Distribution method and system based on crowd density monitoring in disaster emergency hedge transfer

Abstract

The invention discloses a distribution method and system based on crowd density monitoring in disaster emergency hedging transfer. The method comprises the following steps: acquiring personnel position and moving speed data in real time through a sensor network deployed at a hedging channel and a temporary gathering point; identifying and quantifying a personnel gathering area, and positioning an excessive gathering area; the boundary, the people flow direction and the number of people of the excessive gathering area are extracted, and a plurality of alternative shunting paths leading to the standby gathering point are generated in combination with the topological structure of the peripheral channel, the available width and the obstacle information; simulating a shunting process to predict the future load of each path and the gathering point, and selecting a final shunting path from the alternative paths according to a load balancing principle to form a shunting scheme; and finally, integrating into a global transfer path diagram, generating an updated evacuation path, and executing guidance. According to the invention, real-time sensing, dynamic path planning and global load optimization of the crowd gathering risk are realized, so that the evacuation efficiency and the overall safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of disaster risk transfer, in particular to a shunting method and system based on crowd density monitoring in disaster emergency risk transfer. BACKGROUND

[0002] In the field of disaster management and public safety, ensuring the efficient and safe evacuation of large crowds in emergency situations is a significant practical issue. Currently, emergency evacuation in public places mainly relies on pre-established fixed routes and static plans. However, the dynamic and random nature of crowd flow, as well as unexpected obstacles in the environment, often render traditional static solutions ineffective in real emergency situations. The core technical challenge lies in the inability to accurately perceive and quantify the dynamic changes in crowd distribution during the evacuation process in real time, and the lack of adaptive path planning and adjustment capabilities based on these dynamic changes.

[0003] Existing solutions focus more on post-mortem analysis or offline simulation based on ideal conditions, such as optimizing evacuation plans through historical data analysis or computer simulations of set scenarios. While these methods have some guiding value, their greatest flaw is their inability to interact with real-time field data, making it impossible to identify and respond to sudden local high-density aggregation areas during the evacuation process. In addition, even though some systems have access to real-time monitoring data, their decision-making logic is often simple, such as triggering general alarms or instructions based on a single area density threshold, lacking the intelligent shunting capability of multi-area coordination and multi-path load balancing from a global perspective. This leads to problems such as intervention lag, conflicting guidance instructions, or simply transferring congestion risks from one area to another in actual applications, failing to fundamentally optimize the efficiency and safety of the entire evacuation network. SUMMARY

[0004] To overcome the deficiencies in the prior art of relying on static plans and offline simulations, failing to perceive and dynamically respond to crowd aggregation risks in real time, and lacking global load balancing optimization capabilities, resulting in delayed and inefficient emergency evacuation decisions and potentially causing secondary congestion, the present application provides a shunting method based on crowd density monitoring in disaster emergency risk transfer, which can achieve precise and adaptive crowd shunting through real-time data-driven dynamic identification, path planning combined with environmental constraints, and global optimization based on load balancing prediction, thereby improving evacuation efficiency and overall safety.

[0005] To solve the above technical problems, the present application provides a shunting method based on crowd density monitoring in disaster emergency risk transfer, comprising the following steps: Real-time collection of personnel position and movement speed data through a sensor network deployed in the risk avoidance channel and temporary gathering points to form a crowd dynamic distribution dataset; Based on the crowd dynamic distribution dataset, a plurality of regions where people gather are identified, and a density index of each region is calculated to identify an over-concentration region; For each over-concentration region identified, its boundary information, current crowd flow direction and the number of people in the region are extracted; For each over-concentration region, the topological structure, available width and obstacle position information of the surrounding channel are combined to generate a plurality of alternative shunting paths from the over-concentration region to one or more standby gathering points; Based on the number of people in the over-concentration region, the current crowd flow direction and the traffic capacity of each alternative shunting path, shunting simulation is performed to predict the load status of each path and the target gathering point after shunting. According to the load balancing principle, at least one final shunting path is selected for each over-concentration region from the alternative shunting paths to form a shunting scheme; The final shunting path in the shunting scheme is integrated into the original global transfer path graph to generate an updated global transfer path and execute shunting guidance.

[0006] In an embodiment of the present application, the arrangement and data collection of the sensor network include: A plurality of main sensor nodes are deployed in the refuge channel and the temporary gathering point, each node measures the distance and direction of the personnel target in the covered area to obtain the original position data; Auxiliary sensor nodes are deployed at key nodes in the channel to obtain personnel movement direction data; The original position data obtained by the main sensor nodes and the movement direction data obtained by the auxiliary sensor nodes are spatio-temporally fused to correct measurement errors and form a precise coordinate sequence of personnel; Based on the precise coordinate sequence of personnel, continuous tracking is performed, the coordinate displacement and time interval are calculated to obtain the movement speed of personnel; the position coordinates and movement speed of all tracked personnel are summarized to form a crowd dynamic distribution dataset.

[0007] In an embodiment of the present application, based on the crowd dynamic distribution dataset, a plurality of regions where people gather are identified, including: Map the real-time collected personnel position coordinates to the preset spatial grid to generate an initial personnel distribution map; Based on the initial personnel distribution map, adjacent high personnel density grid units are merged through density continuity judgment to form a plurality of initial personnel gathering regions.

[0008] In an embodiment of the present application, after generating the initial personnel distribution map, boundary dynamic correction is added, including: For each boundary grid cell of the initial personnel gathering area, analyze the movement vectors of the personnel inside it in continuous time steps; if the movement vectors of more than half of the personnel in the boundary grid cell are directed towards the inside of the gathering area, the boundary grid cell is marked as a stable incorporated state; if their movement vectors are mainly directed outward or randomly dispersed, it is marked as a dynamic detached state. Based on the stable inclusion and dynamic detachment states of the boundary grid cells, the outline of the initial population gathering area is modified; the boundary grid cells in the stable inclusion state are formally incorporated into the initial population gathering area, while the grid cells in the dynamic detachment state are separated from the initial population gathering area or designated as independent low-density transition zones; thus forming an optimized population gathering area with clear boundaries and well-defined population affiliation.

