A subway station passenger flow scheduling method and system and a medium

Abstract

The application discloses a subway station passenger flow scheduling method and system and a medium, relates to the technical field of urban rail transit passenger flow management and control, and comprises the following steps: dividing a station into multiple passenger flow cells and establishing a topological connection relationship; constructing a partition pedestrian macroscopic basic graph according to the average density and the output flow sequence of each functional partition, determining the critical density and the traffic capacity of each partition; obtaining a dynamic cell transmission model based on the critical density and the traffic capacity, marking the cells as a free state, a critical state or a congestion state according to the comparison result of the cell density and the critical density, and determining the sending capacity and the receiving capacity; comparing the sending capacity and the receiving capacity to determine an imbalance chain, determining a congestion starting cell, a congestion critical cell and a diffusion link from the imbalance chain; and predicting the cell state in a future period and generating a scheduling result. The application realizes quantitative identification and refined scheduling of the congestion diffusion link, and improves the foresight and accuracy of subway station passenger flow management and control.

Description

Technical Field

[0001] This invention relates to the field of urban rail transit passenger flow management technology, and more specifically, to a subway station passenger flow scheduling method, system, and medium. Background Technology

[0002] Subway stations comprise multiple functional areas, including concourses, passageways, escalators, turnstiles, and platforms, all interconnected, with passenger flow continuously shifting between them. Currently, passenger flow management in subway stations primarily relies on video surveillance, turnstile counting, and manual patrols. Dispatchers assess congestion based on video footage and their personal experience, then implement passive intervention measures such as limiting entry, guiding passageways, or controlling escalators.

[0003] However, this passive response approach, relying on human experience, has inherent limitations: intervention only occurs after congestion has formed and been observed, resulting in a significant lag in scheduling and an inability to predict the location and direction of congestion. Furthermore, existing passenger flow data analysis or simulation methods typically only provide rough estimates of total passenger flow or density in specific areas, failing to reflect the coupling relationships between different functional zones, and even less capable of dynamic identification and early warning at critical stages of qualitative changes in passenger flow. Due to the lack of a quantitative analysis mechanism for congestion propagation links, scheduling decisions often remain at a general regional level, failing to formulate refined control strategies for specific facilities and time periods. Therefore, there is a need to research more accurate and forward-looking methods for identifying congestion propagation links and scheduling passenger flow in subway stations.

[0004] Therefore, how to provide a method, system, and medium for passenger flow scheduling in subway stations is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention proposes a method, system and medium for passenger flow scheduling in subway stations. By constructing a dynamic cellular transmission model and identifying imbalance chains, it obtains congestion diffusion links and refined scheduling instructions.

[0006] This invention proposes a method for passenger flow scheduling in subway stations, comprising: Acquire spatial structure data and passenger flow data of the station, divide the station into multiple passenger flow cells and establish topological connection relationships between each cell, and configure the functional area of ​​each passenger flow cell. Based on the average density sequence and output flow sequence of each passenger flow cell, a basic macroscopic map of pedestrians in each zone is constructed to determine the critical density and traffic capacity of each functional area. Based on the critical density and the passage capacity, the critical number of passengers in the passenger flow cells and the passage capacity of each connecting edge are determined, and a dynamic cell transmission model is constructed using the critical number of passengers and the passage capacity of each edge. The dynamic cell transmission model is used to determine the state category of the passenger flow cells based on the density state information of the passenger flow cells, and to determine the passenger flow transmission capacity based on the state association relationship between adjacent passenger flow cells. Based on the difference in passenger flow transmission capacity between the two ends of each connecting edge, connecting edges with mismatched transmission capacity are identified as unbalanced edges, and continuously distributed unbalanced edges are aggregated into unbalanced chains along the passenger flow direction. The unbalanced chains are used to locate the congestion initiation cell, the congestion critical cell, and their diffusion links. Based on the dynamic cell transmission model, the state of passenger flow cells in future time periods is predicted to obtain the state information of the passenger flow cells in future time periods. The scheduling result is generated based on the congestion initiation cell, the congestion critical cell, their diffusion links, and the predicted cell states for future time periods.

[0007] Compared with existing technologies, the beneficial effects of this invention are as follows: 1. This invention divides the station into multiple passenger flow cells and establishes topological connections. Based on the average density sequence and output flow sequence of each functional zone, it constructs a macroscopic basic map of pedestrian traffic in each zone. Using a data-driven self-learning approach, it determines the critical density and capacity of each zone. These parameters are then used as input to the cell transmission model, enabling the model to label cells as free, critical, or congested states based on the comparison between cell density and critical density. Furthermore, it dynamically determines the sending and receiving capabilities based on the state combinations of upstream and downstream cells. Therefore, it can detect and quantify the critical stage of passenger flow transitioning from free flow to congested flow in advance, eliminating the need to wait for actual congestion before intervention, thus elevating scheduling decisions from passive response to proactive prediction.

[0008] 2. This invention compares the upstream sending capacity and downstream receiving capacity of each connection edge, marking the connection edge with a sending capacity greater than its receiving capacity as an unbalanced edge. It then combines consecutively adjacent unbalanced edges in the passenger flow direction into an unbalanced chain. This unbalanced chain determines the congestion initiation cell, the congestion critical cell, and the propagation link, enabling quantitative reasoning about the location of the congestion source, the propagation direction, and the scope of its impact. Furthermore, based on the above information and the predicted cell states for future periods, it generates scheduling results containing the target cell, trigger time period, instruction type, execution object, and control quantity. This allows the scheduling strategy to accurately correspond to specific facilities, equipment, and execution time nodes, significantly improving the precision and executability of passenger flow scheduling.

[0009] Secondly, this application also provides a subway station passenger flow scheduling system for applying the aforementioned subway station passenger flow scheduling method, including: The acquisition module is used to acquire spatial structure data and passenger flow data of the station, divide the station into multiple passenger flow cells and establish the topological connection relationship between each cell, and configure the functional area of ​​each passenger flow cell. Topology module: used to construct a basic macroscopic map of pedestrians in each zone based on the average density sequence and output flow sequence of each passenger flow cell, and to determine the critical density and traffic capacity of each functional area; The determination module is used to determine the critical number of passengers in the passenger flow cells and the edge passage capacity of each connecting edge based on the critical density and the passage capacity, and to construct a dynamic cell transmission model using the critical number of passengers and the edge passage capacity; the dynamic cell transmission model is used to determine the state category of the passenger flow cells based on the density state information of the passenger flow cells, and to determine the passenger flow transmission capacity based on the state association relationship between adjacent passenger flow cells. Identification module: Based on the difference in passenger flow transmission capacity between the two ends of each connection edge, identify the connection edge with mismatched transmission capacity as an unbalanced edge, and aggregate the continuously distributed unbalanced edges into an unbalanced chain along the passenger flow direction, and locate the congestion initiation cell, congestion critical cell and its diffusion link through the unbalanced chain. The module obtains the state information of the passenger flow cells in the future time period based on the dynamic cell transmission model. The generation module is used to generate scheduling results based on the congestion initiation cell, the congestion critical cell, their diffusion links, and the predicted cell states for future time periods.

[0010] Thirdly, this application also provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the methods described in any of the embodiments of the present invention. Attached Figure Description

[0011] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a subway station passenger flow scheduling method provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of a subway station passenger flow scheduling system provided in an embodiment of the present invention. Detailed Implementation

[0012] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0013] See Figure 1 As shown in some embodiments of this application, this embodiment provides a subway station passenger flow scheduling method, including: Acquire spatial structure data and passenger flow data of the station, divide the station into multiple passenger flow cells and establish topological connections between the cells, and configure the functional areas of each passenger flow cell.

[0014] Subway stations typically comprise multiple functional areas, including concourses, passageways, escalators, turnstiles, and platforms. These areas are interconnected, and passenger flow continuously moves between them along specific paths. To perform detailed passenger flow modeling, the station space must first be discretized.

[0015] From the station spatial structure data pre-stored in the passenger flow scheduling server, the area boundaries, facility locations, area areas, connecting edge widths, passenger flow directions, and facility capacity boundaries of each functional zone are extracted. Specifically, the area area excludes inaccessible areas such as obstacles and fixed facilities; the connecting edge width is the actual passage width between adjacent areas; and the passenger flow direction defines the legal movement direction of passengers between adjacent areas.

[0016] Candidate passenger flow areas are determined based on regional boundaries and facility locations. These candidate areas are then divided according to facility capacity boundaries and passenger flow directions, resulting in multiple passenger flow cells. Each cell is assigned a unique cell identifier, its corresponding functional zone, area area, capacity limit, and corresponding passenger flow data collection location.

[0017] Establish the topological connections between cells: Based on the width of the connecting edges and the passenger flow direction, determine the upstream and downstream relationships between adjacent cells and generate a set of directed connecting edges; generate an adjacency matrix based on this set. The passenger flow cells, the set of directed connecting edges, and the adjacency matrix together constitute the passenger flow cell topological model, providing a structural foundation for subsequent state calculations and propagation analysis.

[0018] Based on the average density sequence and output flow sequence of each passenger flow cell, a basic macroscopic map of pedestrians in each zone is constructed to determine the critical density and traffic capacity of each functional area.

[0019] Pedestrian flow characteristics differ fundamentally across functional zones. Movement within passageways is relatively orderly, while queuing and bottlenecks are prone to occur at escalator entrances and turnstiles. Platforms are significantly affected by train arrival and departure schedules. Therefore, it is necessary to independently learn the inherent "density-flow" relationship for each functional zone.

[0020] Based on the data acquisition location corresponding to each cell, obtain the entry count, exit count, and dwell count within the continuous sampling period. Calculate the basic state parameters of each cell based on the count data: Calculate the basic state parameters of each cell. For the... Each passenger flow cell, in the t-th sampling period, is based on its entry count. , departure count Number of people in the previous sampling period Calculate the current number of people using the following formula: When the passenger flow data collection location can directly output the dwell count. At that time, the retention count can be used as The correction value is used to eliminate accumulated errors. Based on this, the current number of people... With area Calculate cell density : Based on the current number of people With capacity limit Calculate remaining capacity Simultaneously, the upstream and downstream cell sets of each cell are determined based on the topological connection relationship. Then, the input flow is determined based on the entry count and the upstream cell set, and the output flow is determined based on the exit count and the downstream cell set.

[0021] All cells are divided into five functional zones: concourse, passageway, escalator, turnstile, and platform. The density of each cell within the same zone is weighted by area to obtain the zone's average density. The output flow of cells with passenger flow pointing outwards from the zone is aggregated to obtain the zone's output flow. These are then arranged in chronological order according to sampling time periods to generate an average density sequence and an output flow sequence.

[0022] Constructing a basic macroscopic pedestrian map for each zone: Pair the average density and output flow rate of the same zone during the same time period to form density-flow rate sampling points, and group them by zone; arrange the sampling points in ascending order of density and connect them sequentially to form a density-flow rate change curve. Compare the output flow rate of adjacent sampling points along the direction of increasing density, and determine the sampling point where the flow rate changes from increasing to non-increasing as the critical sampling point. The corresponding average density is the critical density, and the corresponding output flow rate is the passage capacity. For example, when the density of a station hall zone increases from 0.68 people / m² to 1.42 people / m², the flow rate increases from 420 people / minute to 690 people / minute; after continuing to increase to 1.58 people / m², the flow rate no longer increases, then the critical density is 1.42 people / m², and the passage capacity is 690 people / minute.

[0023] Based on critical density and throughput capacity, the critical number of passengers in each passenger flow cell and the throughput capacity of each connecting edge are determined. A dynamic cell transmission model is then constructed using the critical number of passengers and the throughput capacity of each edge. The dynamic cell transmission model is used to determine the state category of a passenger flow cell based on its density state information and to determine the passenger flow transmission capacity based on the state correlation between adjacent passenger flow cells.

