 2025 to 2030.

[0044] Figure 9 This is a table of national predictive data for 2025-2030.

[0045] Figure 10 This is a table comparing the forecast data for representative cities from 2025 to 2030. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0049] Reference Figures 1 to 10 , is an embodiment of the present invention, which provides an intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration, including the following steps:

[0050] S1. Collect real-time pedestrian density data and electric bicycle traffic data, clean and standardize them, and construct a spatiotemporal distribution map of traffic violation risks.

[0051] S1.1. Obtain pedestrian density data and electric bicycle traffic data, clean outliers, fill in missing fields, supplement historical parameters, and perform normalized mapping.

[0052] Specifically, the system connects to the monitoring systems of large shopping malls and transportation hubs to obtain heat maps of the instantaneous number of people and channel density distribution in different areas. At the same time, it collects e-bike traffic data, including the proportion of people driving against traffic, the proportion of people running red lights, the proportion of people occupying the road, the number of traffic accidents, traffic flow, and the intensity of speed limit policy enforcement.

[0053] The obtained data on the number of instantaneous pedestrian flows and electric bicycle traffic in different regions are cleaned, including deleting abnormal fluctuations in the number of instantaneous pedestrian flows in different regions; filling in the missing fields in the electric bicycle traffic data, supplementing the historically unrecorded width parameters of the evacuation channels of transportation hubs and the bottleneck identifiers of the paths recorded after the end of stadium events; and using the Min-Max normalization method to map the cleaned data on the number of instantaneous pedestrian flows and electric bicycle traffic in different regions to the interval [0, 1].

[0054] S1.2. Construct an urban violation intensity index and integrate the mapped data to generate a spatiotemporal fusion dataset. Through time period density analysis, high-conflict area annotation, and violation ratio aggregation, output a spatiotemporal distribution map of traffic violation risk.

[0055] Specifically, we construct a city violation intensity index and calculate the violation intensity value of each city. The expression is:

[0056] ;

[0057] in, is the violation intensity value of each city;

[0058] The instantaneous number of pedestrians in each region, electric bicycle traffic data, and violation intensity values ​​of each city are integrated to generate a triple structured fusion dataset containing timestamp, location coordinates, and violation intensity value;

[0059] Parse the triplet sequence of timestamp, location coordinates, and violation intensity value in the fused dataset, and identify the average density value of each monitoring point during the five time periods: early morning, morning peak, noon, evening peak, and night;

[0060] Retrieve the path bottleneck identifiers from historical evacuation event characteristics, mark the coordinate points of three high-conflict areas: stadiums, subway transfer stations, and train station exits in the geographic coordinate system, and generate a high-conflict coordinate marking table;

[0061] Load the electric bicycle violation data and aggregate the spatiotemporal distribution means of the wrong-way ratio, red light running ratio, and lane-occupying ratio by grid unit.

[0062] Overlay the average density value, high-conflict coordinate marker table, and spatiotemporal distribution mean to output a spatiotemporal distribution map of traffic violation risks covering both crowd flow hotspots and violation hotspots (see Figure 2 ).

[0063] S2. Based on the spatiotemporal distribution map of traffic violation risks, an improved HD-PSO model is established. The risk index of path nodes is calculated through the dynamic congestion impact factor matrix, and a four-layer dynamic early warning mechanism is constructed to output real-time risk index and hierarchical response instructions.

[0064] S2.1. Analyze the spatiotemporal distribution map of traffic violation risks, extract the coordinates and density ratio data of high-conflict areas, and generate a gridded data table after gridding transformation and principal component analysis dimensionality reduction.

[0065] Specifically, the file data of the spatiotemporal distribution map of traffic violation risks is parsed, the legend mark positions of the three high-conflict areas of "stadium", "subway transfer station" and "train station exit" in the map are located, and the corresponding geographic coordinate point sets are extracted; based on the color mapping data of the spatiotemporal distribution map of traffic violation risks, the corresponding crowd density value interval and the corresponding violation behavior ratio interval are extracted, the spatiotemporal distribution map of traffic violation risks is divided into geographic grids, the average color mapping value of each grid cell is calculated, and converted into a specific value, and a gridded data table is generated, the fields of which include grid ID, center point latitude and longitude, crowd density and violation ratio value; the violation ratio value is generated based on the key features selected by principal component analysis, eigenvalue decomposition is performed to calculate the covariance matrix, the eigenvalues ​​are solved and sorted, so that the cumulative variance contribution rate meets (See Figure 3 );

[0066] in, For the The eigenvalues ​​of the principal components, is the principal component index, is the number of principal components screened, is the sequence number of the original feature, is the total number of original features, For the The eigenvalues ​​of the original features.