[0009] In one embodiment of the present invention, calculating the density index of each region to identify over-clustered areas includes: Calculate multiple core density indicators for each initial cluster area. The core density indicators include the number of people per unit area, the spatial dispersion of the population distribution, and the instantaneous change rate of the population. Each initial clustering region's core density indicators are compared with preset thresholds. When multiple indicators exceed the limit simultaneously, it is determined to be an over-clustered region.

[0010] In one embodiment of the present invention, generating multiple alternative routing paths from an overcrowded area to one or more alternative cluster points includes: Starting from the geometric center of the overcrowded area, the physical space within the preset search range is discretized into a grid, and a dynamic passage cost is assigned to each grid. Starting from the origin, path exploration is initiated simultaneously towards multiple alternative target aggregation points; during the exploration process, grid directions with low cumulative travel costs are prioritized for expansion, and after each expansion, the resource allocation of each exploration path is dynamically adjusted based on the remaining estimated costs of the newly expanded grid and each target point; When the exploration frontier first reaches any target node, the exploration path is recorded and marked as a candidate path; the search process continues until at least one path to each target node is found, or the total number of explorations reaches a preset limit; all successful paths that backtrack to the starting point are output as a set of candidate diversion paths.

[0011] In one embodiment of the present invention, the initial value of the dynamic passage cost is determined by the basic passage capacity of the corresponding location, and is dynamically adjusted upward based on the distance between the grid and the real-time obstacle location and the current personnel distribution density.

[0012] In one embodiment of the present invention, after the step of constructing the dynamic access cost map, a multi-strategy cost perturbation step is added: Define at least two different passage cost perturbation strategies, including: a first strategy that imposes additional cost on grids near the edge of high-density crowds, and a second strategy that imposes additional cost on grids within narrow passages; Based on a differentiated perturbation strategy, multiple versions of the cost map are constructed in parallel; when performing multi-objective exploratory search, the grid cost is read from different versions of the cost map in turn.

[0013] In one embodiment of the present invention, based on the load balancing principle, at least one final distribution path is selected from the candidate distribution paths for each over-aggregated area to form a distribution scheme, including: Set the analog clock and time step; simultaneously allocate the total number of people in the overcrowded area to the starting point of each alternative diversion path according to the preset initial allocation ratio; Within each simulation time step, based on the real-time traffic capacity of different sections of each path, the maximum number of people that can flow forward from the current section within the time step is calculated; people on each path are pushed forward according to this upper limit, and the instantaneous number of people in each section of each path and the target assembly point is updated, i.e., the instantaneous load. During the simulation, the personnel advance speed on each alternative diversion path and the load growth rate of the target assembly point are compared in real time. When a significant decrease in speed is detected on a certain path or an excessively rapid increase in the load of a certain assembly point is detected, the initial personnel allocation ratio assigned to this path or pointing to this assembly point is automatically reduced, and the reduced ratio is redistributed to other paths with lighter loads. Repeat the above steps until all personnel have arrived at the target assembly point by the end of the simulation, or the simulation reaches a steady state; record the final personnel allocation ratio and corresponding path selection that can make the advancement speed of each path relatively balanced and the final load of each assembly point not exceed its capacity under steady state, thus forming a diversion plan.

[0014] To address the aforementioned technical problems, this invention also provides a diversion system based on population density monitoring for disaster emergency evacuation and relocation, used to implement the above method, comprising: The data acquisition module is used to collect real-time data on the location and movement speed of people through a sensor network deployed in emergency escape routes and temporary assembly points, forming a dynamic population distribution dataset. The clustering identification module is used to identify multiple areas where people are clustered based on the dynamic distribution dataset of the population, calculate the density index of each area, and identify areas of excessive clustering. The regional parameter extraction module is used to extract the boundary information, current crowd flow direction, and number of people in the area for each identified overcrowded area; The alternative route planning module is used to generate multiple alternative diversion paths from each overcrowded area to one or more backup hubs, taking into account the topology of the surrounding channels, available width, and obstacle location information. The diversion simulation decision module is used to perform diversion simulation based on the number of people in the over-gathering area, the current direction of crowd flow, and the capacity of each alternative diversion path. It predicts the load status of each path and target gathering point after diversion, and selects at least one final diversion path for each over-gathering area from the alternative diversion paths according to the load balancing principle, thus forming a diversion plan. The global path integration and execution module is used to integrate the final routing path in the routing scheme into the original global transfer path graph, generate the updated global transfer path, and execute the routing guidance.

[0015] The technical solution of the present invention has the following advantages compared with the prior art: The disaster emergency evacuation and relocation method and system based on population density monitoring described in this invention constructs a decision-making and execution chain of real-time monitoring, intelligent identification, dynamic planning, and global optimization. First, a sensor network deployed at key nodes continuously collects personnel location and speed data, forming the real-time data foundation for all subsequent analysis and decision-making, solving the problem of traditional methods relying on offline data or non-real-time information. Based on this real-time dataset, dynamic density indicators for each area are calculated, enabling objective and quantitative identification of overcrowded areas and achieving precise location of core risk points, replacing unreliable methods that rely on manual observation or subjective judgment using fixed cameras.

[0016] Based on the accurate identification of risk sources, this solution further extracts key parameters such as boundaries, pedestrian flow direction, and number of people for each identified over-crowded area. These parameters provide specific optimization objectives and constraints for subsequent path planning. Combining the actual physical conditions of the surrounding passageways for each target area (such as topology, available width, and obstacle locations), a path search algorithm generates multiple alternative diversion paths pointing to different backup gathering points. This step ensures that the planned paths are practically feasible, rather than simply representing the theoretical shortest distance.

[0017] Furthermore, a traffic diversion simulation and prediction mechanism based on the load balancing principle was introduced. By simulating the distribution of pedestrian flow under different diversion strategies, the load status of each alternative path and target gathering point after diversion execution was predicted. This simulation and prediction process allows decision-making to proactively assess the impact of different schemes on the global evacuation network. Finally, based on the load balancing principle, the diversion path that can optimally balance the pressure of each path and avoid creating new congestion points was selected from the alternative schemes, thus forming the final diversion scheme. Finally, by integrating the optimized diversion scheme into the global transfer path map and implementing guidance, a closed loop from analysis and decision-making to on-site control was completed.