[0024] Transform macroscopic parameters at the partition level into microscopic parameters at the cell level. Cell critical population: Multiply the critical density of the partition to which each cell belongs by the area of ​​that cell. Edge traffic capacity: Match each connecting edge to the traffic capacity of its respective partition; connecting edges with the same traffic direction within the same partition are allocated traffic capacity according to their width proportion.

[0025] By incorporating the critical number of cells and edge throughput into the Cellular Transmission Model (CTM), a dynamic cellular transmission model is formed. This dynamism is reflected in two aspects: key parameters are learned through preprocessing steps rather than manually preset; and sending / receiving capabilities are dynamically determined based on the current state of the cell and the combination of upstream and downstream states.

[0026] The dynamic cellular transport model labels cell states based on a comparison between density and critical density: a free state is defined as a density less than the critical density and a number of people less than the critical number; a congested state is defined as a density greater than the critical density and a number of people greater than the critical number; and the rest are critical states. The critical state represents an intermediate state where there is not yet complete congestion but traffic is no longer smooth, and it is a key window for proactive intervention.

[0027] The sending / receiving capacity is determined based on the combination of upstream and downstream states: when the upstream is in a free state and the downstream is in a free or critical state, the sending capacity is taken from the current number of people in the upstream, and the receiving capacity is taken from the remaining capacity in the downstream; when the critical state is included but the congested state is not included, the sending / receiving capacity is jointly limited by the critical number of people, the edge passage capacity, and the remaining capacity; when the congested state is included, it is further strongly constrained by the zone passage capacity.

[0028] Based on the difference in passenger flow transmission capacity between the two ends of each connecting edge, connecting edges with mismatched transmission capacity are identified as unbalanced edges. The continuously distributed unbalanced edges are aggregated into unbalanced chains along the passenger flow direction. The unbalanced chains are used to locate the congestion initiation cell, the congestion critical cell, and their diffusion links.

[0029] For each directed connection edge, if the upstream sending capacity is greater than the downstream receiving capacity, it is marked as an unbalanced edge. Multiple unbalanced edges that are consecutively adjacent along the passenger flow direction form an unbalanced chain, indicating a complete diffusion path of congestion from its source to the bottleneck. Specifically, this is identified through the adjacency relationship of directed connections: along the passenger flow direction, it is determined whether the downstream cell of one unbalanced edge is the upstream cell of another unbalanced edge; if so, they are continuously connected until no further unbalanced edges exist.

[0030] The downstream cell at the end of the imbalance chain is identified as the downstream bottleneck cell, and its receiving capacity is used as the end link constraint. This constraint is sequentially transmitted to each upstream imbalance edge in the opposite direction of passenger flow, generating the link constraint traffic for each imbalance edge. The upstream cell of the first imbalance edge in the imbalance chain is identified as the congestion initiation cell (source), and cells whose density reaches or exceeds the critical density are identified as congestion critical cells (positions about to change). The directed connection path corresponding to the imbalance edge is identified as the diffusion link (propagation direction).

[0031] Based on the dynamic cell transmission model, the state of passenger flow cells in future time periods is predicted to obtain the state information of passenger flow cells in future time periods.

[0032] The scheduling result is generated based on the congestion initiation cell, the congestion critical cell, its diffusion links, and the predicted cell states for future periods.

[0033] The prediction of cell states for future time periods adopts a recursive approach: the number of people in the current time period is used as the starting value for the first prediction time period; within each time period, the sending capacity, receiving capacity, and passage capacity of each connected edge are read according to the dynamic model, and the minimum of the three is taken as the predicted transfer flow; the predicted input and output flow of each cell are summarized; the predicted number of people is updated according to the passenger flow conservation, and then the predicted density and remaining capacity are calculated; the result of the current time period is used as the input of the next time period, and so on until the states of all future time periods are obtained.

[0034] Generate scheduling results: Determine the target cell based on the congestion critical cell, the congestion initiation cell, and the diffusion link; use the future time period when the target cell first reaches the critical density as the trigger time period; determine the instruction type (gate flow restriction, channel diversion, platform guidance, escalator release control) based on the functional zone to which the target cell belongs; determine the execution object based on the cell identifier and spatial structure data; and determine the control quantity based on the output flow, receiving capacity, and link constraint flow of the target cell.

[0035] The final output scheduling result contains five elements: target cell, trigger time period, instruction type, execution object, and control variable, which can be directly sent to the execution end. For example, in a certain scenario, the system identifies the distribution cell behind the turnstile as the congestion initiation cell, and the channel entrance cell and escalator entrance cell as the congestion critical cells, with the diffusion path being turnstile → channel → escalator entrance. The system generates a scheduling result, determining the target cell based on the congestion critical cell, the congestion initiation cell, and the diffusion path. These are the core areas requiring focused control measures. The future time period when the target cell first reaches the critical density is used as the trigger time period, ensuring that the scheduling measures are initiated precisely before it enters the critical state, achieving a "just-in-time" intervention effect.

[0036] The instruction type is determined based on the functional zone to which the target cell belongs: When the target cell belongs to the turnstile zone, the instruction type is turnstile flow control, which controls the rate of passenger flow entering the station by adjusting the turnstile release ratio. When the target cell belongs to the passageway zone, the instruction type is passageway diversion, which guides some passengers to other routes through directional facilities. When the target cell belongs to the platform zone, the instruction type is platform evacuation, which accelerates passenger flow evacuation through broadcasts or personnel guidance. When the target cell belongs to the escalator zone, the instruction type is escalator release control, which avoids passenger congestion at escalator entrances by adjusting the release rhythm at escalator entrances.

[0037] The execution target is determined based on the cell identifier of the target cell and the facility location in the spatial structure data. The abstract cell number is mapped to a specific physical device, such as a numbered turnstile group, a numbered channel guide screen, or a numbered escalator control cabinet. The control variables are determined based on the output flow, receiving capacity, and link constraint flow of the target cell. For example, the turnstile release ratio is reduced to a certain percentage, the channel diversion ratio is adjusted to a certain value, and the release capacity at the escalator entrance is limited to a specific number of people.

[0038] The final output scheduling result is a complete and executable control instruction, which includes five elements: target cell (where to execute), trigger time period (when to execute), instruction type (what action to perform), execution object (what equipment to use to execute), and control quantity (to what extent to execute). It can be directly sent to the station passenger flow scheduling execution terminal for automated or semi-automated execution.

[0039] Understandably, existing technologies rely on human experience and video surveillance, allowing intervention only after congestion has formed and been observed, resulting in significant delays in scheduling. This invention addresses this by dividing the station into multiple passenger flow cells and establishing topological connections. Based on the average density sequence and output flow sequence of each functional zone, a macroscopic basic map of pedestrian traffic is constructed. Using a data-driven self-learning approach, the critical density and capacity of each zone are determined. These parameters are then used as input to a cell transmission model, enabling the model to label cells as free, critical, or congested states based on the comparison between cell density and critical density. Furthermore, the model dynamically determines sending and receiving capabilities based on the state combinations of upstream and downstream cells. Therefore, the system can proactively detect and quantify the transition from free flow to congested flow at the critical stage, eliminating the need to wait for actual congestion to form before intervention. This elevates scheduling decisions from passive response to proactive prediction, reducing the probability and severity of congestion.

[0040] Existing technologies cannot identify the origin of congestion, its propagation path, or the areas ultimately affected, limiting scheduling decisions to general manual intervention or station-level flow control. This invention compares the upstream transmitting capacity with the downstream receiving capacity of each connection edge, marking edges with transmitting capacity exceeding receiving capacity as imbalance edges. It then combines consecutively adjacent imbalance edges along the passenger flow direction into an imbalance chain, using this chain to determine the congestion initiation cell, the critical congestion cell, and the propagation path. This enables quantitative reasoning regarding the location of congestion sources, propagation direction, and scope of impact. Consequently, scheduling measures can intervene at the source of congestion, rather than passively responding only to downstream congestion, thus improving the efficiency of congestion management.

[0041] In some embodiments of this application, dividing a station into multiple passenger flow cells and establishing topological connections between these cells includes: Extract the area boundaries, facility locations, area areas, connecting edge widths, passenger flow directions, and facility capacity boundaries of each functional area from the spatial structure data.

[0042] Station spatial structure data is typically pre-stored in the database of a passenger flow scheduling server, and its sources can include architectural design drawings, geographic information data, or on-site survey data. Six key types of information are extracted from this data: area boundaries define the geometric range of each functional area; facility locations include the coordinates of fixed facilities such as turnstiles, escalators, stairs, and platform screen doors; area area needs to deduct the area occupied by obstacles such as columns, walls, and ticket machines, as using the total geometric area would underestimate the actual passenger flow density; the width of connecting edges also needs to deduct the area occupied by obstacles, which determines the maximum passenger flow per unit time; passenger flow direction defines the legal movement direction of passenger flow between adjacent areas, with the entry direction being outside the station → turnstile → concourse → passageway → escalator → platform, and the exit direction being the opposite, requiring separate modeling for bidirectional passenger flow in transfer passageways; facility capacity boundaries include the maximum throughput capacity of turnstiles, the rated transport capacity of escalators, and the maximum number of people safely waiting on the platform, providing a basis for subsequent cell capacity limit settings.

[0043] Candidate passenger flow areas are determined based on regional boundaries and facility locations, and then divided into multiple passenger flow cells according to facility capacity boundaries and passenger flow directions.

[0044] The core idea of ​​the Cellular Transport Model (CTM) is to discretize a continuous space into a series of interconnected cells, each with a unified state at any given time. Passenger flow is transmitted between adjacent cells along directed connecting edges. The determination of candidate passenger flow areas is based on preliminary functional zoning according to area boundaries and facility locations. Any continuous space with relatively consistent passenger flow patterns and no obvious internal obstacles or facility boundaries can be considered a candidate area. The station concourse level may include entrance and exit areas, security checkpoints, and ticket offices; passageways may include straight sections, turning sections, and intersection sections; platforms may include central waiting areas, end waiting areas, and platform screen door queuing areas. After determining the candidate area, it needs to be further segmented according to the following three aspects: First, facility capacity boundary, when there are multiple facilities with different capacity characteristics in the same candidate area, segment according to capacity differences; Second, passenger flow direction, when there are multiple different passenger flow directions, segment according to direction or establish multiple directional attribute layers; Third, spatial geometric features, segment when the area width, shape or obstacle distribution changes significantly, to ensure that the passage conditions inside each cell are relatively consistent.

[0045] Configure cell information for each passenger flow cell.

[0046] Cell identifiers are unique codes used for indexing and referencing, and can be in the form of "zone abbreviation + serial number". The functional zone identifier indicates whether the cell belongs to the concourse, passageway, escalator, turnstile, or platform zone, determining which zone's critical density and throughput capacity to match subsequently. The area, after being divided and allocated to each cell, is the denominator for density calculation. The capacity limit setting is based on facility capacity boundaries, the product of area area and maximum safe density, and safety regulations; it forms the basis for calculating remaining capacity. The smaller the remaining capacity, the smaller the receiving capacity. Passenger flow data collection locations correspond to physical devices such as turnstile counting equipment and video detection equipment, used to accurately map real-time collected passenger flow data to the corresponding cells.

[0047] Based on the width of the connecting edge and the direction of passenger flow, the upstream and downstream relationships between adjacent passenger flow cells are determined, and a set of directed connecting edges is generated.

[0048] Based on the passenger flow direction, for each pair of adjacent cells, determine whether passenger flow is allowed to move from one cell to another. If allowed, establish a directed connection edge from the upstream cell to the destination cell. The attributes of each directed connection edge include the starting cell, the ending cell, the edge width, the flow direction, and the subsequent allocated edge capacity. The generation of the set of directed connection edges must ensure completeness (establishing a connection between all adjacent cell pairs where passenger flow may transfer) and consistency (the direction must be consistent with the actual flow direction).

