Urban rail transit emergency command method based on digital twinning

Abstract

The invention belongs to the field of urban rail transit emergency command, and particularly relates to an urban rail transit emergency command method based on digital twinning, which comprises the following steps: collecting multi-source data and setting an emergency trigger condition; deploying a lightweight self-adaptive adjustment twinborn model; according to the rail transit guidance method, the problem of how to estimate the flow direction of crowds and issue an evacuation strategy under the condition that a fire occurs in a scene of urban rail transit and a camera is shielded by smoke or powered off is solved; according to the invention, the method can achieve the estimation of the flow direction of the crowd under the condition of no sight line, and gives an evacuation command accordingly.

Description

Technical Field

[0001] This invention belongs to the field of urban rail transit emergency command, specifically relating to an urban rail transit emergency command method based on digital twins. Background Technology

[0002] The Chinese patent application number is CN202010332849.4, which discloses an intelligent emergency command integration platform and method for rail transit. This invention provides an intelligent emergency command integration platform and method that establishes an integrated platform capable of providing pre-event warnings, in-event monitoring and command dispatch, post-event cause tracing for public safety incidents in rail transit, and helping to improve management processes.

[0003] While this invention addresses the issue of independent systems and poor information interoperability within the rail transit operation and management technology field, it fails to consider how to observe the actual situation and issue commands when cameras are obscured by smoke or power outages during emergency command. Smoke degrades camera images, leading to a significant underestimation of passenger flow and missed detections of critical conditions. Thermal imaging is prone to reflections and obstructions, causing false alarms and a general decrease in the confidence of emergency command commands. Underestimation of crowd density can lead to flow control measures or excessive closures causing secondary congestion. Therefore, there is an urgent need for a digital twin-based emergency command method for urban rail transit to address the problem of estimating crowd flow and issuing evacuation strategies when fires occur in urban rail transit scenarios and cameras are obscured by smoke or power outages. Summary of the Invention

[0004] This invention provides an emergency command method for urban rail transit based on digital twins, aiming to solve the following problem: how to estimate the flow of people and issue evacuation strategies when a fire occurs in an urban rail transit scenario, and when cameras are obscured by smoke or power is lost.

[0005] The technical solution adopted by this invention to solve the above-mentioned technical problems is: a digital twin-based urban rail transit emergency command method, comprising:

[0006] Collect and preprocess multi-source data; construct a camera availability scoring formula to determine if the camera is clear; set a camera availability scoring safety threshold and judge the camera availability score based on the safety threshold; construct a comprehensive linear scoring formula; set emergency trigger judgment conditions;

[0007] After an emergency is triggered, the process is executed by the station-level edge simulator, including: constructing a data-side density estimation formula and a correction formula; the edge simulator performs physical-side estimation by constructing a population conservation formula; normalizing the population density and population flow rate obtained from the data-side density estimation formula and the population conservation formula; and constructing a weighted fusion formula.

[0008] By setting an activation threshold, the risk score is compared with the activation threshold to obtain the activation seed; the risk seed is used as the center to expand into a local simulation domain; emergency plans are generated from the local simulation domain, and a multi-objective function evaluation function is constructed to score the emergency plans generated by the local simulation domain; the edge simulator decomposes the preferred plan obtained after executing the multi-objective function evaluation function formula into a standardized set of atomic instructions; the control unit determines the delivery level of each atomic instruction according to its priority and target device type.

[0009] As a preferred embodiment, the specific steps for determining whether a camera is clear using the camera availability scoring formula are as follows:

[0010] By constructing a camera availability scoring formula, the system determines whether a camera is clear, which can lead to inaccuracies and delays in security command orders. The camera availability scoring formula is as follows:

[0011] ,

[0012] It is a rating of camera usability. It is the weight of the smoke index. The smoke index for the camera. It is the ambiguity weighting coefficient. It is a normalization operation. Indicates the ambiguity term. Let L be the Laplace response variance, and L be the Laplace operator. It is the weight of the camera confidence score. It refers to the confidence level of the camera image. It is the weight of the camera's frame drop rate. It refers to the camera's frame drop rate.

[0013] As a preferred embodiment, the specific steps for constructing the comprehensive linear scoring formula are as follows:

[0014] By constructing a comprehensive linear scoring formula, the comprehensive linear scoring formula is as follows:

[0015] ,

[0016] To determine the safety status of urban rail transit stations using the comprehensive linear score, N represents the total number of indicators involved in the comprehensive linear score. M represents the weight of the F-th indicator, where M is the weight of the indicator. It is a normalization operation. Let O be the score for the F-th indicator, O be the indicator score, and F be the index of the summation symbol.

[0017] As a preferred embodiment, the specific steps for setting the emergency trigger determination conditions are as follows:

[0018] The emergency trigger conditions are as follows: if a fire occurs and the camera is powered off or obscured by smoke, and the camera's availability score is lower than the safety threshold, an emergency alarm is triggered; if the comprehensive linear score is lower than the set threshold, an emergency alarm is triggered; after the emergency alarm is triggered, in order to confirm the flow of people without cameras, activation requests are sent to multiple operators to obtain de-identified aggregated data at a grid size of 10–25m and a granularity of 30–60s; in normal mode, data from multiple operators is not retrieved.