[0067] It should be noted that the contribution rate threshold is based on the empirical balance point of principal component analysis in the transportation field, that is, it is set at 85%, which is the optimal trade-off between retaining key features (such as spatiotemporal patterns of congestion) and avoiding a surge in computational complexity. From a theoretical point of view, in PCA applications in the transportation field, 85%-95% is the empirical threshold for retaining key information. A value below 85% may lose significant features (such as spatiotemporal patterns of congestion), while a value above 95% will increase computational complexity and easily introduce noise.

[0068] S2.2. Based on the violation ratio and real-time normal intervention intensity in the grid data table, calculate the violation ratio of grid cells. Combined with the traffic infrastructure data and real-time traffic flow, calculate the conflict probability value for each grid cell.

[0069] Specifically, the violation ratio value field is extracted from the grid data table, including the historical benchmark values ​​of the three dimensions of the ratio of driving against traffic, the ratio of running red lights, and the ratio of occupying the road; the historical policy intervention data is loaded simultaneously to obtain the real-time policy intervention intensity (such as the number of electric bicycle inspections and punishments in Beijing in 2024, which is 157), the time attenuation coefficient, and the intervention effect coefficient; based on the dynamic violation ratio formula, the time attenuation coefficient of each grid unit is calculated. The dynamic violation ratio is expressed as:

[0070] ;

[0071] in, Expressing behavior In time The proportion of violations, Indicates the violation type identifier. =1,2,3 represent driving against traffic, running a red light and occupying the road respectively. Indicates behavior without intervention The initial ratio, represents the time variable, Represents the time decay coefficient, which is used to control the decay rate of the impact of historical violations. represents the policy intervention effect coefficient, Indicates time the intensity of real-time policy intervention;

[0072] Obtain lane capacity based on traffic infrastructure data (such as main roads =10,000 vehicles / hour); access the real-time traffic flow monitoring interface to obtain real-time traffic flow (such as Beijing's morning rush hour = 1200 vehicles / hour); calibrate the collision intensity coefficient using historical accident data (such as retrograde =1.5, running a red light =1.3, occupy the road = 1.1), and the collision probability value is calculated for each grid cell. The expression is:

[0073] ;

[0074] in, Indicates time The probability of conflict, Indicates the number of violation types. represents the conflict intensity coefficient, represents the lane capacity, Indicates real-time traffic flow.

[0075] S2.3. Establish an improved HD-PSO model, calculate the path congestion risk index based on the path node parameters in the historical evacuation event data, associate the conflict probability values ​​to generate a dynamic congestion impact factor matrix, and calculate the dynamic risk index based on the real-time traffic flow.

[0076] Specifically, we retrieve the path node parameters from the historical evacuation event data, load the stadium exit node capacity, subway transfer station node capacity, and train station exit node capacity, set the maximum tolerable delay time, and calculate the path congestion risk index for each grid cell node by node in combination with the crowd density value to quantify the node congestion risk. The expression is:

[0077] ;

[0078] in, Indicates time The path congestion risk index is used to measure the path congestion risk in time. The congestion level, the larger the value, the higher the congestion risk. represents the node index, Indicates the number of nodes, Representation node The lane capacity, Indicates the maximum tolerable delay time;

[0079] Load the path congestion risk index and conflict probability value; through the spatial location mapping relationship (node ​​coordinates → grid ID), associate and match the path congestion risk index and conflict probability value of the same geographical location to generate a dynamic congestion impact factor matrix; retrieve the historical traffic flow database to obtain historical accident benchmark values; extract the current travel delay time and the baseline travel time in the non-congestion state from the real-time traffic condition platform; based on the path congestion risk index and conflict probability value in the dynamic congestion impact factor matrix, combined with the real-time traffic flow, calculate the dynamic risk index, which is expressed as follows:

[0080] ;

[0081] in, Represents a dynamic risk index, which is used to comprehensively quantify safety and efficiency risks. represents the weight parameter of security risk, represents the weight parameter of traffic efficiency, Indicates the historical accident benchmark value, Indicates the current traffic delay time. Indicates the benchmark travel time in a non-congested state.