[0049] An adjacency matrix is ​​generated based on the set of directed edges, and the topological connection relationship is formed by the passenger flow cells, the set of directed edges, and the adjacency matrix.

[0050] The adjacency matrix A is an N×N square matrix. When there is a directed connection edge between the i-th passenger flow cell and the j-th passenger flow cell, ... =1. When there is no directed connection between the i-th passenger flow cell and the j-th passenger flow cell. =0. For example, if there exists a directed connection edge from cell 2 to cell 5, then =1. If there is no connecting edge from cell 3 to cell 7, then =0, the adjacency matrix stores all topological connections in a regular two-dimensional array, supports efficient matrix operations, and the connection of subway station cells only occurs between adjacent cells, which has sparsity and can be stored using an efficient sparse matrix method.

[0051] Ultimately, the passenger flow cell topology model is constructed from the passenger flow cell set, the directed connection edge set, and the adjacency matrix. This model has three functions: spatial description, accurately describing the spatial attributes and connectivity of each computational unit; computational support, quickly obtaining the upstream and downstream sets of any cell by querying the adjacency matrix, and calculating input / output flow; and propagation analysis, providing a graph theory basis for identifying imbalance chains, and finding consecutive adjacent imbalance edge sequences forming imbalance chains in the passenger flow direction through path search algorithms.

[0052] Understandably, existing technologies for passenger flow analysis typically use the total geometric area as a spatial parameter, ignoring impassable areas occupied by obstacles such as pillars, walls, and ticket machines. This leads to calculated passenger density values ​​being lower than actual values, and the risk of congestion being underestimated. This solution precisely extracts the area and connecting edge width from the spatial structure data and deducts the areas occupied by obstacles, ensuring that the physical parameters of the cells accurately match the actual passage conditions. This improves the accuracy of density calculation and reduces the risk of misjudgment of congestion due to deviations in spatial parameters.

[0053] Passenger flow patterns differ fundamentally across different functional areas within a station. Passenger flow in passageways is unidirectional, escalator entrances are prone to queuing, and platform density is affected by train arrival and departure schedules. Existing technologies typically treat all areas uniformly, failing to reflect these differences. This solution differentiates candidate areas based on three dimensions: facility capacity boundaries, passenger flow direction, and spatial geometric features. This ensures relatively uniform traffic conditions within each cell, enhancing the model's ability to characterize passenger flow features across different functional areas and reducing modeling distortion caused by area homogenization.

[0054] This solution assigns a unique cell identifier, functional zone, area, capacity limit, and passenger flow data collection location to each cell, making each cell a computable and traceable independent entity. Through the correspondence between cell identifiers and passenger flow data collection locations, real-time collected passenger flow data can be accurately mapped to the corresponding spatial units, improving data alignment accuracy and reducing state perception bias caused by the disconnect between physical space and the logical model.

[0055] This topological model uses a framework of passenger flow cell set, directed edge set, and adjacency matrix. It also features spatial description capabilities, accurately depicting the spatial attributes and connectivity of each computational unit; computational support capabilities, allowing for rapid retrieval of upstream and downstream sets for any cell by querying the adjacency matrix; and propagation analysis capabilities, providing a graph theory foundation for identifying imbalance chains by using path search algorithms to find consecutive adjacent imbalance edge sequences along the passenger flow direction. This structured framework enables subsequent passenger flow state calculations and congestion propagation analysis to be efficiently completed on a unified graph structure, improving the overall process coordination and scalability, and reducing processing overhead caused by inconsistent data formats.

[0056] In some embodiments of this application, the method further includes: obtaining the entry count, departure count, and dwell count of each cell during a continuous sampling period based on the passenger flow data collection location corresponding to each passenger flow cell.

[0057] The data collection locations for each cell are determined during the cell configuration phase. Turnstile cells primarily acquire data through turnstile counting devices, accurately recording the direction and time of each passenger's passage. Station hall, passageway, and platform cells primarily acquire data through video detection devices, using image recognition to count the number of people entering, leaving, and remaining in real time; however, accuracy may decrease in high-density scenarios. Escalator cells acquire data through escalator entrance video detection devices or load detection devices. The system collects data at fixed sampling intervals. The entry count represents the total number of people entering the cell from upstream, the exit count represents the total number of people leaving the cell downstream, and the remaining count represents the actual number of people remaining in the cell at the end of the time period; theoretically, these three values ​​satisfy a conservation relationship.

[0058] The count data for each passenger flow cell during the continuous sampling period is obtained based on the passenger flow data collection location corresponding to each passenger flow cell. The number of people in each passenger flow cell during the current time period is determined based on the count data.

[0059] However, due to the existence of detection errors, the actual data may not be completely consistent and need to be corrected in subsequent steps.

[0060] After obtaining the entry count, exit count, and dwell count for each cell, it is necessary to determine the number of people in each cell during the current sampling period. The method for determining the number of people can be flexibly selected according to the data collection conditions, and there are two parallel schemes.

[0061] The first approach uses a recursive calculation based on the principle of passenger flow conservation. The law of passenger flow conservation is one of the fundamental laws of traffic flow theory. Its core idea is that, within any given time period, the change in the number of people within a closed area equals the number of people entering the area minus the number of people leaving the area. For the i-th passenger flow cell in the t-th sampling period, the calculation is based on the entry count of that cell in the t-th sampling period. , departure count The number of people in the previous sampling period The number of people in the current time period is determined using the following formula:

[0062] The physical meaning of this calculation formula is: the number of people in the current time period equals the number of people in the previous time period plus the net inflow of people in this time period (entry minus departure). The advantage of this recursive calculation method is that it utilizes the physical constraint of passenger flow conservation, maintaining the continuity and consistency of the number of people over time. As long as the initial number of people is accurate and the errors in the entry and departure counts are within a controllable range, the recursive result can accurately reflect the actual changes in the number of people. Its potential risk lies in the fact that counting errors accumulate during the recursive process. If there is a positive deviation in the entry count or a negative deviation in the departure count for a certain time period, this deviation will continue to affect the number of people calculated for all subsequent time periods, leading to a continuous amplification of the error.

[0063] To overcome the problem of error accumulation in recursive calculations, the second approach directly uses the dwell count as a correction value for the number of people in the current time period. This is applicable when the passenger flow data collection location (such as video detection equipment) can directly output the dwell count. When this number of people remains in the area, it will be directly used as the number of people in the current time period. The value of , i.e. = The dwell count is a direct observation of the actual number of people staying in a cell by the detection equipment. It is not affected by historical recursion errors, so it can be used as an independent data source to correct the recursion results.

[0064] In practical applications, the two approaches can be combined to leverage their respective advantages. A typical implementation is to primarily use recursive calculation under normal circumstances, supplemented by dwell count. Recursive calculation provides a continuous trajectory of population changes, while dwell count is used to periodically correct accumulated errors during the recursive process. Specifically, recursive and detected values ​​can be calculated simultaneously at each sampling period. When the deviation between the two exceeds a preset threshold, the detected value is used as the correction value to reset the population count, and the corrected value becomes the starting point for subsequent recursions. Another implementation is to directly use the dwell count as the population value when the detection equipment is highly accurate and can stably output dwell counts, completely avoiding the accumulation of recursive errors. The choice can be flexibly made based on actual data acquisition conditions.

[0065] After determining the number of people in each cell at the current time period. Next, two key derived state parameters need to be calculated: cell density and remaining capacity. These two parameters are the direct basis for determining the cell's state (free, critical, or congested) and its receiving capacity. The density and remaining capacity of each passenger flow cell are determined based on the number of people in the current time period.

[0066] The formula for calculating cell density is: ,in This represents the density of the i-th passenger flow cell in the t-th sampling period. This represents the current number of people in the cell. This refers to the area of ​​the cell. Density, physically speaking, is the number of people per unit area and is a core indicator for measuring crowding levels. Compared to absolute numbers, density more objectively reflects the differences in crowding levels between cells of different areas. A cell with an area of ​​100 square meters and 100 people has a density of 1.0 people / square meter, indicating a relatively smooth flow. A cell with an area of ​​20 square meters and 40 people has a density of 2.0 people / square meter, indicating a relatively crowded state. Therefore, density is the core benchmark for judging whether a cell has reached a critical state.

[0067] The formula for calculating the remaining capacity is: ,in This represents the remaining capacity of the i-th passenger flow cell during the t-th sampling period. This represents the upper limit of the cell's capacity. This represents the current number of people in the cell. The remaining capacity physically refers to the number of people the cell can still accommodate in its current state. When the remaining capacity is positive, the cell has the capacity to accept more passengers. When the remaining capacity approaches zero or becomes negative, the cell has reached or exceeded its safe capacity limit and can no longer receive more passengers from upstream. Remaining capacity is the direct basis for determining a cell's receiving capacity. In the dynamic cell transmission model, the receiving capacity of a downstream cell is directly limited by its remaining capacity; the smaller the remaining capacity of a downstream cell, the smaller the passenger flow it can receive from upstream.

[0068] While both density and remaining capacity are calculated based on the current number of people, they reflect different dimensions of the cell's state: density reflects the degree of crowding (number of people per unit area), used to compare with the critical density to determine the cell's state. Remaining capacity reflects the capacity margin (how many more people can be accommodated), used to determine the cell's upper limit of receiving capacity. They complement each other, together forming a complete description of cell state perception. The upstream and downstream cell sets of each passenger flow cell are determined based on topological connections, and the input and output flows of each passenger flow cell are determined based on counting data.

[0069] After determining the number of people and derived state parameters in each cell, it is necessary to further determine the passenger flow exchange relationship between each cell and its neighboring cells, i.e., the input flow and output flow. The input flow and output flow describe the direction and amount of passenger flow transfer between cells, and are the basic data for subsequent flow calculations in the cellular transport model.

[0070] First, based on the set of directed edges and the adjacency matrix in the passenger flow cell topology model, the upstream cell set and downstream cell set of each cell are determined. For the i-th cell, its upstream cell set... This is the set of all cells that have directed edges pointing to the i-th cell. The set of its downstream cells. This refers to the set of all cells whose i-th cell has directed edges pointing to it. By querying the adjacency matrix, we can quickly obtain all upstream sources and all downstream destinations for any given cell.

[0071] After determining the upstream and downstream cell sets, the entry and exit counts need to be allocated according to the upstream and downstream relationships to determine the specific flow on each directed connection edge. Entry Count This represents the total number of people entering the i-th cell from all upstream cells during the t-th sampling period. To determine the specific number of people contributed by each upstream cell, it is necessary to decompose the calculation based on the sending capacity of each upstream cell, the throughput capacity of the connected edges, and the passenger flow distribution ratio. Similarly, the departure count... This represents the total number of people flowing from the i-th cell to all downstream cells during the t-th sampling period. It needs to be decomposed according to the receiving capacity of each downstream cell and the passage capacity of the connecting edge.

[0072] The input flow is determined based on the entry count and the upstream cell set: for the i-th cell, the input flow it receives from upstream cell j is equal to the number of people sent from the j-th cell to the i-th cell via the directed connection edge j→i during the t-th sampling period. This flow value is limited by the sending capacity of upstream cell j (how much passenger flow can the upstream provide), the edge throughput capacity of connection edge j→i (how much passenger flow can the connection edge handle), and the receiving capacity of downstream cell i (how much passenger flow can the downstream accept). The minimum value among these three constraints determines the actual input flow.