[0018] In summary, the technical solution of the present invention, through the orderly combination and synergistic effect of the above-mentioned features, can produce the following beneficial effects: First, the real-time data-driven dynamic identification of crowd gatherings significantly improves the timeliness and objectivity of identifying high-risk areas during evacuation, providing a reliable basis for early intervention.

[0019] Second, by combining personalized path planning with real-time environmental constraints, the feasibility and efficiency of diversion suggestions are improved.

[0020] Third, through global load balancing optimization based on simulation prediction, the diversion decision not only focuses on alleviating the current local congestion, but also takes into account the balance and stability of the overall traffic of the entire evacuation network, thereby shortening the overall evacuation time at the system level and further improving the safety of the entire process by avoiding secondary congestion. Attached Figure Description

[0021] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of the steps of the diversion method based on population density monitoring in disaster emergency evacuation and relocation according to the present invention; Figure 2 This is a flowchart illustrating the arrangement of the sensor network and data acquisition in this invention. Figure 3 The flowchart of the steps in this invention to identify multiple areas of population concentration and over-concentration areas based on a dynamic population distribution dataset; Figure 4 The present invention provides a flowchart of the steps for generating multiple alternative diversion paths from an over-aggregated region to one or more alternative cluster points. Figure 5 The present invention provides a flowchart of the steps for selecting at least one final diversion path from candidate diversion paths to form a diversion scheme for each over-aggregated region based on the load balancing principle. Figure 6The structural framework diagram of the diversion system based on population density monitoring in disaster emergency evacuation and relocation according to the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0023] Reference Figure 1 As shown, this invention proposes a diversion method based on population density monitoring during disaster emergency evacuation, comprising the following steps: By deploying sensor networks in emergency escape routes and temporary assembly points, real-time data on people's location and movement speed can be collected to form a dynamic population distribution dataset. To obtain real-time dynamic information about the population, it is necessary to collect data on people's location and movement speed and create a dynamic population distribution dataset. For example, infrared sensors or cameras can be deployed in emergency escape routes and temporary assembly points to periodically record the time and approximate location of people passing through, or manual patrols can be used to record the number of people and their approximate direction of movement within a specific area. After summarizing and preliminary processing, this raw data can form a dataset reflecting the distribution and movement trends of the population.

[0024] Based on this dynamic population distribution dataset, it is necessary to identify multiple areas of population aggregation, calculate the density index of each area, and identify areas of over-aggregation. Specifically, the monitoring area can be divided into several fixed-size grids, and the number of people in each grid can be counted to obtain the population distribution. Subsequently, a fixed density threshold can be set. When the population density of a grid or adjacent grid groups exceeds this threshold, it is identified as a population aggregation area. For these aggregation areas, their average population density can be calculated and compared with the preset over-aggregation threshold to determine whether over-aggregation exists.

[0025] Furthermore, for each identified overcrowded area, it is necessary to extract its boundary information, current crowd flow direction, and the number of people in the area. For example, boundary information can be obtained by polygon fitting of the geometric contour of the clustered area. The crowd flow direction can be roughly determined by observing the overall movement trend of people in the area, such as through manual observation or statistical analysis using direction sensors. The number of people in the area can be obtained by counting all the people in that area.

[0026] Based on this, for each overcrowded area, multiple alternative diversion paths are generated from that area to one or more backup assembly points, taking into account the topology, available width, and obstacle location information of surrounding passageways. For example, multiple fixed evacuation routes can be pre-defined as alternative paths, selected based on historical experience or map information. Alternatively, a graph search algorithm can be used to find the shortest path from the overcrowded area to the backup assembly point within a pre-defined passageway network, considering the static width information of the passageways as a basis for path selection.

[0027] Subsequently, based on the number of people in the overcrowded area, the current flow direction of the crowd, and the capacity of each alternative diversion path, a diversion simulation is conducted to predict the load status of each path and the target assembly point after diversion. Then, according to the load balancing principle, at least one final diversion path is selected from the alternative diversion paths for each overcrowded area, forming a diversion plan. Specifically, the total number of people in the overcrowded area and the maximum capacity of each alternative path are proportionally allocated to predict the final number of people on each path and at the target assembly point after evacuation. Then, based on the predicted load of each path and assembly point, the paths with lower loads are selected as the final diversion paths to avoid new congestion.

[0028] Finally, the final diversion paths in this diversion plan are integrated into the existing global evacuation path map to generate updated global evacuation paths and implement diversion guidance. For example, the selected final diversion paths can be highlighted and overlaid on the existing evacuation map, and people can be guided to evacuate according to the updated paths through broadcasts, signs, or manual guidance.

[0029] This application dynamically identifies overcrowded areas by collecting real-time data on people's location and movement speed, and generates alternative diversion routes by combining this data with channel information. Through diversion simulation and load balancing principles, the final diversion route is selected and integrated into a global evacuation path map for guidance. Therefore, this application effectively solves the problems of traditional static evacuation schemes, such as their inability to respond to dynamic changes in the population in real time and their lack of adaptive path planning capabilities. It avoids intervention delays, conflicting instructions, and congested evacuations, thereby improving the overall efficiency and safety of disaster evacuation.

[0030] In actual disaster evacuation scenarios, the raw data collected by sensor networks may have problems such as measurement errors, insufficient positioning accuracy, or lack of information on the direction of personnel movement. This will directly affect the accuracy and reliability of subsequent crowd density identification and diversion path planning, thereby reducing the effectiveness of the entire diversion method.

[0031] In this regard, refer to Figure 2As shown, this application further proposes specific implementation methods for the deployment of sensor networks and data acquisition to ensure higher accuracy and reliability of the acquired dynamic population distribution dataset. Specifically, the method includes the following steps: First, multiple master sensor nodes are deployed along escape routes and at temporary assembly points. These master sensor nodes are key devices for acquiring basic personnel location information and can be implemented using various technologies, such as ultra-wideband (UWB) based positioning modules, Wi-Fi or Bluetooth signal strength (RSSI) positioning modules, visual sensors (cameras) combined with image recognition technology, or LiDAR. Through ranging and orientation, each master sensor node can independently or collaboratively acquire the approximate location coordinates of personnel targets within its coverage area, i.e., raw location data.