[0082] It is further explained that the weight parameters of safety risk and traffic efficiency refer to the improved HD-PSO model, which sets the particle position as the weight parameter of safety risk and traffic efficiency, and the fitness function is to minimize the dynamic risk index, and output the optimized weight parameters of safety risk and traffic efficiency.

[0083] By analyzing traffic accident and violation data (see Figure 4 ), and found that the accident rate is strongly correlated with violation behavior. In the evacuation scenario, avoiding casualties is the primary goal (for example, a wrong-way accident may lead to mass casualties), so a higher weight is given to safety risk, set to 0.6; the weight parameter of traffic efficiency is to take into account evacuation efficiency (for example, avoiding secondary risks caused by excessive congestion) under the premise of ensuring safety, so it is set to 0.4.

[0084] It should be noted that an improved HD-PSO model is established, and the weight parameters of safety risk and traffic efficiency are dynamically adjusted through particle swarm optimization to solve the problem that traditional static weights cannot adapt to emergency scenarios (such as fire and stampede); the fitness function simultaneously optimizes safety and efficiency, suppresses the probability of conflict, and controls delay time.

[0085] S2.4. Generate a comprehensive node risk table based on the path congestion risk index and the dynamic risk index, apply the four-layer warning rules to mark the risk status and trigger the corresponding response instructions.

[0086] Specifically, the path congestion risk index and the dynamic risk index are combined and associated through the node code field to generate a node risk comprehensive table. Based on the node risk comprehensive table, the four-layer dynamic warning rules are applied to determine the status of each node:

[0087] when and When , the mark status is low risk;

[0088] when and When , the mark status is medium risk;

[0089] when or When , the mark status is high risk;

[0090] when When , it is marked as critical risk;

[0091] Generate an early warning status table based on low risk, medium risk, high risk and critical risk;

[0092] Trigger graded response instructions based on the warning status table:

[0093] When the risk is low, no instructions are generated;

[0094] When the risk is medium, a yellow warning instruction is generated, which requires strengthening monitoring and preparing to evacuate people;

[0095] When the risk is high, an orange warning instruction is generated, which requires immediate dispatch of additional guides and adjustment of signage.

[0096] When the risk is critical, a red warning instruction is generated, which is to start the emergency plan and open the backup channel (see Figure 5 ).

[0097] Further explanation: based on historical accident data and congestion patterns, the low path risk threshold is set to 0.3, the medium path risk threshold is set to 0.6, and the high path risk threshold is set to 0.9; based on multi-objective optimization and accident correlation verification, the dynamic risk safety threshold is set to 0.4, and the dynamic risk warning threshold is set to 0.7.

[0098] S3. Based on the real-time risk index and hierarchical response instructions, the NSGA-III algorithm is used to optimize the evacuation path weight parameters and resource allocation factors. Constraints are applied to solve the multi-objective optimal solution and the optimized evacuation path weight parameters and resource allocation factors are output.

[0099] S3.1. Filter high-risk nodes from the warning status table, associate them with graded response instructions, generate a high-risk node historical data table, initialize the evacuation path weight parameters and resource allocation factor optimization parameters of the NSGA-III algorithm, and define the multi-objective optimization function (see Figure 6 ).

[0100] Specifically, the node code set with a warning status of "high risk" or "critical risk" is filtered out from the warning status table; the hierarchical response instruction table is loaded at the same time, and the response measure description of the corresponding node is extracted. Based on the risk node code set, the historical evacuation record database is retrieved, and the average time spent by the risk node in the most recent evacuation event is output to form the evacuation time history value, the accident rate of the risk node under the same warning status is output to form the accident rate statistics, and the amortized cost of the risk node executing the response instruction is output to form the resource consumption cost value, ultimately generating a high-risk node historical data table.

[0101] According to the high-risk node historical data table, initialize the NSGA-III algorithm parameters: chromosome encoding uses binary form, in which represents the evacuation path weight parameter, Represents the resource allocation factor; sets the population size individuals and the maximum number of iterations, and defines the multi-objective optimization function as minimizing evacuation time, minimizing safety risks, and minimizing resource costs.