[0073] The output flow is determined based on the departure count and the downstream cell set: for the i-th cell, the output flow sent to the downstream cell k is equal to the number of people sent by the i-th cell to the k-th cell through the directed connection edge i→k during the t-th sampling period. This flow value is also subject to the triple constraints of the sending capacity of the upstream cell i, the edge throughput capacity of the connection edge i→k, and the receiving capacity of the downstream cell k.

[0074] Understandably, by configuring corresponding passenger flow data collection locations for each cell and integrating multiple collection methods such as gate counting devices, video detection devices, and escalator load detection devices, the stability and robustness of passenger flow data acquisition are improved, reducing dependence on a single data source and the risk of data loss due to collection blind spots. Furthermore, by simultaneously collecting entry counts, exit counts, and dwell counts within the same sampling period, the passenger flow status is characterized in three dimensions: upstream inflow, downstream outflow, and in-area dwell. This provides diversified data support for passenger flow estimation, enhances the richness and redundancy of passenger flow status description, and reduces the uncertainty caused by incomplete information. In the headcount determination stage, this solution offers two parallel approaches: recursive calculation based on the law of passenger flow conservation and direct use of dwell count as a correction value. The former utilizes the physical constraint of passenger flow conservation to maintain the continuity and consistency of headcount over time, improving the temporal coherence of headcount estimation and reducing state jumps caused by missing or abnormal data. The latter uses direct observations for independent correction, suppressing the cumulative effect of counting errors over time during the recursive process and reducing the risk of state drift caused by continuous error amplification. The two approaches can be flexibly selected or combined according to actual data collection conditions. In the composite mode of recursive calculation as the main approach and dwell count as the auxiliary approach, the continuous trajectory of headcount changes is maintained, and the deviation is controlled within a reasonable range through periodic correction, thereby improving the accuracy and reliability of headcount determination and reducing state deviations caused by a single data source or a fixed algorithm. In the derived state parameter calculation stage, this scheme simultaneously determines two dimensions: cell density and remaining capacity. Density reflects the degree of congestion per unit area and is used to determine the state category; remaining capacity reflects the remaining number of people that can be accommodated and is used to determine the upper limit of receiving capacity. These two dimensions work together to describe the cell state, improving the comprehensiveness of state perception and reducing the risk of misjudgment that may be caused by a single indicator. In the flow determination stage, this scheme determines the upstream and downstream sets of each cell based on topological connections. Incoming counts are assigned to upstream connection edges, and outgoing counts are assigned to downstream connection edges, giving the determination of input and output flow a clear spatial orientation, improving the accuracy of flow allocation, and reducing propagation description deviations caused by directional ambiguity. Simultaneously, both input and output flow are constrained by three conditions: upstream sending capacity, connection edge throughput capacity, and downstream receiving capacity. The actual flow is taken as the minimum of these three conditions. This constraint mechanism improves the physical consistency of flow calculation and reduces the risk of discrepancies between model output and actual throughput conditions. Through the coordinated processing of the above-mentioned counting and data collection, number of people estimation, state parameter calculation and traffic allocation, this scheme provides a complete and reliable cellular state description for the subsequent construction of the partitioned macroscopic basic graph and the operation of the dynamic cellular transmission model, thereby improving the stability and accuracy of the overall method.

[0075] In some embodiments of this application, the step of constructing a macroscopic basic pedestrian map of each functional zone based on the average density sequence and output flow sequence includes: Based on the functional area to which each passenger flow cell belongs, the density and output flow of each cell in each functional area are summarized to generate the average density sequence and output flow sequence of each functional area.

[0076] Pedestrian flow characteristics differ fundamentally across functional zones. The station hall, with its open space, allows for multi-directional gathering, dispersal, and turning of passenger flow, exhibiting a distinct two-dimensional characteristic. Passenger passageways, being narrow and elongated, have a relatively unidirectional flow, displaying one-dimensional linear movement. Escalators, constrained by mechanical transport capacity, exhibit intermittent and rhythmic passenger flow. Turnstiles are affected by physical spacing between equipment and passenger operation time. Platforms are influenced by train arrival and departure rhythms, resulting in periodic fluctuations in density. Therefore, a unified density-flow relationship cannot be established for all cells as a whole; instead, separate macroscopic basic pedestrian maps must be constructed for each functional zone. To this end, based on the functional zone attributes assigned to each passenger flow cell, all passenger flow cells are first divided into four sets: station hall cell set V, passageway cell set C, escalator cell set E, turnstile cell set G, and platform cell set P. The density of each cell within the same functional zone is weighted and summed according to the area of ​​the zone to obtain the average density of the zone for each sampling period.

[0077] The average density of a zone is calculated using a weighted average based on the area of ​​the zone, rather than an arithmetic average of the densities of individual cells. The physical reason for this is that different cells have different areas, and larger cells contribute more to the overall crowding level of the zone. For example, a 100-square-meter cell with a density of 0.5 people / square meter (50 people) has an arithmetic average of 1.25 people / square meter, while a 20-square-meter cell with a density of 2.0 people / square meter (40 people) has a weighted average of approximately 0.75 people / square meter. The latter is closer to the actual overall crowding level of the zone. The area-weighted method more accurately reflects the overall state of the zone, providing an objective data basis for subsequent determination of critical density.

[0078] The output flow of cells within the same functional zone that point outwards from the zone is aggregated to obtain the zone output flow for each sampling period.

[0079] The calculation of zone output flow is not simply an sum of the output flows of all cells within a zone. Instead, it requires identifying cells located at the zone boundary whose passenger flow direction points outwards – these are the zone output cells. Only the output flow of these cells represents passenger flow leaving the zone and entering other functional zones or outside the station, and should therefore be included in the zone output flow. Passenger flow transfer within a functional zone occurs as an internal redistribution between different cells within that zone and does not represent the zone's capacity to transport passenger flow outwards. Including the output flow between all internal cells would double-count the transfer of the same passenger flow between multiple cells, leading to a significant overestimation of the zone output flow and affecting the accuracy of the zone's macro-level pedestrian map. For example, in a channel zone composed of three connected channel cells, the flow from cell 1 to cell 2 is an internal transfer; only the flow from cell 3 to the escalator entrance cell (crossing the zone boundary) should be included in the zone output flow. The specific implementation method is as follows: For each functional partition, traverse all cells and determine whether there are cells in the downstream cell set that do not belong to this functional partition. If there are, the cell is determined as the partition output cell, and its output flow to the outside of the downstream partition is included in the partition output flow.

[0080] Arrange the average density and output flow of each zone in the order of sampling time periods to generate an average density sequence and an output flow sequence.

[0081] After obtaining the average density and output flow of each sampling period, the discrete data needs to be organized into a time series in chronological order. The average density and output flow of multiple consecutive sampling periods are arranged chronologically, with the two series strictly aligned in the time dimension. Serialization facilitates subsequent data pairing and curve fitting, maintains the temporal order of passenger flow changes, and ensures that the sequence data from multiple consecutive periods covers the complete density range from free flow to congested flow. In one implementation, a sliding time window (e.g., the past 30 minutes or 1 hour) can be maintained. The window slides forward over time, continuously incorporating new sampling data and discarding outdated historical data. This sliding window mechanism allows the macroscopic basic map of pedestrian traffic in each zone to adaptively track changes in passenger flow patterns. When station passenger flow patterns undergo seasonal changes, operational adjustments, or are affected by unforeseen events, the sliding window automatically discards historical data inconsistent with current characteristics, ensuring that the determination of critical density and capacity is always based on data samples most relevant to the current scenario.

[0082] Understandably, by constructing separate macro-level pedestrian maps for each functional zone, different passenger flow characteristics—such as the two-dimensional multi-directional gathering and dispersal of the station hall, the one-dimensional linear flow of passageways, the mechanical transport constraints of escalators, the equipment spacing and operation time constraints of turnstiles, and the periodic fluctuations of train arrivals and departures on platforms—can be accurately characterized within the independent density-flow relationships of each zone. This improves the modeling's adaptability to different scenarios and the accuracy of representing differentiated traffic patterns, while reducing the confusion and distortion of zone features caused by unified modeling. In calculating the average density of zones, an area-weighted average is used instead of an arithmetic average, ensuring that the contribution of larger cells to the overall congestion level of the zone is reasonably reflected. This avoids the distortion of the average zone state by outliers in small-area cell density, improving the objectivity and accuracy of density description and reducing the risk of misjudgment due to improper statistical methods. In summarizing the output flow of zones, output cells located at zone boundaries with outward-pointing passenger flow are identified, allowing for the redistribution of internal passenger flow and cross-zone flow. The system strictly distinguishes between zoned output traffic, only including the latter in the zoned output traffic count. This avoids overestimation of output traffic caused by repeated calculations of the same batch of passenger flow across multiple cells, thus improving the data quality upon which the macro-level basic map is constructed. By organizing discrete sampled data into strictly aligned average density and output traffic sequences in chronological order, the system maintains the temporal sequence information of passenger flow status changes, ensuring a clear temporal correspondence for subsequent density-traffic pairings. Furthermore, the continuous multi-period sequence data can cover the complete density range from free flow to congested flow, providing ample data support for the construction of the macro-level basic map. By introducing a sliding time window mechanism, historical data is dynamically eliminated as the window slides, while new sampled data is continuously incorporated. This allows for adaptive tracking of changes in passenger flow patterns. Seasonal fluctuations, operational adjustments, and sudden passenger flow events can all trigger data updates, ensuring that the determination of critical density and capacity is always based on data samples most relevant to the current scenario, thus improving the adaptability and timeliness of the parameters. In summary, this solution constitutes a complete chain from partitioned differential modeling, area-weighted aggregation, boundary output identification, time series organization to sliding window updates, improving the accuracy and reliability of subsequent critical parameter identification at each stage.

[0083] In some embodiments of this application, determining the critical density and traffic capacity of each functional zone includes: The average density of the same functional zone during the same sampling period is paired with the output flow rate to form density flow rate sampling points, and the sampling points are grouped according to the functional zone.

[0084] After obtaining the average density sequence and output flow sequence for each functional zone, it is necessary to establish a correspondence between the two to form raw data samples describing the "density-flow" relationship of that zone. For each functional zone, the average density and output flow are paired for the same sampling period to form a density-flow sampling point (ρft, qft). Each sampling point represents a "state snapshot" of the functional zone at a certain moment; density reflects the degree of congestion, and output flow reflects the rate of outward passenger flow in that state. As the sampling period progresses, the data collection covers a wide range from low density to high density, constituting raw data describing the basic characteristics of pedestrian flow in that zone. All sampling points are grouped by functional zone. Different zones have different spatial geometric characteristics and pedestrian movement patterns, resulting in different density-flow curves. Therefore, their respective macroscopic basic maps must be constructed separately.

[0085] The sampling points within the same group are arranged in ascending order of average density. Adjacent sampling points are connected sequentially to form density-flow variation curves. A basic macroscopic map of pedestrians in the zone is constructed with average density as the horizontal axis and output flow as the vertical axis.

[0086] The sampling points for the same functional zone are arranged in ascending order of average density to minimize the density difference between adjacent sampling points; when densities are the same, they are sorted by time period. The sorted sampling points are then connected sequentially, with average density on the horizontal axis and output flow on the vertical axis, to form a basic macroscopic pedestrian flow map for each zone. This curve reveals the basic traffic flow characteristics of the functional zone: at lower densities, pedestrian interference is minimal, and flow increases with density; flow reaches its maximum after density reaches a critical value; further increases in density cause the speed reduction effect to outweigh the flow gain from increased density, and flow begins to decrease, entering a congested state. Therefore, this curve fully describes the evolution from free flow to congested flow.

[0087] Along the direction of increasing average density, the output flow rate of adjacent sampling points is compared sequentially, and the sampling point where the output flow rate changes from increasing to non-increasing is determined as the critical sampling point.