[0032] Secondly, auxiliary sensor nodes are deployed at key nodes in the passageway to acquire data on the direction of pedestrian movement. These auxiliary sensor nodes are designed to supplement the precise direction of movement information that the main sensor nodes may not be able to directly provide. These nodes are typically deployed at critical locations where the direction of pedestrian flow may change significantly, such as turns, intersections, and entrances / exits. Auxiliary sensor nodes can employ pedestrian counting and direction determination modules based on infrared beam or microwave radar, or visual sensors with more refined direction recognition capabilities. By monitoring the order and timing of pedestrian movement through these key points, auxiliary sensor nodes can provide localized direction of movement information, which is crucial for understanding the overall flow trend of the crowd.

[0033] Next, the raw position data acquired by the main sensor node and the movement direction data acquired by the auxiliary sensor node are spatiotemporally fused to correct measurement errors and form a precise personnel coordinate sequence. The raw position data may contain deviations caused by random errors, systematic errors, or multipath effects, while the direction data provided by the auxiliary sensor has high local accuracy. The fusion algorithm comprehensively considers the position information reported by the main sensor node at different time points and combines it with the movement direction constraints provided by the auxiliary sensor node to smooth and correct the personnel's trajectory, thereby effectively eliminating or reducing measurement errors in the raw data and generating a continuous and highly accurate personnel coordinate sequence.

[0034] Finally, based on the precise coordinate sequence of individuals, continuous tracking is performed. By calculating coordinate displacement and time intervals, the movement speed of each individual is obtained. The position coordinates and movement speeds of all tracked individuals are then aggregated to form a dynamic population distribution dataset. After obtaining the precise coordinate sequence, the system continuously tracks each individual, associating different coordinate points of the same individual to form their complete movement trajectory. This can be achieved through a target association algorithm. Once a continuous trajectory is established, the movement speed of an individual can be obtained by calculating the Euclidean distance (displacement) between adjacent precise coordinate points and dividing by the corresponding time interval. Ultimately, the system aggregates the real-time position coordinates and calculated movement speeds of all tracked individuals to form a dynamic population distribution dataset containing the position and speed information of each individual at a specific moment.

[0035] The real-time collected data on people's location and movement speed forms a dynamic population distribution dataset. To accurately and efficiently identify areas of high population density and over-density from this raw, discrete data, [further details are needed]. Figure 3 As shown, this application further proposes a method for identifying multiple areas of crowd gathering based on a dynamic population distribution dataset, calculating density indices for each area, and identifying over-crowded areas. Specifically, it includes the following steps: mapping real-time collected personnel location coordinates to a preset spatial grid to generate an initial personnel distribution map. This aims to discretize the continuous physical space to facilitate the quantification and analysis of personnel distribution. This step first divides physical spaces such as escape routes and temporary assembly points into a series of grid units with fixed sizes and shapes, such as square or hexagonal grids. Subsequently, the system traverses all real-time collected personnel location coordinates, determines the grid unit to which each coordinate point belongs, and accumulates the number of personnel within the corresponding grid unit. Finally, a two-dimensional or three-dimensional initial personnel distribution map is formed, where each grid unit records the number of personnel within it, thus intuitively reflecting the spatial distribution and local density of personnel.

[0036] Based on this, and using the initial population distribution map, adjacent high-density grid cells are merged through density continuity judgment to form several initial population clusters. The purpose of this step is to identify meaningful population clusters from discrete grid data. First, the system sets a density threshold, marking grid cells in the initial population distribution map with more than this threshold as high-density grid cells. Next, by analyzing the spatial adjacency relationships between these high-density grid cells (e.g., using four-neighbor or eight-neighbor judgment), all adjacent high-density grid cells are merged. This merging process can be implemented using a connected component algorithm, ensuring that only spatially continuous grid cells with consistently high densities are grouped into a single entity. Each merged set of continuous high-density grid cells is then identified as an initial population cluster, thus transforming scattered individual data into clusters with clear boundaries and internal connectivity.

[0037] Furthermore, after generating the initial population clustering areas, a method is provided to calculate the density indices of each area and identify overcrowded areas. This method includes calculating multiple core density indices for each initial clustering area. These core density indices specifically include the number of people per unit area, the spatial dispersion of population distribution, and the instantaneous rate of change in the number of people.

[0038] Among these, the number of people per unit area is the most intuitive density indicator. It is obtained by dividing the total number of people in the clustered area by the actual area occupied by that area. For example, the number of grid cells occupied by all identified people in the area can be counted, and the ratio of the total number of people to the area can be calculated. The spatial dispersion of personnel distribution reflects the evenness of personnel distribution within the clustered area. It can be measured using statistical methods such as standard deviation, variance, or Herfindahl-Hirschman index. If people are highly concentrated in a small part of the area, the dispersion is low; if people are evenly distributed, the dispersion is high. The instantaneous rate of change of personnel number is used to capture the dynamic trend of personnel change in the clustered area, reflecting the speed of personnel inflow or outflow. For example, it can be calculated by comparing the total number of people at the current time step with the previous time step and dividing by the time interval. A positive value indicates that the number of people is increasing rapidly, and a negative value indicates that the number of people is decreasing rapidly. Through these multi-dimensional indicators, the true situation of the clustered area can be assessed more comprehensively and accurately.

[0039] Subsequently, the system compares the core density indicators of each initial gathering area with the corresponding preset thresholds. For each calculated core density indicator, the system pre-sets one or more corresponding critical values. These thresholds can be dynamically adjusted based on the specific environment, carrying capacity, and disaster type of the evacuation route and temporary assembly point. When multiple indicators exceed the limits simultaneously, the system determines that the area is an over-gathering area.

[0040] In some embodiments described above in this application, an initial population clustering region is formed by mapping personnel locations to a preset spatial grid and merging adjacent high-density grid cells. However, in practical applications, this static grid merging-based method may result in inaccurate identification of clustering region boundaries, especially when personnel are continuously moving. It becomes difficult to accurately define which boundary areas truly belong to the clustering region, thus affecting the accuracy of subsequent density index calculations and overcrowded region identification.