[0102] S3.2. Execute NSGA-III optimization according to the defined multi-objective optimization function and initialization parameters, and output the optimized evacuation path weight parameters and resource allocation factors (see Figure 7 ).

[0103] Specifically, based on the defined multi-objective optimization function and initialization parameters, three types of risk transmission constraints are applied to evaluate the population feasibility of the initialization parameters: the crowd density constraint requires loading real-time monitoring density values ​​from the evacuation network parameter library; the channel width constraint forces the evacuation channel width; and the risk transmission constraint combines the warning status table to apply an accident rate correction rule to critical risk nodes: if the node status is "critical risk", the accident rate is corrected; otherwise, the original value remains unchanged.

[0104] The NSGA-III optimization process is executed: non-dominated sorting is performed, the objective function value of each individual is calculated, and the levels are divided according to the Pareto dominance relationship (for example, level 1 is the optimal non-dominated solution set); uniformly distributed reference points are constructed in the three-dimensional target space, a simulated binary crossover operation is performed, and polynomial mutation is performed; offspring are selected based on the Pareto level and the minimum distance to the reference point (prioritizing the retention of individuals with level 1 and closest distance to the reference point), and the iteration is carried out until convergence; the Pareto optimal solution set is extracted from the final population to generate the optimized evacuation path weight parameters and resource allocation factors.

[0105] S4. Based on the optimized evacuation path weight parameters and resource allocation factors, and combined with the real-time environmental status, the DRL-LSTM hybrid model is used to predict the crowd density and dynamically update the path weight parameters to generate an optimized evacuation strategy.

[0106] S4.1. Construct an environmental state vector and apply the LSTM gating mechanism to output the crowd density prediction value. Combined with the actual number of accidents, the optimized evacuation path weight parameters are updated through DRL.

[0107] Specifically, the system loads historical crowd density time series data (including key nodes such as stadium exits and subway transfer stations) to generate a timestamp-location coordinate-density value triple sequence. It also simultaneously accesses the data stream of real-time environmental monitoring equipment and integrates fire spread trends, structural damage rates, and toxic gas concentrations to generate an environmental state vector.

[0108] The LSTM gating mechanism is applied to process the input triple sequence: the input gate controls the inflow of new information (such as sudden changes in current passenger flow), the forget gate filters out invalid historical data (such as outdated congestion records), and the output gate generates a predicted passenger flow density value;

[0109] Based on the environmental state vector, the crowd density value predicted by LSTM and the actual number of accidents monitored in real time are loaded at the same time. The optimized evacuation path weight parameter is updated through DRL. The expression is:

[0110] ;

[0111] in, Indicates time The updated evacuation path weight parameter of the iterative step is Indicates time The iterative step evacuation path weight parameter, Represents the learning rate, which is used to control the update step size of the evacuation path weight parameters and balance the convergence speed and stability. Represents the predicted crowd density value, Indicates the actual number of accidents.

[0112] It should be noted that the LSTM gating mechanism accurately captures the dynamics of pedestrian flow and reduces errors; the DRL gradient update achieves a balance between safety and efficiency. When the predicted pedestrian density value is higher than the actual value and accidents occur frequently, the weight parameter of the evacuation path is adjusted upward (safety first); when the predicted pedestrian density value is lower than the actual value and accidents occur frequently, the weight parameter of the evacuation path is adjusted downward (safety first).

[0113] S4.2. Based on the updated path weight parameters, combined with the safety factor and efficiency factor, calculate the priority weight of each path, and integrate the resource allocation factor and the predicted crowd density value to generate an optimized evacuation strategy.

[0114] Specifically, based on the output updated path weight parameters and the structural damage rate in the environmental state vector, the safety factor is calculated (safety factor = 1-structural damage rate); the channel width constraint value (such as main channel W = 3.0 meters) in the evacuation network parameter library is retrieved to calculate the efficiency coefficient (efficiency coefficient = channel width / reference width);

[0115] Based on the safety factor, efficiency factor and updated optimized path weight parameters, the priority weight of each path is recalculated. The expression is:

[0116] ;

[0117] The priority weights of each path are combined with the resource allocation factor and the STM crowd density prediction value to dynamically generate an optimized evacuation strategy including resource allocation plans and dynamic avoidance instructions.