[0088] The identification of critical sampling points is achieved by comparing the output flow of adjacent sampling points sequentially along the density-increasing direction: starting from the lowest density point, the output flow of the current sampling point is compared with the previous one. When it is detected that the output flow of a sampling point no longer increases relative to the previous sampling point (less than or equal to), and the output flow of the previous sampling point increases relative to the even earlier sampling point, then the current sampling point is determined as a critical sampling point. This point corresponds to the "peak" on the density-flow curve, the position where the flow changes from "increasing" to "non-increasing," that is, the turning point where the curve changes from rising to falling (or flattening). This judgment logic is based on the classic theory of the basic pedestrian flow graph. The maximum value point of the macroscopic basic graph is the critical point, corresponding to the state of maximizing traffic capacity, and also the critical state where congestion is about to occur.

[0089] The average density corresponding to the critical sampling point is used as the critical density, and the output flow rate corresponding to the critical sampling point is used as the passage capacity.

[0090] The average density corresponding to the critical sampling point is used as the critical density of the functional zone, which is the density threshold for transitioning from free flow to congested flow. Passenger flow is smooth when the actual density is below this value, and decreases in traffic efficiency and increases the risk of congestion when it approaches or exceeds this value. The corresponding output flow rate is used as the traffic capacity, which is the maximum passenger flow that can be transported outward per unit time. This serves as an important reference benchmark for subsequently determining the side traffic capacity and sending / receiving capacity. Through the above steps, the critical density and traffic capacity of each functional zone are determined independently. Due to differences in spatial geometry, pedestrian movement patterns, and traffic conditions, the critical parameters vary among different functional zones. These values ​​are generated from sampling point data of multiple consecutive passenger flow periods of the same type, objectively reflecting the real traffic characteristics of each zone under actual operating conditions, without relying on manual experience thresholds.

[0091] Understandably, by pairing the average density and output flow rate of the same functional zone during the same time period to form density-flow sampling points, each sampling point becomes a snapshot of the relationship between the congestion level and output rate of that zone at a specific moment, providing physically meaningful original data samples for subsequent curve fitting. By grouping the sampling points according to functional zones, zones with different spatial geometric features and motion patterns can independently construct macroscopic basic maps, improving the scene adaptability of the modeling and the accuracy of the zone characteristic representation, and reducing feature confusion and parameter deviation caused by zone mixing. By arranging the sampling points in ascending order of density and connecting them sequentially to form a density-flow change curve, the density difference between adjacent points is minimized, improving the continuity and smoothness of the curve's description of actual traffic patterns, and reducing curve distortion caused by improper data point sorting. This curve fully reveals the evolution process from free flow to congested flow; when the density is low, the flow rate increases with increasing density, reaches its maximum after reaching a critical value, and further increases in density cause the flow rate to begin to decrease, making the critical value... The identification of boundary states has clear physical meaning and theoretical basis, improving the interpretability of critical point determination and reducing the subjective arbitrariness caused by relying on empirical thresholds. By comparing the output flow of adjacent sampling points point by point along the density increasing direction, the point where the flow changes from increasing to non-increasing is accurately identified as the critical sampling point, corresponding to the peak position on the density-flow curve, that is, the state of maximizing traffic capacity and the critical state of impending congestion. This improves the accuracy and objectivity of critical position identification and reduces the risk of critical point deviation caused by subjective judgment. Using the average density corresponding to the critical sampling point as the critical density and the corresponding output flow as the traffic capacity, these two key parameters are directly derived from the actual operation data of the zone itself rather than being manually preset, improving the adaptability and scene matching of the parameters and reducing the risk of misjudgment caused by fixed thresholds not matching the actual situation. At the same time, different functional zones have different critical parameters due to differences in spatial geometry, movement patterns and traffic conditions, further improving the personalized accuracy of state determination for each zone. In summary, this solution constitutes a complete link from data pairing, group modeling, ordered arrangement, curve construction, point-by-point comparison to critical extraction, which improves the accuracy, objectivity and adaptability of critical density and traffic capacity determination.

[0092] In some embodiments of this application, determining the critical number of passengers in a passenger flow cell and the edge throughput of each connected edge based on critical density and throughput capacity includes: Multiply the critical density of the functional zone to which each cell belongs by the area of ​​that cell to obtain the critical number of people in that cell.

[0093] The critical density of a zone is an "intensity" indicator, measured in "people / square meter," reflecting the pedestrian density threshold per unit area. However, when the cellular transport model performs state marking, it needs to determine whether the current number of people within a cell has reached the critical value. This is a "total" indicator, not an "intensity" indicator. Different cells within the same functional zone have different areas; a large cell may already have a large number of people at a low density, while a small cell may still have a limited number of people at a high density. Therefore, the critical density of a zone must be multiplied by the actual area of ​​each cell to convert it into the critical number threshold for each cell. First, all passenger flow cells in the passenger flow cell topology model are traversed. For the i-th cell, the critical density of its functional zone f is obtained. Then obtain the area of ​​the region configured for that cell. The critical number of cells can be calculated using the following formula:

[0094] The physical meaning of this calculation formula is: when the number of people in a cell reaches the critical number, the density of the cell is exactly equal to the critical density of its functional zone, that is, the cell is exactly in a critical state between free flow and congested flow. The critical number is a threshold in the form of "total" quantity, which directly reflects the different capacity characteristics of the cell due to differences in area.

[0095] For example, the critical density of a passageway is 1.86 people per square meter. If a passageway cell has an area of ​​36 square meters, its critical capacity is approximately 67 people (1.86 × 36 ≈ 67). Another passageway cell has an area of ​​20 square meters, and its critical capacity is approximately 37 people (1.86 × 20 ≈ 37). These two cells belong to the same functional zone and have the same critical density, but due to their different areas, the absolute number of people they can accommodate when reaching the critical state differs. The larger cell reaches the critical state when the number of people reaches 67, while the smaller cell reaches the critical state when the number of people reaches 37. This difference directly reflects the spatial geometry and must be accurately characterized by the critical capacity.

[0096] Each connecting edge is matched to the passage capability of its respective functional partition, and connecting edges with the same passage direction within the same functional partition are grouped into a set of connecting edges in the same direction.

[0097] A zone's throughput capacity is the maximum passenger flow that the entire functional zone can output per unit time. However, the actual transmission of passenger flow between cells is accomplished through directed connections. The throughput capacity of each connection edge is allocated from the zone's throughput capacity. The functional zone to which a connection edge belongs is determined by the functional zone of its upstream cell, because the passenger flow carried by the connection edge originates from the zone of its upstream cell. For the k-th connection edge, it is matched with the throughput capacity of its functional zone to obtain the throughput capacity of that zone. As the values ​​to be assigned. After matching is completed, the connecting edges need to be categorized according to two dimensions: functional zone and direction of travel. Connecting edges with the same direction of travel within the same functional zone should be grouped into the same set of connecting edges in the same direction. The necessity of this categorization is that: the same functional zone may have multiple passenger flow exits in different directions (such as a station hall zone having three directions leading to passage A, passage B, and escalator C at the same time). Connecting edges in different directions carry passenger flows in different directions and should share the passage capacity of the zone respectively; only connecting edges in the same direction should compete for the passage capacity quota in that direction.

[0098] Based on the proportion of the width of each connecting edge to the width of the set of connecting edges in the same direction, the passage capacity is allocated to each connecting edge to obtain the edge passage capacity of each connecting edge.

[0099] After classifying the set of edges connected in the same direction, it is necessary to perform fine-grained allocation of each edge within the set. For the set of edges connected in the same direction... Let it contain m directed connecting edges, with widths of each edge being respectively... Then the total width of the set is For the k-th connecting edge, its width is... The percentage is: ; —No. The percentage of the width of the connecting edge; —No. The width of the connecting edge; —The total width of the connecting edges within the set of edges pointing in the same direction; Then, the passage capacity of this functional partition in this direction is allocated to each connecting edge according to the aforementioned width ratio. For the k-th connecting edge, its edge passage capacity is:

[0100] —The traversal capacity of the k-th connecting edge; in This refers to the accessibility of this functional area.

[0101] The physical basis of this allocation method is that, under the same traffic conditions, the passenger flow capacity of a connecting edge is directly proportional to its width; the wider the passage, the greater the passenger flow that can pass through per unit time. Therefore, allocating the passage capacity of a zone according to its width proportion is a precise mapping of the physical traffic conditions of each connecting edge. For example, in a typical embodiment, there are three directed connecting edges in the same passage zone, with widths of 2.4 meters, 1.8 meters, and 1.2 meters, respectively, for a total width of 5.4 meters. The width proportions of the three connecting edges are 44.4%, 33.3%, and 22.2%, respectively. If the passage capacity of this passage zone is 540 people per minute, then the edge passage capacities of the three connecting edges are 240 people, 180 people, and 120 people per minute, respectively. The edge passage capacity of each connecting edge is directly proportional to its width, accurately reflecting the differences in its physical traffic conditions.

[0102] Understandably, by multiplying the "intensity" index of partition critical density by the actual area of ​​each cell, and converting it into the "total" index of critical number of cells, the cell transport model can determine the state based on the absolute number of cells reaching the critical state when performing state labeling. This improves the accuracy of state judgment and reduces the risk of misjudging cells of different sizes due to area differences by using a uniform density threshold. Larger cells reach the critical value only when there are more cells, while smaller cells enter the critical state when there are fewer cells. This differentiated threshold ensures that the critical number of cells for each cell is precisely matched with its actual physical spatial scale, improving the model's ability to characterize spatial heterogeneity. Simultaneously, by distributing partition passage capacity proportionally to the width of each connecting edge, the edge of each connecting edge... The throughput capacity is directly proportional to its actual physical throughput conditions, improving the rationality of assigning capacity values ​​to connecting edges and reducing the risk of distortion caused by underestimating the capacity of wide channels and overestimating the capacity of narrow channels due to directly using uniform values ​​for partitions. By grouping connecting edges with the same throughput direction within the same functional partition into a set of connecting edges in the same direction, passenger flow exits in different directions independently share their respective throughput capacity quotas, avoiding capacity interference between different flow directions and improving the rationality of throughput capacity allocation in the directional dimension. Through a refined allocation method based on width ratio calculation, the sum of the throughput capacity of each connecting edge within the same direction set is exactly equal to the throughput capacity of that partition, ensuring that the total output capacity of the partition is not exceeded and achieving a reasonable distribution of capacity among the connecting edges, thus improving the consistency and physical rationality of model parameter configuration. This scheme constitutes a complete link from the scale transformation from intensity to total amount, from uniform to differential threshold weighting, from partition to edge capacity decomposition, and from mixed to unidirectional classification and allocation, improving the accuracy, rationality, and mapping precision of the dynamic cellular transmission model parameter configuration to physical throughput conditions.

[0103] In some embodiments of this application, when the dynamic cellular transport model is used to determine the state category of a passenger flow cell based on the density state information of the passenger flow cells, and to determine the passenger flow transport capacity based on the state association relationship between adjacent passenger flow cells, it includes: The upstream and downstream cells of each connecting edge are determined according to its passenger flow direction.

[0104] In the passenger flow cellular topology model, each directed connection edge points from its starting cell to its ending cell; the starting cell is the upstream cell, and the ending cell is the downstream cell. Passenger flow can only move from upstream cells to downstream cells along directed connections; it cannot flow in the opposite direction. This upstream-downstream relationship is determined based on the passenger flow direction in the station's spatial structure data. For example, when passenger flow is allowed to move from concourse cell A to passageway cell B, a directed connection edge A→B is established, where A is the upstream cell and B is the downstream cell.

[0105] When the cell density is less than the critical density and the number of cells is less than the critical number of cells, the cell is marked as free. When the cell density is greater than the critical density and the number of cells is greater than the critical number of cells, the cell is marked as congested. The remaining cells are marked as critical states.