[0041] To address this, this application further proposes adding dynamic boundary correction after generating the initial population distribution map. This dynamic boundary correction aims to solve the problem of ambiguous or inaccurate boundaries of the initial gathering area. By introducing a time dimension and analyzing population movement behavior, the outline of the gathering area can more accurately reflect the actual gathering state and affiliation of the population, thereby improving the accuracy and effectiveness of subsequent density calculations and diversion decisions. Specifically, this correction includes: for each boundary grid cell of the initial population gathering area, analyzing the movement vectors of the people inside it in continuous time steps. Here, the boundary grid cell refers to the grid cell located at the edge of the initial population gathering area. Analyzing the movement vectors of the people inside it in continuous time steps means that the system continuously monitors the displacement direction and magnitude of the people within these boundary grid cells in short time intervals. This can be obtained by calculating the position difference between adjacent time points through real-time personnel location data collected by the sensor network. For example, a time window can be set, and multiple movement vectors of the people within the window can be statistically analyzed to determine their overall movement trend. If the movement vectors of more than half of the people in the boundary grid cell point inward into the gathering area, then the boundary grid cell is marked as a stable inclusion state. A stable inclusion state indicates that the majority of people within the boundary grid cell tend to move towards the center of the cluster or remain within the cluster, suggesting that the grid cell should be considered part of the cluster. The "more than half" criterion is a statistical method, or it could be other preset percentage thresholds to ensure robustness. If their movement vectors primarily point outwards or are randomly dispersed, it is marked as a dynamic detached state. A dynamic detached state indicates that the majority of people within the boundary grid cell tend to leave the cluster, or their movement direction is not clearly directional and exhibits randomness, suggesting that the people within the grid cell may not belong to the core cluster or are leaving the area.

[0042] Based on the stable inclusion and dynamic detachment states of boundary grid cells, the outline of the initial population gathering area is corrected. Outline correction involves adjusting the geometric boundaries of the initial population gathering area after identifying the state of the boundary grid cells. This ensures that the shape and extent of the gathering area dynamically adapt to the actual movement and distribution of the population. Specifically, boundary grid cells in a stable inclusion state are formally incorporated into the initial population gathering area. Formal incorporation means that these grid cells marked as stable inclusion become part of the gathering area, and the people within them are included in the total number and density calculations of the gathering area. Simultaneously, grid cells in a dynamic detachment state are either removed from the initial population gathering area or designated as independent low-density transition zones. Removal removes dynamically detached grid cells from the original gathering area to avoid incorrectly including people who do not belong to the gathering area in the calculation. Designating them as independent low-density transition zones is a more refined approach. It acknowledges that these areas are adjacent to the core gathering area but have lower density and higher population mobility, serving as buffer zones or being monitored separately, rather than being directly included in the density calculation of the core gathering area. This results in an optimized population gathering area with clear boundaries and well-defined population affiliation. The optimized population concentration area is the result of dynamic correction, and its boundary can more accurately reflect the actual concentration range of the population, avoiding errors caused by static division. Clear population affiliation means that each grid cell and the people within it can be accurately determined whether they belong to a certain concentration area, transition zone, or non-concentration area, providing more reliable basic data for subsequent density calculation and distribution decisions.

[0043] During disaster evacuation and relocation, once overcrowded areas are identified, multiple alternative diversion routes need to be planned from the current location to backup relocation points, referring to... Figure 4 As shown, this application proposes a method for generating multiple alternative diversion paths from an overcrowded region to one or more alternative node points. The method includes: discretizing the physical space within a preset search range into grids, starting from the geometric center of the overcrowded region, and assigning a dynamic passage cost to each grid; simultaneously initiating path exploration towards multiple alternative target node points from the starting point; prioritizing the expansion of grid directions with lower cumulative passage costs during the exploration process, and dynamically adjusting the resource allocation of each exploration path based on the remaining estimated costs of the newly expanded grid and each target point after each expansion; recording the reached exploration path and marking it as an alternative path when the exploration front first reaches any target node point; continuing the search process until at least one arrival path is found for each target node point, or the total number of explorations reaches a preset upper limit; and outputting all arrival paths that successfully trace back to the starting point as a set of alternative diversion paths.

[0044] Specifically, using the geometric center of the overcrowded area as the starting point means that after identifying the overcrowded area, a representative starting point is determined by calculating the average of the coordinates of all personnel within that area, or the average of the center coordinates of the grid cells occupied by that area. This starting point serves as the starting point for the path planning algorithm, ensuring the logic and consistency of the path planning.

[0045] Discretizing the physical space within a preset search range into a grid refers to transforming the continuous physical space containing escape routes and temporary rendezvous points into a series of discrete cells of uniform size through grid partitioning. These grids can be squares, hexagons, or other regular shapes, with each cell representing a small region in the physical space. The preset search range can be set according to the actual scenario requirements, such as a certain radius range centered on the starting point, or an area containing all potential backup rendezvous points. This discretization process enables complex path planning problems to be efficiently computed numerically and searched graphically in a computer.

[0046] Assigning a dynamic passage cost to each grid cell means assigning a numerical value to each discretized spatial unit, representing the cost or difficulty required to traverse that grid. This cost is "dynamic," meaning it can be adjusted based on real-time environmental factors (e.g., but not limited to, inherent properties of the passageway, potentially hazardous areas, or general real-time congestion) to reflect the current ease or difficulty of passage. In this way, path planning algorithms can avoid high-cost areas, guiding people to choose safer and more efficient routes.

[0047] Starting from the origin and simultaneously exploring multiple candidate target nodes, this means that the path planning algorithm does not search for a single target, but rather explores all pre-defined candidate target nodes simultaneously, centered on the origin. This can be achieved by using multi-target search algorithms that maintain a priority queue containing all grids to be explored and sorted according to their cumulative cost.

[0048] The exploration process prioritizes expanding along grid directions with lower cumulative travel costs. This means that during path exploration, the algorithm continuously evaluates the cumulative travel cost from the starting point to the currently explored grid, and combines this with the estimated cost (heuristic function) from the current grid to the target node. It then prioritizes expanding along the grid with the lowest estimated total cost. This strategy ensures the algorithm can efficiently move towards the optimal path, avoiding unnecessary exploration.