[0118] Based on the safety factor (corrected by the structural damage rate) and the efficiency factor (limited by the channel width constraint), the priority weights of each path are recalculated. Combined with the resource allocation factor and the predicted value of crowd density, an optimized evacuation strategy is generated, including a priority path sequence, a resource allocation plan, and dynamic avoidance instructions.

[0119] S5. Execute optimized evacuation strategies, implement evacuation guidance and emergency response in a real-time environment, verify evacuation effectiveness through the digital twin platform, and output an evacuation execution report.

[0120] S5.1. Execute optimized evacuation strategies and resource allocation plans, combine with the warning status table, continuously update the path weight parameters, and dynamically optimize the evacuation strategy (see Figure 9 ).

[0121] Specifically, it implements an optimized evacuation strategy, dynamically displays route arrows on electronic indicator screens, broadcasts real-time multi-language guidance, and automatically switches to an alternate route (e.g., gymnasium east exit to subway station entrance D) when real-time gas concentration exceeds the standard based on dynamic avoidance instructions (e.g., gas concentration > 100 ppm). It also implements a resource allocation plan: deploying guides and dynamically adding them based on real-time passenger flow growth rates (e.g., adding one guide for every 10% increase in passenger flow). It also activates emergency lighting equipment and adjusts the illumination range based on aisle width constraints.

[0122] When the warning status table indicates "critical risk," the backup channel gate (such as the stadium's underground emergency exit) is physically opened, a drone swarm is dispatched to project laser path guidance, and an audible and visual alarm system is triggered to prompt emergency evacuation. Facial recognition records are used to record the time from the release of multilingual guidance to the last person arriving in the safe zone, forming the actual evacuation time. Surveillance video streams are analyzed to calculate the ratio of collisions and stampedes to the total number of people evacuated, forming the actual accident rate. Gas mask collection records are scanned to calculate the percentage of actual number of gas masks issued to the total number of gas masks delivered, forming the resource utilization rate. When the actual evacuation time exceeds the predicted time, the path weight is adjusted in real time, and the path weight parameters are continuously updated to achieve dynamic optimization of the evacuation strategy.

[0123] S5.2. Reconstruct a virtual scenario on the digital twin platform based on the real-time environmental status and updated evacuation strategy parameters. Output the evacuation time interval and accident rate distribution through Monte Carlo simulation. Verify the effectiveness of the strategy by combining the path overlap rate. Generate an evacuation execution report.

[0124] Specifically, based on the real-time environmental status of the evacuation strategy in the dynamic optimization process (such as the fire spread trend of 2.5 square meters / minute, the structural damage rate of 0.3) and the updated evacuation strategy parameters (such as =0.57, =0.85), reconstructing a 3D virtual scene on the digital twin platform, accurately restoring physical parameters such as channel width and node capacity, running Monte Carlo simulations to cover extreme scenarios such as sudden surges in passenger flow and equipment failures, outputting evacuation time confidence intervals and accident rate distributions, and verifying that the results are consistent with the traffic efficiency prediction data (see Figure 8 );

[0125] At the same time, the virtual path is compared with the actual trajectory, the path overlap rate is calculated to verify the effectiveness of the strategy, and a simulation verification report containing position coordinates, time intervals and accident rate distribution is generated (see Figure 10 ); Integrate actual execution data and simulation results to generate a structured indicator comparison table to show the compliance status of evacuation time, accident rate and path overlap rate; create a parameter optimization trajectory diagram based on dynamic adjustment records, with the horizontal axis as the time series and the vertical axis as the weight parameter change curve, reflecting the real-time optimization process, and finally output an evacuation execution report.

[0126] This embodiment also provides a computer device suitable for the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration proposed in the above embodiment.

[0127] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0128] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0129] In summary, the present invention reduces risk identification delay by: integrating the spatiotemporal distribution of behaviors such as driving against traffic and running red lights in real time with the dynamic violation ratio based on the conflict probability matrix; triggering a hierarchical response by combining the path congestion index with the dynamic risk index; and using an improved HD-PSO model to output the Pareto optimal solution. The fitness function is used to compress evacuation time, reduce accident rates, and improve resource utilization.