[0106] After obtaining the critical number of people and edge passage capacity for each cell, the Dynamic Cellular Transport Model (CTM) begins operation. The primary feature distinguishing this model from traditional CTM is the introduction of the "state" concept for cells. Each cell is labeled as one of three states based on its current density and the comparison of the number of people with the critical value: When the density is less than the critical density and the number of people is less than the critical number, it is marked as a free state, indicating that the cell is in a smooth flow state with high efficiency and sufficient remaining capacity. When the density is greater than the critical density and the number of people is greater than the critical number, it is marked as a congested state, indicating that the cell has entered a congested state, with significantly reduced efficiency and extremely limited remaining capacity. All other situations are marked as critical states, meaning that one of the density or number of people exceeds a threshold while the other has not, or both fluctuate around the threshold. The critical state is the core innovation of this invention, distinguishing it from traditional technologies. Traditional methods only differentiate between "smooth flow" and "congested" states, failing to identify the transition phase between them. The introduction of the critical state allows the system to initiate proactive intervention during this transition phase, gaining valuable scheduling response time before congestion fully develops. The combined judgment of density and number of people has significant physical meaning: density reflects the degree of crowding per unit area and is a core indicator for judging flow efficiency. The number of people reflects whether the absolute number has reached the capacity threshold. For cells with a large area, there may be situations where the density is low but the number of people has exceeded the critical value (e.g., a cell density of 1.0 people / square meter in a 100-square-meter area but the number of people reaches 100). In this case, although the density has not reached the critical value, large-scale gatherings of people already pose a safety hazard. Using both conditions in combination can avoid misjudgment by a single indicator.

[0107] The upstream cell state and downstream cell state of each connected edge are combined to obtain the transmission state combination.

[0108] After completing the state labeling, the upstream cell state and downstream cell state of each directed connection edge are combined to form a transmission state combination (a total of 3×3=9 possibilities). The physical significance of the transmission state combination is that the transmission of passenger flow depends not only on the upstream sending capacity or the downstream receiving capacity, but also on the matching relationship between the upstream and downstream states. The transmission dynamics characteristics under different combinations are fundamentally different.

[0109] The transmitting and receiving capabilities are determined based on the combination of transmission states: When the transmission state combination is an upstream free state and a downstream free state or a critical state, the current number of people upstream is used as the sending capacity, and the remaining capacity downstream is used as the receiving capacity.

[0110] When the transmission state combination includes a critical state but not a congested state, the upstream current number of users, the critical number of users, and the side passage capacity jointly limit the upstream transmission capacity, and the downstream remaining capacity and the side passage capacity jointly limit the downstream reception capacity.

[0111] When the transmission state combination includes a congestion state, the upstream transmission capacity is jointly limited by the upstream functional zone's throughput capacity, edge throughput capacity, and the current number of upstream users, while the downstream reception capacity is jointly limited by the downstream remaining capacity, the downstream functional zone's throughput capacity, and edge throughput capacity.

[0112] Based on the combination of transmission states, three rules are used to determine the sending and receiving capabilities: The first scenario is a free state upstream and a free or critical state downstream: upstream passenger flow is smooth, and the sending capacity is determined by the current number of upstream passengers. , For sending capability, This represents the current number of people upstream.

[0113] Downstream congestion has not yet occurred; receiving capacity is based on the remaining capacity downstream. ,in, For receiving capability, This represents the upper limit of the cell's capacity. The current number of people in the cell is relatively large, and the transfer flow is determined by the edge throughput and the smaller of the two values, which is within the efficient transmission range.

[0114] The second scenario includes a critical state but not a congested state: When the upstream threshold is reached, the sending capacity is no longer simply equal to the current number of people, but is constrained by the critical number of people, the edge passage capacity, and the remaining downstream capacity. When the number of people exceeds the critical number, the excess portion represents the passenger flow that has entered a congested state, and its movement speed decreases significantly. When the downstream threshold is reached, the receiving capacity cannot simply be the full value of the remaining capacity; the limitation imposed by the edge passage capacity on the incoming flow per unit time must be considered. Both sending and receiving capacities are jointly limited by the edge passage capacity and the remaining capacity.

[0115] The third scenario includes congestion: When upstream is congested, the sending capacity is subject to a hard upper limit constraint on the partition's throughput capacity. The partition throughput capacity represents the maximum output traffic that can be maintained under congestion conditions and cannot be exceeded under any circumstances. When downstream is congested, the receiving capacity is also strongly constrained by the partition's throughput capacity. The sending capacity is jointly limited by the upstream partition's throughput capacity, the edge throughput capacity, and the current number of users upstream. The receiving capacity is jointly limited by the downstream's remaining capacity, the downstream partition's throughput capacity, and the edge throughput capacity. The partition throughput capacity, as a hard constraint, ensures that the model will not calculate transfer traffic exceeding the actual throughput capacity under congestion conditions.

[0116] Understandably, by accurately determining the upstream and downstream cell relationships based on the start and end points of directed connections, the direction of passenger flow transmission gains clear spatial directionality, improving the clarity of the transmission relationship description and reducing computational bias caused by directional ambiguity. Introducing three state markers—free state, critical state, and congested state—elevates the cell state description from a traditional binary distinction to a ternary fine division. The introduction of the critical state allows the system to identify passenger flow during the transition from free flow to congested flow, improving the ability to detect congestion precursors and reducing the risk of delayed proactive intervention due to coarse state division. By employing a dual-condition judgment of density and number of people, the state markers are simultaneously constrained in both the degree of congestion and absolute number, avoiding misjudgments caused by a single indicator in large cells where density is low but the number of people exceeds the threshold, thus improving the comprehensiveness and accuracy of state discrimination. By combining upstream and downstream states… The state combination method combines transmission states, ensuring that transmission capacity determination depends not only on the state at one end but also on the matching relationship between the states at both ends. This improves the granularity of transmission scenario identification and reduces transmission capacity misestimation caused by ignoring state correlations. Based on this, differentiated sending and receiving capacity determination rules are applied to three typical transmission scenarios: upstream free state and downstream free state or critical state; containing critical state but not congested state; and containing congested state. This ensures that the calculated transmission capacity matches the actual state. In the free state scenario, capacity is determined by the current number of passengers and remaining capacity to ensure efficient transmission. In the critical state scenario, a triple constraint of critical number of passengers, edge passage capacity, and remaining capacity is introduced to reflect the declining trend of passage efficiency. In the congested state scenario, zone passage capacity is used as a hard constraint to limit traffic output exceeding physical feasibility. This improves the scenario adaptability and physical consistency of transmission capacity calculation and reduces traffic estimation bias caused by single rules. Through the progressive processing of state labeling, combination judgment, and differentiated rules, this scheme improves the accuracy of passenger flow transmission capacity determination and its responsiveness to actual traffic conditions.

[0117] In some embodiments of this application, based on the difference in passenger flow transmission capacity between the passenger flow cells at both ends of each connecting edge, connecting edges with mismatched transmission capacity are identified as unbalanced edges, and continuously distributed unbalanced edges are aggregated into unbalanced chains along the passenger flow direction. When locating congestion initiation cells, congestion critical cells, and their diffusion links through unbalanced chains, the following steps are included: When the upstream sending capacity of a connection edge is greater than the downstream receiving capacity, the connection edge is marked as an unbalanced edge.

[0118] The dynamic cellular transmission model determines the upstream transmission capacity of each connection edge. and downstream receiving capabilities Then, compare each directed connection edge one by one. At that time, passenger flow can be transmitted smoothly, and the connection side remains in normal condition. When more passengers attempt to enter the downstream area, but there is insufficient space or passageway downstream to accommodate them, this connection is marked as an imbalance edge. An imbalance edge is a "warning signal" of congestion; congestion begins with supply-demand imbalances at key nodes and gradually accumulates and spreads. If an imbalance edge is identified and intervened in its early stages, further congestion can be prevented.

[0119] Along the passenger flow direction, consecutive unbalanced edges that are downstream cells of one unbalanced edge and upstream cells of another unbalanced edge are combined into an unbalanced chain, and the downstream cell at the end of the unbalanced chain is determined as the downstream bottleneck cell.

[0120] The receiving capacity of the downstream bottleneck cell is used as the end link constraint and is sequentially transmitted to each upstream imbalance edge in the imbalance chain in the opposite direction of passenger flow, generating the link constraint traffic of each imbalance edge.

[0121] A single imbalance edge only reflects a local bottleneck and cannot describe the congestion propagation path. In real-world scenarios, congestion often propagates downstream from the starting point of the imbalance along the passenger flow direction. To characterize the propagation relationship, consecutive adjacent imbalance edges in the passenger flow direction need to be combined into an imbalance chain. All imbalance edges are traversed, and the topological model is queried to determine if its downstream cell is also the upstream cell of another imbalance edge. If so, they are continuously connected along the passenger flow direction until no further consecutive imbalance edges exist. For example, three imbalance edges—A (behind the turnstile), B (at the entrance of the passage), C (in the middle of the passage), and D (at the escalator entrance)—are connected end-to-end, forming the imbalance chain A→B→C→D.

[0122] The downstream cell at the end of the unbalanced chain is designated as the downstream bottleneck cell, and its receiving capacity serves as the constraint on the end link. The bottleneck cell is the final point on the entire unbalanced chain that restricts passenger flow, and its insufficient receiving capacity is the root cause of all upstream unbalanced edges. The receiving capacity of the bottleneck cell determines the total flow that the entire unbalanced chain can ultimately pass through, and this constraint must be propagated forward level by level.

[0123] The specific reverse transmission method is as follows: The receiving capacity of the downstream bottleneck cell is obtained as the initial value of the end-link constraint. The process proceeds upstream from the end, traversing in the opposite direction of passenger flow. For each imbalanced edge, the currently transmitted constraint value is compared with the edge's throughput capacity. The smaller value is taken as the link constraint flow for that edge, and then this value is used as the constraint value for the next imbalanced edge upstream. This progressively smaller value transmission method ensures that the actual throughput of each link does not exceed its own physical throughput capacity, nor does it exceed the downstream bottleneck's receiving capacity.

[0124] For example, the bottleneck cell D has a receiving capacity of 315 people / minute. The flow rate to C→D (edge ​​capacity 360 people / minute) is taken as the smaller value, 315. Continuing to B→C (520 people / minute), the flow rate remains 315. The flow rate to A→B (610 people / minute) also remains 315. If the flow rate of an edge is less than the current constraint value, such as B→C only 280 people / minute, then the link constraint flow rate for that edge is 280, and it continues to be transmitted upstream at 280. Through reverse transmission, link constraint flow rates are generated for each unbalanced edge. The actual transmittable upper limit after bottleneck constraint correction ensures that the model does not output overly optimistic estimates beyond physical feasibility.

[0125] The upstream cell of the first unbalanced edge in the unbalanced chain is determined as the congestion initiation cell, the cell whose cell density reaches the critical density is determined as the congestion critical cell, and the directed connection path corresponding to each unbalanced edge in the unbalanced chain is determined as the diffusion link.

[0126] After calculating the link-constrained flow, three key pieces of information are extracted from the imbalanced chain. The congestion initiation cell is the upstream cell of the first imbalanced edge in the chain, representing the earliest point of supply-demand imbalance and the source of congestion. Limiting or diverting flow at the source is more efficient than intervention downstream. The congestion critical cell is identified by comparing the cell density in the future time period with the critical density, indicating the specific location where congestion is about to occur. The diffusion link is the directed connection path corresponding to each imbalanced edge in the chain, representing a continuous physical channel from the congestion initiation cell to the bottleneck cell in physical space. It indicates the actual spatial path of congestion propagation, providing a clear roadmap for scheduling intervention.