[0049] After each expansion, the algorithm dynamically adjusts the resource allocation of each exploration path based on the estimated remaining cost of the newly expanded grid and each target point. This means that after each expansion operation, the algorithm updates its internal state and re-evaluates the priority of all active exploration branches. For example, if a newly expanded grid significantly reduces the estimated total cost to reach a target node, the exploration path pointing to that target node may have its weight increased in the priority queue, thus receiving more computational resources in subsequent steps or being expanded earlier. This dynamic adjustment mechanism makes the search process more adaptive and efficient.

[0050] When the exploration front first reaches any target node, the exploration path is recorded and marked as a candidate path. This means that once the search algorithm's extended front touches the boundary of any candidate target node, it signifies that a complete path from the starting point to that node has been found. At this point, the algorithm backtracks and records all the grid sequences traversed by this path, saving them as a valid candidate branching path.

[0051] The search continues until at least one path to each target node is found, or the total number of explorations reaches a preset limit. This means the path search does not stop immediately after finding the first path. Instead, it continues to ensure that at least one feasible path is found for each candidate target node. Meanwhile, to prevent the search from going indefinitely or consuming excessive computational resources, a limit is set on the total number of explorations. Once this limit is reached, the search terminates even if not all target nodes have found paths. This ensures the comprehensiveness of the path set and the controllability of the computation.

[0052] Outputting all successful arrival paths back to the starting point as a set of candidate diversion paths means that after the search process is completed, all successfully recorded complete paths from the geometric center of the overcrowded area to each candidate target node will be aggregated and output as a set. This set contains diverse diversion options, providing a rich data foundation for subsequent diversion simulations and decisions.

[0053] Specifically, the initial value of the dynamic passage cost is determined by the basic passage capacity of the corresponding location, and is dynamically adjusted upwards based on the distance between the grid and the real-time obstacle location, and the current personnel density. In essence, the basic passage capacity refers to the upper limit of personnel flow or passage speed that a specific passage segment can handle under ideal conditions. This is typically related to the physical properties of the passage, such as its width, structural strength, slope, lighting conditions, and the presence of fixed facilities. During system initialization or map building, the basic passage capacity of each grid or passage segment within escape routes and temporary assembly points can be pre-assessed and set based on these physical properties. For example, wider, flatter grids can have a lower initial passage cost, while narrower, sloping grids have a higher initial cost. This setting provides a benchmark for path planning based on the inherent properties of the infrastructure, ensuring that path selection initially favors passages with better physical conditions.

[0054] Meanwhile, real-time obstacles refer to objects that suddenly appear during a disaster or evacuation, obstructing passage, such as collapsed building debris, damaged facilities, stranded vehicles or personnel, etc. The system monitors environmental changes within the evacuation route in real time through a sensor network, identifying and locating these obstacles. When a grid cell approaches a real-time obstacle, its passage cost is dynamically adjusted upwards based on the distance to the obstacle. The closer the distance, the greater the increase in passage cost, until a grid cell completely occupied by the obstacle is assigned an extremely high (or even infinitely high) passage cost, effectively avoiding paths passing through or adjacent to dangerous areas. This dynamic adjustment mechanism enables path planning to respond quickly to emergencies and improve path safety.

[0055] Furthermore, the current population density refers to the ratio of the number of people in a specific grid or its surrounding area to the area of ​​that region. The system continuously calculates the population density of each region using real-time collected population location data. When the population density of a grid area is high, it indicates that the area is already congested or about to become congested. At this time, the passage cost of that grid will be dynamically adjusted upwards accordingly. The higher the density, the greater the increase in passage cost, thereby guiding diversion paths to avoid currently congested areas, preventing further exacerbation of congestion, and ensuring that people can pass through other relatively open paths more quickly. This adjustment based on real-time density helps to achieve dynamic and balanced distribution of crowds, improving overall transfer efficiency.

[0056] Specifically, after constructing the dynamic passage cost map, this embodiment further proposes adding a multi-strategy cost perturbation step. Specifically, the method defines at least two different passage cost perturbation strategies: The first strategy aims to impose additional costs on grid cells near the edges of high-density crowds. This means that during path planning, the system identifies the boundaries of areas with high population density and assigns higher passage costs to grid cells adjacent to these boundaries. For example, a buffer zone can be set, and the passage cost of any grid cell falling within this buffer zone will be significantly increased to avoid guiding people into or past areas that may cause congestion or panic. The second strategy imposes additional costs on grid cells within narrow passages. This is because, in disaster scenarios, narrow passages inherently have lower passage capacity and higher risk, and should be avoided as primary diversion paths even if the current population density is not high. For example, the system can pre-identify areas where the passage width is below a certain threshold and impose a fixed high passage cost on all grid cells within these areas to reduce their probability of being selected.

[0057] Based on these differentiated perturbation strategies, the system constructs multiple versions of the traffic cost map in parallel. This means that in addition to the original dynamic traffic cost map generated based on basic traffic capacity, obstacle locations, and pedestrian density, at least two or more cost maps perturbed by different strategies are generated. Each version of the cost map adjusts the traffic cost from a specific risk perspective; for example, one version may focus more on avoiding high-density edges, while another version focuses more on avoiding narrow passages. When performing multi-objective exploratory search to generate alternative diversion paths, the system reads the grid cost from different versions of the cost map in turn. For example, at each step of path exploration or every few steps, the path search algorithm switches the referenced cost map, thus comprehensively considering multiple risk factors during path generation. This turnaround reading mechanism ensures that the final generated alternative diversion paths not only consider real-time pedestrian density and obstacles but also take into account potential congestion risks and structural bottlenecks.