[0130] 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 the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration, characterized by: include, Collect real-time pedestrian density data and electric bicycle traffic data, clean and standardize them, and construct a spatiotemporal distribution map of traffic violation risks; Based on the spatiotemporal distribution map of traffic violation risks, an improved HD-PSO model is established. The risk index of path nodes is calculated using a dynamic congestion impact factor matrix. A four-layer dynamic early warning mechanism is constructed to output real-time risk indexes and hierarchical response instructions. Based on the real-time risk index and hierarchical response instructions, the NSGA-III algorithm is used to optimize the evacuation path weight parameters and resource allocation factors. The constraints are applied to solve the multi-objective optimal solution and the optimized evacuation path weight parameters and resource allocation factors are output. Based on the optimized evacuation path weight parameters and resource allocation factors, and combined with real-time environmental conditions, the DRL-LSTM hybrid model predicts crowd density and dynamically updates the path weight parameters to generate an optimized evacuation strategy. Execute optimized evacuation strategies, implement evacuation guidance and emergency response in a real-time environment, verify evacuation results through the digital twin platform, and output evacuation execution reports.

2. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration according to claim 1, characterized in that: The electric bicycle traffic data includes the proportion of wrong-way driving, the proportion of running red lights, the proportion of road occupation, the number of traffic accidents, traffic flow and speed limit policy enforcement intensity parameters.

3. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration according to claim 1, characterized in that: The construction of the spatiotemporal distribution map of traffic violation risks includes the following steps: Obtain pedestrian density data and electric bicycle traffic data, clean outliers, fill in missing fields, supplement historical parameters, and perform normalized mapping; Construct an urban violation intensity index and integrate the mapped data to generate a spatiotemporal fusion dataset. Through time period density analysis, high-conflict area labeling and violation ratio aggregation, the spatiotemporal distribution map of traffic violation risks is output.

4. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective coordination according to claim 1, characterized in that: The output of real-time risk index and graded response instructions includes the following steps: Analyze the spatiotemporal distribution map of traffic violation risks, extract the coordinates and density ratio data of high-conflict areas, and generate a gridded data table after gridding transformation and principal component analysis dimensionality reduction; Based on the violation ratio and real-time normal intervention intensity in the grid data table, the violation ratio of grid units is calculated. Combined with the traffic infrastructure data and real-time traffic flow, the conflict probability value is calculated for each grid. An improved HD-PSO model was established to calculate the path congestion risk index based on the path node parameters in historical evacuation event data. The conflict probability values ​​were associated to generate a dynamic congestion impact factor matrix, and the dynamic risk index was calculated based on real-time traffic flow. A comprehensive node risk table is generated based on the path congestion risk index and the dynamic risk index, and the four-layer warning rules are applied to mark the risk status and trigger corresponding response instructions.

5. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective coordination according to claim 4, characterized in that: The risk status includes low risk, medium risk, high risk and critical risk.

6. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective coordination according to claim 1, characterized in that: The outputting of optimized evacuation path weight parameters and resource allocation factors includes the following steps: Filter high-risk nodes from the warning status table, associate hierarchical response instructions, generate a high-risk node historical data table, initialize the evacuation path weight parameters and resource allocation factor optimization parameters of the NSGA-III algorithm, and define a multi-objective optimization function; According to the defined multi-objective optimization function and initialization parameters, NSGA-III optimization is performed to output the optimized evacuation path weight parameters and resource allocation factors.

7. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective coordination according to claim 1, characterized in that: The generating of the optimized evacuation strategy comprises the following steps: Construct an environmental state vector, apply the LSTM gating mechanism to output the crowd density prediction value, combine it with the actual number of accidents, and use DRL to update the optimized evacuation path weight parameters; Based on the updated path weight parameters, combined with the safety factor and efficiency coefficient, the priority weight of each path is calculated, and the resource allocation factor and the predicted value of crowd density are integrated to generate an optimized evacuation strategy.

8. The intelligent traffic evacuation method based on dynamic path optimization and multi-objective coordination according to claim 1, characterized in that: The output of the evacuation execution report includes the following steps: Execute optimized evacuation strategies and resource allocation plans, combine with the warning status table, continuously update path weight parameters, and dynamically optimize evacuation strategies; On the digital twin platform, a virtual scene is reconstructed based on the real-time environmental status and updated evacuation strategy parameters. The evacuation time interval and accident rate distribution are output through Monte Carlo simulation. The effectiveness of the strategy is verified by combining the path overlap rate, and an evacuation execution report is generated.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent traffic evacuation method based on dynamic path optimization and multi-objective collaboration described in any one of claims 1 to 8 are implemented.

Images