[0127] Understandably, by comparing the upstream sending capacity and downstream receiving capacity of each connection edge, connection edges with sending capacity greater than receiving capacity are identified as imbalance edges. This upgrades the detection of passenger flow supply and demand imbalance from relying on manual experience to automatic quantitative identification based on model calculation, improving the objectivity and timeliness of imbalance detection and reducing missed detections and delays caused by relying on visual observation and experience. By aggregating consecutive adjacent imbalance edges into imbalance chains along the passenger flow direction, the description of congestion is upgraded from a single bottleneck to a complete description of the diffusion link, revealing the continuous propagation pattern of congestion spreading from the starting position downstream, reducing intervention strategy errors caused by focusing only on local imbalances and ignoring the overall propagation law. By identifying the downstream cell at the end of the imbalance chain as the bottleneck cell and using its receiving capacity as the end link constraint, the upper limit of the entire imbalance chain is accurately located to the final limiting link, identifying the fundamental constraints of congestion and reducing ineffective interventions caused by misjudging the relationship between the source of congestion and the bottleneck. By propagating the end-link constraints step-by-step to each upstream imbalanced edge along the opposite direction of passenger flow and taking the smaller value between the capacity and constraint value of each edge as the link constraint flow, the downstream bottleneck's restriction is precisely propagated back to each upstream link. This ensures that the actual flow of each connecting edge on the entire imbalanced chain does not exceed the downstream bottleneck's capacity and its own physical capacity, improving the global consistency and physical feasibility of flow allocation and reducing the risk of optimistic estimation due to local capacity overestimation. By extracting three key pieces of information from the imbalanced chain—the congestion initiation cell, the congestion critical cell, and the diffusion link—the location of the congestion source, the specific location where congestion is about to occur, and the spatial path of congestion propagation can be accurately located simultaneously. This improves the targeting of scheduling interventions. Source location allows flow restriction measures to be applied to the location where the supply-demand imbalance first occurs, critical location allows preventive measures to be deployed in areas where congestion is about to occur, and path location allows diversion measures to be deployed step-by-step along the propagation direction, reducing the loss of intervention effectiveness due to positioning ambiguity. In summary, this solution constitutes a complete link from imbalance edge identification, imbalance chain aggregation, bottleneck location, constraint backpropagation to the extraction of three elements, which improves the completeness, accuracy and support capability for scheduling decisions in the identification of congestion diffusion links.

[0128] In some embodiments of this application, obtaining the state information of passenger flow cells in future time periods includes: The number of people in each cell during the current sampling period is taken as the starting number of the first future prediction period.

[0129] After completing the construction of the dynamic cellular transport model and the identification of imbalance chains, it is necessary to predict the passenger flow status in future periods in order to obtain information on density change trends and congestion risks in advance when congestion has not yet actually occurred or is still in a critical state.

[0130] Within each future prediction period, the upstream transmission capacity, downstream reception capacity, and edge passage capacity of each connection edge are read according to the dynamic cellular transmission model to determine the predicted transfer traffic of each connection edge.

[0131] The prediction starting point is the actual number of people in each cell during the current sampling period (period t), which is used as the number of people in the previous period for the first future prediction period (period t+1). For each directed connection edge in period t+1, the upstream sending capacity, downstream receiving capacity, and edge throughput capacity are read (the sending / receiving capacity is determined by the cell state in period t as input), and the minimum value of the three is taken as the predicted transfer flow. Based on the "barrel effect," the actual transmission flow is limited by the weakest link.

[0132] Based on the predicted transfer flow, the flow directed to the current cell is summarized to obtain the predicted input flow, and the flow directed from the current cell to the downstream cell is summarized to obtain the predicted output flow.

[0133] The predicted number of people for the current period is updated based on the number of people in the previous period, the predicted input flow, and the predicted output flow.

[0134] Cell density is determined based on the predicted number of people and the area of ​​the region, and the remaining capacity is determined based on the predicted number of people and the upper limit of the capacity.

[0135] After determining the predicted transfer flow for each connection edge, calculate the predicted input and output flow for each cell: the predicted input flow is the sum of the predicted transfer flows from all upstream cells to that cell, and the predicted output flow is the sum of the predicted transfer flows from that cell to all downstream cells. Update the predicted number of people according to the law of conservation of passenger flow: Predicted number of people = Number of people in the previous period + Predicted input flow - Predicted output flow. For period t+1, the number of people in the previous period is the actual number of people in period t. For subsequent periods, the number of people in the previous period is the updated value of the previous predicted period. Further calculate the predicted density (predicted number of people / area) and the predicted remaining capacity (capacity limit - predicted number of people). The predicted number of people, cell density, and remaining capacity for the current period are used as input data for the next future prediction period. This process is repeated sequentially along the future prediction period sequence until the cell states for all future periods are obtained.

[0136] After completing the (t+1)th time period, the predicted number of people, density, and remaining capacity are used as inputs for the (t+2)th time period. This process is repeated sequentially along the prediction time period sequence. The prediction confidence gradually decreases as the time is pushed further back, and is usually controlled within 5 to 15 sampling periods to balance foresight and reliability.

[0137] Understandably, by using the actual number of people in each cell during the current sampling period as the starting value for prediction, the initial conditions of the prediction are precisely aligned with the actual state, improving the accuracy of the prediction starting point and reducing the deviation in the prediction results caused by the initial value bias. By reading the upstream sending capacity, downstream receiving capacity, and edge throughput capacity of each connection edge during each prediction period and taking the minimum of the three as the predicted transfer flow, the determination of the actual transmission flow follows the shortest limit constraint principle, avoiding optimistic estimates caused by ignoring any limiting factor, and improving the physical rationality and reliability of the predicted flow calculation. By separately summarizing the flow pointing to the current cell and the flow pointing downstream from the current cell to obtain the prediction input and prediction output, the law of passenger flow conservation is accurately maintained during the prediction process, improving the physical consistency of the number update process. This approach reduces prediction errors caused by inconsistent traffic statistics. By determining density based on the ratio of predicted number of people to area and remaining capacity based on the difference between the upper limit of capacity and the predicted number of people, the calculation methods for predicted state parameters are completely consistent with actual state parameters, improving the comparability between predicted and actual states and reducing misjudgments caused by differences in parameter definitions. By using the prediction results of the current period as input data for the next period and iteratively extrapolating along the prediction period sequence, the dynamic evolution of passenger flow is continuously characterized, improving the prediction's ability to reflect cumulative effects over time and reducing the risk of time-series breaks caused by isolated predictions. By controlling the number of prediction periods within a reasonable range, sufficient forward-looking information is obtained to support predictions while controlling the risk of error accumulation and amplification with increasing number of extrapolations. In summary, this scheme constitutes a complete prediction chain from initial value setting, traffic extreme value constraints, conservation updates, consistent parameter calculation to time-series extrapolation, improving the accuracy, physical rationality, and support value for scheduling decisions in predicting future passenger flow.

[0138] In some embodiments of this application, when generating scheduling results based on the congestion initiation cell, the congestion critical cell, their propagation links, and the predicted future time period cell states, the following methods are included: The target cell is determined based on the congestion critical cell, the congestion initiation cell, and the diffusion link, and the future time period when the target cell first reaches the critical density is used as the trigger time period.

[0139] After identifying congestion propagation links and predicting cell states for future periods, the analysis results need to be transformed into executable scheduling instructions.

[0140] Identifying target cells is the primary task. The target cell set is selected based on a comprehensive analysis of congestion-critical cells, congestion-initiating cells, and diffusion links: congestion-initiating cells are the sources of congestion; scheduling them can reduce downstream passenger flow at its source. Congestion-critical cells are those whose density is predicted to reach or exceed a critical density within a predicted time period; scheduling them allows for intervention before congestion occurs. Key intermediate cells on the diffusion link play a role in absorbing and transmitting congestion pressure; scheduling them can cut off the diffusion link. These three dimensions respectively cover "addressing the root cause," "addressing the symptoms," and "intercepting the flow."

[0141] After identifying the target cell, its triggering time period needs to be determined. The predicted density of the target cell is iterated through in each future predicted time period, and the time period when the critical density is first reached is taken as the triggering time period, ensuring that scheduling measures are initiated before congestion occurs. The triggering time periods may differ for different target cells and must be determined separately for each.

[0142] The instruction type is determined based on the functional zone to which the target cell belongs: if it belongs to the turnstile zone, it is determined to be a turnstile flow restriction instruction; if it belongs to the passage zone, it is determined to be a passage flow diversion instruction; if it belongs to the platform zone, it is determined to be a platform evacuation instruction; and if it belongs to the escalator zone, it is determined to be an escalator release control instruction.

[0143] The instruction type is matched according to the functional zone to which the target cell belongs: the turnstile zone corresponds to the turnstile flow restriction instruction (adjusting the release ratio). The passage zone corresponds to the passage diversion instruction (guiding and diverting the flow through directional facilities). The platform zone corresponds to the platform evacuation instruction (broadcast / personnel guidance to accelerate evacuation). The escalator zone corresponds to the escalator release control instruction (adjusting the entrance release rate). Each instruction type is adapted to the physical characteristics and passenger flow patterns of the target zone.

[0144] The target cell is determined based on its cell identifier and the location of facilities in the spatial structure data.

[0145] The determination of the execution target is based on the target cell identifier and spatial structure data, mapping cell numbers to specific physical facilities (turnstiles, guide facilities, escalator control cabinets, etc.). During the cell partitioning phase, each cell is configured with a corresponding data acquisition location, which is then mapped back to a list of operable physical devices during the scheduling and execution phase.

[0146] The control variables are determined based on the target cell's output flow, receiving capability, and link constraint flow.

[0147] The control quantity is determined based on three parameters: the output flow rate of the target cell (reflecting the current transport pressure), the receiving capacity (reflecting the remaining acceptance space), and the link constraint flow rate (the "upper limit reference" for downstream bottlenecks). Control measures should not exceed the link constraint flow rate limit. The gate flow control quantity is the reduction in the release ratio (calculated based on the difference between the link constraint flow rate and the current output flow rate). The channel diversion control quantity is the diversion ratio (compared to the channel's capacity limit and the current flow rate). Escalator release control commands directly use the link constraint flow rate as the upper limit of the release quantity. The platform evacuation control quantity is the target evacuation rate value (calculated based on the difference between the predicted density and the critical density).

[0148] The final output scheduling result includes five elements: target cell, trigger period, instruction type, execution object, and control variable, which can be directly sent to the scheduling execution end for automatic or semi-automatic execution.

[0149] In summary, the beneficial effects of this invention are as follows: 1. This invention divides the station into multiple passenger flow cells and establishes topological connections. Based on the average density sequence and output flow sequence of each functional zone, it constructs a macroscopic basic map of pedestrian traffic in each zone. Using a data-driven self-learning approach, it determines the critical density and capacity of each zone. These parameters are then used as input to the cell transmission model, enabling the model to label cells as free, critical, or congested states based on the comparison between cell density and critical density. Furthermore, it dynamically determines the sending and receiving capabilities based on the state combinations of upstream and downstream cells. Therefore, it can detect and quantify the critical stage of passenger flow transitioning from free flow to congested flow in advance, eliminating the need to wait for actual congestion before intervention, thus elevating scheduling decisions from passive response to proactive prediction.

[0150] This invention compares the upstream transmitting capacity and downstream receiving capacity of each connection edge, marking connections with a transmitting capacity greater than their receiving capacity as unbalanced edges. It then combines consecutively adjacent unbalanced edges along the passenger flow direction into an unbalanced chain. This unbalanced chain determines the congestion initiation cell, the congestion critical cell, and the propagation link, enabling quantitative reasoning about the location of the congestion source, the propagation direction, and the scope of its impact. Furthermore, based on the above information and the predicted cell states for future time periods, it generates scheduling results including target cells, triggering time periods, instruction types, execution objects, and control variables. This allows scheduling strategies to be precisely mapped to specific facilities, equipment, and execution time nodes, significantly improving the precision and executability of passenger flow scheduling.