[0058] In actual dynamic relocation processes, due to the randomness of personnel behavior, the appearance of sudden obstacles, or real-time changes in path capacity, pre-set diversion plans may fail to adapt to the actual situation, leading to new congestion on some paths or excessively rapid load increases at some assembly points, thereby affecting overall relocation efficiency and safety. In response, refer to... Figure 5 As shown, this application further proposes a method to select at least one final distribution path from the alternative distribution paths for each over-aggregated area based on the load balancing principle, forming a distribution scheme, specifically including the following steps: First, to simulate the crowd diversion process in detail, a simulation clock needs to be set and divided into a series of discrete time steps. For example, the time steps can be set to 1 second, 5 seconds, or 10 seconds to capture the dynamic changes in crowd flow. Then, the total number of people in the overcrowded area is virtually and simultaneously allocated to the starting points of each alternative diversion path according to a pre-set initial allocation ratio. This initial allocation ratio can be initially set based on path length, capacity, historical data, or expert experience; for example, 30% of the total number of people could be allocated to path A, 40% to path B, and 30% to path C.

[0059] Within each set simulation time step, the system calculates the maximum number of people that can safely and efficiently pass through each segment of the path, based on the real-time traffic capacity of different sections. Real-time traffic capacity considers factors such as channel width, current obstacle conditions, and personnel density. The simulation system then propels people forward along the path according to this calculated maximum number of people. Simultaneously, the system updates the number of people in each segment of the path and at the final destination assembly point in real time—the instantaneous load—to reflect the actual flow of people.

[0060] During the simulation, the system continuously monitors and compares in real time the personnel advancement speed on each alternative diversion path and the personnel load growth rate at each target assembly point. For example, the advancement speed can be assessed by calculating the average distance people travel on the path per unit time, or the load growth rate can be assessed by the increase in the number of people at the assembly point per unit time. Once the system detects that the personnel advancement speed on a certain path is significantly lower than expected, indicating that the path may be congested or have reduced traffic efficiency; or if it detects that the personnel load at a certain target assembly point is growing too rapidly, potentially leading to overload, the system will immediately trigger a dynamic adjustment mechanism. At this time, the system will automatically reduce the initial personnel allocation ratio assigned to the congested path or overloaded assembly point, and intelligently reallocate this reduced personnel ratio to other alternative diversion paths or target assembly points with lighter loads and higher traffic efficiency.

[0061] The aforementioned dynamic adjustment and personnel mobilization simulation steps will be repeated until all personnel requiring relocation have successfully reached their target assembly points, or the entire simulation system reaches a steady state. A steady state refers to a state where, during continuous simulation, personnel flow and load distribution tend to stabilize, without drastic fluctuations or the need for frequent adjustments. After the simulation reaches a steady state, the system will record the final personnel allocation ratio that ensures relatively balanced personnel movement speeds along each diversion path and that the final personnel load at all target assembly points does not exceed their preset capacity. Simultaneously, the system will also record the path selections corresponding to these allocation ratios. These recorded final allocation ratios and path selections together constitute the final diversion plan, used to guide the actual personnel evacuation and relocation.

[0062] Reference Figure 6 As shown, in order to implement the above method, this application proposes a diversion system based on population density monitoring for disaster emergency evacuation, including: The data acquisition module is used to collect real-time data on the location and movement speed of people through a sensor network deployed in emergency escape routes and temporary assembly points, forming a dynamic population distribution dataset. The clustering identification module is used to identify multiple areas where people are clustered based on the dynamic distribution dataset of the population, calculate the density index of each area, and identify areas of excessive clustering. The regional parameter extraction module is used to extract the boundary information, current crowd flow direction, and number of people in the area for each identified overcrowded area; The alternative route planning module is used to generate multiple alternative diversion paths from each overcrowded area to one or more backup hubs, taking into account the topology of the surrounding channels, available width, and obstacle location information. The diversion simulation decision module is used to perform diversion simulation based on the number of people in the over-gathering area, the current direction of crowd flow, and the capacity of each alternative diversion path. It predicts the load status of each path and target gathering point after diversion, and selects at least one final diversion path for each over-gathering area from the alternative diversion paths according to the load balancing principle, thus forming a diversion plan. The global path integration and execution module is used to integrate the final routing path in the routing scheme into the original global transfer path graph, generate the updated global transfer path, and execute the routing guidance.

[0063] The data acquisition module continuously provides accurate personnel location and speed information, laying a solid foundation for subsequent decision-making. The clustering identification module and area parameter extraction module can quickly and accurately locate overcrowded areas and extract key parameters, ensuring the timeliness and relevance of diversion decisions. The alternative route planning module can comprehensively consider multiple factors to generate diverse diversion routes, increasing the flexibility and robustness of the plan. The diversion simulation decision-making module, through sophisticated simulation, can predict and optimize diversion effects, effectively avoiding secondary congestion and overload at gathering points, ensuring a smooth and safe transfer process. Finally, the global route integration and execution module seamlessly integrates dynamic diversion plans into the overall transfer strategy and provides effective guidance, significantly improving the efficiency and safety of disaster evacuation and reducing the risk of casualties, providing strong technical support for disaster management.

[0064] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A diversion method based on population density monitoring in disaster emergency evacuation, characterized in that, Includes the following steps: By deploying a sensor network in emergency escape routes and temporary assembly points, real-time data on people's location and movement speed are collected to form a dynamic population distribution dataset. Based on the dynamic distribution dataset of the population, multiple areas where people gather are identified, density indices of each area are calculated, and areas of overcrowding are identified. For each identified overcrowded area, extract its boundary information, current crowd flow direction, and number of people in the area; For each overcrowded area, multiple alternative diversion paths are generated from the overcrowded area to one or more backup gathering points, taking into account the topology of the surrounding channels, available width, and obstacle location information. Based on the number of people in the overcrowded area, the current direction of population flow, and the capacity of each alternative diversion path, a diversion simulation is conducted to predict the load status of each path and target gathering point after diversion. Based on the load balancing principle, at least one final distribution path is selected from the alternative distribution paths for each over-aggregated area to form a distribution scheme. The final routing path in the routing scheme is integrated into the original global transfer path graph, generating an updated global transfer path and executing the routing guidance.

2. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 1, characterized in that: The deployment and data acquisition of sensor networks include: Multiple main sensor nodes are deployed in the evacuation routes and temporary assembly points. Each node measures the distance and orientation of personnel targets within its coverage area to obtain raw location data. Deploy auxiliary sensor nodes at key nodes in the passageway to acquire data on the direction of personnel movement; The raw position data acquired by the main sensor node and the movement direction data acquired by the auxiliary sensor node are spatiotemporally fused to correct measurement errors and form a precise coordinate sequence of personnel. Continuous tracking is performed based on the precise coordinate sequence of personnel. The movement speed of personnel is obtained by calculating the coordinate displacement and time interval. The position coordinates and movement speed of all tracked personnel are summarized to form a dynamic distribution dataset of the crowd.

3. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 1, characterized in that: Based on the dynamic population distribution dataset, multiple areas where people congregate were identified, including: The real-time collected personnel location coordinates are mapped to a preset spatial grid to generate an initial personnel distribution map; Based on the initial population distribution map, adjacent high-density grid cells are merged by determining density continuity to form several initial population gathering areas.

4. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 3, characterized in that: After generating the initial personnel distribution map, dynamic boundary correction is added, including: For each boundary grid cell of the initial personnel gathering area, analyze the movement vectors of the personnel inside it in continuous time steps; if the movement vectors of more than half of the personnel in the boundary grid cell are directed towards the inside of the gathering area, the boundary grid cell is marked as a stable incorporated state; if their movement vectors are mainly directed outward or randomly dispersed, it is marked as a dynamic detached state. Based on the stable inclusion and dynamic detachment states of the boundary grid cells, the outline of the initial population gathering area is modified; the boundary grid cells in the stable inclusion state are formally incorporated into the initial population gathering area, while the grid cells in the dynamic detachment state are separated from the initial population gathering area or designated as independent low-density transition zones; thus forming an optimized population gathering area with clear boundaries and well-defined population affiliation.

5. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 3 or 4, characterized in that: Calculating density indices for each region identifies areas of over-aggregation, including: Calculate multiple core density indicators for each initial cluster area. The core density indicators include the number of people per unit area, the spatial dispersion of the population distribution, and the instantaneous change rate of the population. Each initial clustering region's core density indicators are compared with preset thresholds. When multiple indicators exceed the limit simultaneously, it is determined to be an over-clustered region.

6. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 1, characterized in that: Generate multiple alternative routing paths from overcrowded areas to one or more alternative hubs, including: Starting from the geometric center of the overcrowded area, the physical space within the preset search range is discretized into a grid, and a dynamic passage cost is assigned to each grid. Starting from the origin, path exploration is initiated simultaneously towards multiple alternative target aggregation points; during the exploration process, grid directions with low cumulative travel costs are prioritized for expansion, and after each expansion, the resource allocation of each exploration path is dynamically adjusted based on the remaining estimated costs of the newly expanded grid and each target point; When the exploration frontier first reaches any target node, the exploration path is recorded and marked as a candidate path; the search process continues until at least one path to each target node is found, or the total number of explorations reaches a preset limit; all successful paths that backtrack to the starting point are output as a set of candidate diversion paths.

7. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 6, characterized in that: The initial value of the dynamic passage cost is determined by the basic passage capacity of the corresponding location, and is dynamically adjusted upwards based on the distance between the grid and the real-time obstacle location and the current personnel distribution density.

8. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 7, characterized in that: After constructing the dynamic passage cost map, a multi-strategy cost perturbation step is added: Define at least two different passage cost perturbation strategies, including: a first strategy that imposes additional cost on grids near the edge of high-density crowds, and a second strategy that imposes additional cost on grids within narrow passages; Based on a differentiated perturbation strategy, multiple versions of the cost map are constructed in parallel; when performing multi-objective exploratory search, the grid cost is read from different versions of the cost map in turn.

9. The diversion method based on population density monitoring in disaster emergency evacuation according to claim 1, characterized in that: Based on load balancing principles, at least one final distribution path is selected from the alternative distribution paths for each over-aggregated area to form a distribution scheme, including: Set the analog clock and time step; simultaneously allocate the total number of people in the overcrowded area to the starting point of each alternative diversion path according to the preset initial allocation ratio; Within each simulation time step, based on the real-time traffic capacity of different sections of each path, the maximum number of people that can flow forward from the current section within the time step is calculated; people on each path are pushed forward according to this upper limit, and the instantaneous number of people in each section of each path and the target assembly point is updated, i.e., the instantaneous load. During the simulation, the personnel advance speed on each alternative diversion path and the load growth rate of the target assembly point are compared in real time. When a significant decrease in speed is detected on a certain path or an excessively rapid increase in the load of a certain assembly point is detected, the initial personnel allocation ratio assigned to this path or pointing to this assembly point is automatically reduced, and the reduced ratio is redistributed to other paths with lighter loads. Repeat the above steps until all personnel have arrived at the target assembly point by the end of the simulation, or the simulation reaches a steady state; record the final personnel allocation ratio and corresponding path selection that can make the advancement speed of each path relatively balanced and the final load of each assembly point not exceed its capacity under steady state, thus forming a diversion plan.

10. A diversion system based on population density monitoring for disaster emergency evacuation, used to implement the method described in any one of claims 1 to 9, characterized in that: include: The data acquisition module is used to collect real-time data on the location and movement speed of people through a sensor network deployed in emergency escape routes and temporary assembly points, forming a dynamic population distribution dataset. The clustering identification module is used to identify multiple areas where people are clustered based on the dynamic distribution dataset of the population, calculate the density index of each area, and identify areas of excessive clustering. The regional parameter extraction module is used to extract the boundary information, current crowd flow direction, and number of people in the area for each identified overcrowded area; The alternative route planning module is used to generate multiple alternative diversion paths from each overcrowded area to one or more backup hubs, taking into account the topology of the surrounding channels, available width, and obstacle location information. The diversion simulation decision module is used to perform diversion simulation based on the number of people in the over-gathering area, the current direction of crowd flow, and the capacity of each alternative diversion path. It predicts the load status of each path and target gathering point after diversion, and selects at least one final diversion path for each over-gathering area from the alternative diversion paths according to the load balancing principle, thus forming a diversion plan. The global path integration and execution module is used to integrate the final routing path in the routing scheme into the original global transfer path graph, generate the updated global transfer path, and execute the routing guidance.

Images