[0151] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this embodiment provides a subway station passenger flow scheduling system for applying the above-described subway station passenger flow scheduling method, including: The acquisition module is used to acquire spatial structure data and passenger flow data of the station, divide the station into multiple passenger flow cells and establish the topological connection relationship between each cell, and configure the functional area of ​​each passenger flow cell. Topology module: used to construct a basic macroscopic map of pedestrians in each zone based on the average density sequence and output flow sequence of each passenger flow cell, and to determine the critical density and traffic capacity of each functional area; The determination module is used to determine the critical number of passengers in the passenger flow cells and the edge passage capacity of each connecting edge based on the critical density and the passage capacity, and to construct a dynamic cell transmission model using the critical number of passengers and the edge passage capacity; the dynamic cell transmission model is used to determine the state category of the passenger flow cells based on the density state information of the passenger flow cells, and to determine the passenger flow transmission capacity based on the state association relationship between adjacent passenger flow cells. Identification module: Based on the difference in passenger flow transmission capacity between the two ends of each connection edge, identify the connection edge with mismatched transmission capacity as an unbalanced edge, and aggregate the continuously distributed unbalanced edges into an unbalanced chain along the passenger flow direction, and locate the congestion initiation cell, congestion critical cell and its diffusion link through the unbalanced chain. The module obtains the state information of the passenger flow cells in the future time period based on the dynamic cell transmission model. The generation module is used to generate scheduling results based on the congestion initiation cell, the congestion critical cell, their diffusion links, and the predicted cell states for future time periods.

[0152] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the subway station passenger flow scheduling method according to any of the above-described method embodiments.

[0153] It is understandable that the above-mentioned subway station passenger flow scheduling method, system and medium have the same beneficial effects, and will not be elaborated here.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for passenger flow scheduling in subway stations, characterized in that, include: Acquire spatial structure data and passenger flow data of the station, divide the station into multiple passenger flow cells and establish topological connection relationships between each cell, and configure the functional area of ​​each passenger flow cell. Based on the average density sequence and output flow sequence of each passenger flow cell, a basic macroscopic map of pedestrians in each zone is constructed to determine the critical density and traffic capacity of each functional area. Based on the critical density and the passage capacity, the critical number of passengers in the passenger flow cells and the passage capacity of each connecting edge are determined, and a dynamic cell transmission model is constructed using the critical number of passengers and the passage capacity of each edge. The dynamic cell transmission model is used to determine the state category of the passenger flow cells based on the density state information of the passenger flow cells, and to determine the passenger flow transmission capacity based on the state association relationship between adjacent passenger flow cells. Based on the difference in passenger flow transmission capacity between the two ends of each connecting edge, connecting edges with mismatched transmission capacity are identified as unbalanced edges, and continuously distributed unbalanced edges are aggregated into unbalanced chains along the passenger flow direction. The unbalanced chains are used to locate the congestion initiation cell, the congestion critical cell, and their diffusion links. Based on the dynamic cell transmission model, the state of passenger flow cells in future time periods is predicted to obtain the state information of the passenger flow cells in future time periods. The scheduling result is generated based on the congestion initiation cell, the congestion critical cell, their diffusion links, and the predicted cell states for future time periods.

2. The subway station passenger flow scheduling method according to claim 1, characterized in that, When dividing the station into multiple passenger flow cells and establishing the topological connections between these cells, the following steps are included: Extract the area boundaries, facility locations, area areas, connecting edge widths, passenger flow directions, and facility capacity boundaries of each functional area from the spatial structure data; Candidate passenger flow areas are determined based on the area boundaries and facility locations, and the candidate passenger flow areas are divided according to the facility capacity boundaries and passenger flow directions to obtain multiple passenger flow cells; Configure cell information for each of the passenger flow cells; Based on the width of the connecting edge and the direction of passenger flow, the upstream and downstream relationships between adjacent passenger flow cells are determined, and a set of directed connecting edges is generated. An adjacency matrix is ​​generated based on the set of directed edges, and the topological connection relationship is formed by the passenger flow cells, the set of directed edges, and the adjacency matrix.

3. The subway station passenger flow scheduling method according to claim 2, characterized in that, When determining the critical density and capacity of each functional zone, the following should be included: The average density of the same functional zone during the same sampling period is paired with the output flow rate to form density flow rate sampling points, and the sampling points are grouped according to the functional zone. The sampling points within the same group are arranged in ascending order of average density, and adjacent sampling points are connected sequentially to form a density-flow rate change curve. The average density is used as the horizontal axis and the output flow rate is used as the vertical axis to construct the macroscopic basic map of pedestrians in the partition. Along the direction of increasing average density, the output flow rate of adjacent sampling points is compared sequentially, and the sampling point where the output flow rate changes from increasing to non-increasing is determined as the critical sampling point. The average density corresponding to the critical sampling point is taken as the critical density, and the output flow rate corresponding to the critical sampling point is taken as the passage capacity.

4. The subway station passenger flow scheduling method according to claim 1, characterized in that, When determining the critical number of passengers in the passenger flow cell and the edge throughput of each connected edge based on the critical density and the throughput capacity, the following steps are included: Multiply the critical density of the functional zone to which each cell belongs by the area of ​​that cell to obtain the critical number of people in that cell. Match each connecting edge to the passage capacity of its respective functional zone, and group connecting edges with the same passage direction within the same functional zone into a set of connecting edges in the same direction; Based on the proportion of the width of each connecting edge to the width of the set of connecting edges in the same direction, the passage capacity is allocated to each connecting edge to obtain the edge passage capacity of each connecting edge.

5. The subway station passenger flow scheduling method according to claim 4, characterized in that, The dynamic cellular transport model is used to determine the state category of a passenger flow cell based on its density state information, and to determine passenger flow transport capacity based on the state correlation between adjacent passenger flow cells, including: Determine the upstream and downstream cells of each connecting edge based on its passenger flow direction; When the cell density is less than the critical density and the number of cells is less than the critical number of cells, the cell is marked as free; when the cell density is greater than the critical density and the number of cells is greater than the critical number of cells, the cell is marked as congested; the remaining cells are marked as critical. The upstream cell state and downstream cell state of each connected edge are combined to obtain the transmission state combination; The transmitting and receiving capabilities are determined based on the aforementioned combination of transmission states: When the transmission state combination is an upstream free state and a downstream free state or a critical state, the current number of people upstream is used as the transmission capacity, and the remaining capacity downstream is used as the reception capacity. When the transmission state combination includes a critical state but not a congested state, the upstream transmission capacity is jointly limited by the current number of upstream users, the critical number of upstream users, and the side passage capacity, and the downstream reception capacity is jointly limited by the downstream remaining capacity and the side passage capacity. When the transmission state combination includes a congestion state, the upstream transmission capacity is jointly limited by the passage capacity, edge passage capacity, and current number of people in the upstream functional zone, and the downstream reception capacity is jointly limited by the remaining capacity, passage capacity, and edge passage capacity of the downstream functional zone.

6. The subway station passenger flow scheduling method according to claim 5, characterized in that, Based on the differences in passenger flow transmission capacity between the passenger flow cells at both ends of each connecting edge, connecting edges with mismatched transmission capacity are identified as unbalanced edges. Continuously distributed unbalanced edges are aggregated into unbalanced chains along the passenger flow direction. When locating congestion initiation cells, congestion threshold cells, and their propagation links through these unbalanced chains, the following steps are taken: When the upstream transmission capacity of a connection edge is greater than the downstream reception capacity, the connection edge is marked as an unbalanced edge. Along the passenger flow direction, a series of unbalanced edges, in which the downstream cell of one unbalanced edge is the upstream cell of another unbalanced edge, are combined to form the unbalanced chain. The downstream cell at the end of the unbalanced chain is determined as the downstream bottleneck cell. The receiving capacity of the downstream bottleneck cell is used as the end link constraint and is sequentially transmitted to each upstream imbalance edge in the imbalance chain in the opposite direction to the passenger flow, generating the link constraint traffic of each imbalance edge. The upstream cell of the first unbalanced edge in the unbalanced chain is determined as the congestion initiation cell, the cell whose cell density reaches the critical density is determined as the congestion critical cell, and the directed connection path corresponding to each unbalanced edge in the unbalanced chain is determined as the diffusion link.

7. The subway station passenger flow scheduling method according to claim 1, characterized in that, When obtaining the state information of the passenger flow cells in future time periods, the following is included: The number of people in each cell during the current sampling period is taken as the starting number of the first future prediction period; Within each future prediction period, the upstream transmission capacity, downstream reception capacity, and edge passage capacity of each connection edge are read according to the dynamic cellular transmission model to determine the predicted transfer traffic of each connection edge. Based on the predicted transfer flow, the flow pointing to the current cell is summarized to obtain the predicted input flow, and the flow from the current cell to the downstream cell is summarized to obtain the predicted output flow. Update the predicted number of people for the current period based on the number of people in the previous period, the predicted input flow, and the predicted output flow; The cell density is determined based on the predicted number of people and the area of ​​the region, and the remaining capacity is determined based on the predicted number of people and the upper limit of the capacity. The predicted number of people, cell density, and remaining capacity for the current period are used as input data for the next future prediction period. This process is repeated sequentially along the future prediction period sequence until the cell states for all future periods are obtained.

8. The subway station passenger flow scheduling method according to claim 7, characterized in that, When generating scheduling results based on the congestion initiation cell, the congestion critical cell, their propagation links, and the predicted future time period cell states, the following are included: The target cell is determined based on the congestion critical cell, the congestion initiation cell, and the diffusion link, and the future time period when the target cell first reaches the critical density is used as the trigger time period. The instruction type is determined based on the functional zone to which the target cell belongs: if it belongs to the gate zone, it is determined to be a gate flow restriction instruction; if it belongs to the channel zone, it is determined to be a channel flow diversion instruction; if it belongs to the platform zone, it is determined to be a platform evacuation instruction; and if it belongs to the escalator zone, it is determined to be an escalator release control instruction. The execution target is determined based on the cell identifier of the target cell and the facility location in the spatial structure data; The control quantity is determined based on the output flow, receiving capability, and link constraint flow of the target cell.

9. A subway station passenger flow dispatching system, characterized in that, include: The acquisition module is used to acquire spatial structure data and passenger flow data of the station, divide the station into multiple passenger flow cells and establish the topological connection relationship between each cell, and configure the functional area of ​​each passenger flow cell. Topology module: used to construct a basic macroscopic map of pedestrians in each zone based on the average density sequence and output flow sequence of each passenger flow cell, and to determine the critical density and traffic capacity of each functional area; The determination module is used to determine the critical number of passengers in the passenger flow cells and the edge passage capacity of each connecting edge based on the critical density and the passage capacity, and to construct a dynamic cell transmission model using the critical number of passengers and the edge passage capacity; the dynamic cell transmission model is used to determine the state category of the passenger flow cells based on the density state information of the passenger flow cells, and to determine the passenger flow transmission capacity based on the state association relationship between adjacent passenger flow cells. Identification module: Based on the difference in passenger flow transmission capacity between the two ends of each connection edge, identify the connection edge with mismatched transmission capacity as an unbalanced edge, and aggregate the continuously distributed unbalanced edges into an unbalanced chain along the passenger flow direction, and locate the congestion initiation cell, congestion critical cell and its diffusion link through the unbalanced chain. The module obtains the state information of the passenger flow cells in the future time period based on the dynamic cell transmission model. The generation module is used to generate scheduling results based on the congestion initiation cell, the congestion critical cell, their diffusion links, and the predicted cell states for future time periods.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method according to any one of claims 1-8.

Images