[0019] As a preferred embodiment, the specific steps for constructing the data-side density estimation formula and the correction formula are as follows:

[0020] Based on the estimated number of people With grid area The formula for data density estimation is as follows:

[0021] ,

[0022] The estimated number of people per grid cell g, Estimate the area of ​​the population in each grid cell g. For density estimation of raster g, where g is the index of the raster cell;

[0023] The edge simulator determines the persistence confidence of each topological node by analyzing data from line-of-sight probes. and scaling factor used for density correction Furthermore, a correction update formula is constructed to correct the formula for data-side density. The correction formula is as follows:

[0024] ,

[0025] It is a data-side density estimation of node v. It is the density correction factor for node v. It is an assigned symbol, where v represents a spatial unit with physical meaning in the station topology.

[0026] As a preferred embodiment, the specific steps of the edge simulator in performing physical-side estimation by constructing a population conservation formula are as follows:

[0027] To provide effective data for the edge simulation layer, the edge simulator performs physical-side estimation by constructing a population conservation formula, which is as follows:

[0028] ,

[0029] Indicates the time step. The estimated number of people in the next node v, where P is the estimated number of people. Let TI be the discrete time step, and TI be the current time in the discrete time step. For time TI node The estimated number of people within the station, where v represents a spatially meaningful physical unit in the station topology. The inflow term represents the time step. Within, the cumulative number of people flowing into node v via all channels G pointing to node v. Let G be the set of all incoming edges pointing to node v, G be the directed edges of node v, and in be the symbol in graph theory representing the set of all directed edges terminating at v. This represents a specific edge in the set of edges belonging to node v. Indicates the time step The total number of people flowing out through all channels indicated by node v. Let v represent the set of outgoing edges of all edges pointing to node v. `out` represents a specific edge in the set, and `out` is a notation in graph theory representing the set of all directed edges originating from node `v`. This represents the flow rate of people on edge G at time TI. For the flow rate, This represents the net local source item at time TI node v, where s represents the number of people added or removed within that node.

[0030] As a preferred embodiment, the specific steps for constructing the weighted fusion formula are as follows:

[0031] The formula for weighted fusion is:

[0032] ,

[0033] The weights after fusion Here, represents the weights on the data side, and Data represents the normalized density estimate of the data side. This is a normalized physical density estimate.

[0034] As a preferred embodiment, the specific steps for generating an emergency plan from a local simulation domain are as follows:

[0035] The edge simulator generates several candidate emergency control schemes within a local simulation domain. Each candidate scheme consists of several atomic actions, including: gate timing adjustment, evacuation indicator route switching, station broadcast text release, fan speed adjustment, and one-way lockdown measures. The edge simulator performs lightweight and rapid simulation on each candidate scheme within the local domain and evaluates key performance indicators within a time window to ensure the feasibility of the scheme.

[0036] As a preferred embodiment, the specific steps for scoring the emergency response plan generated in the local simulation domain by constructing a multi-objective function evaluation function are as follows:

[0037] The edge simulator performs executability checks on candidate solutions that have undergone lightweight simulation evaluation. It scores the emergency solutions generated in the local simulation domain by constructing a multi-objective evaluation function, the formula of which is:

[0038] ,

[0039] Rate the overall merits and demerits of the proposed solutions. To restore the time weighting coefficient, For recovery time, The peak weighting coefficient for the number of stranded people. This represents the peak number of stranded individuals. The reverse flow penalty weight coefficient, As a penalty for reverse flow, For safety margin weighting coefficients, For safety margin.

[0040] As a preferred embodiment, the specific steps by which the control unit determines the delivery level of each atomic instruction based on its priority and the target device type are as follows:

[0041] The first layer prioritizes wired connections within the station, sending the data directly when the wired equipment link is uncongested; the second layer uses vehicle-mounted trunks, sending the data when the wired equipment link is congested or unavailable; the third layer uses vehicle-mounted trunks and multi-hop trunks with the operator's private network, sending the data when neither the first nor the second layer is available.

[0042] The beneficial effects of this invention are as follows:

[0043] 1. By setting emergency trigger conditions, after triggering, activation requests are sent to multiple operators to retrieve mobile terminal data from operator grids. De-identified statistical indicators are aggregated in each grid, thereby mapping operator observations into spatially granular clues of crowd density and flow without exposing personal identification information. This successfully provides a basis for evacuation decisions when there is no line of sight or cameras are unavailable.

[0044] 2. To prevent the degradation or unavailability of aggregated data from mobile operators, the edge simulator corrects the data by parsing data from line-of-sight probes, constructs a population conservation formula for physical estimation, and constructs a weighted fusion formula, so that the system can still maintain reliable situational awareness and issue evacuation instructions in a timely manner to ensure personnel safety.

[0045] 3. By adopting an executability verification and a layered robust delivery mechanism, the system ensures that instructions are safe, reachable, and rollbackable, thereby improving the success rate of evacuation execution.

[0046] Legend

[0047] Figure 1 This is a flowchart of an urban rail transit emergency command method based on digital twins.

[0048] Figure 2 This is a rendering of an emergency command method for urban rail transit based on digital twins. Detailed Implementation

[0049] To make the technical means, creative features, and achieved objectives and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention. Unless otherwise specified, the experimental methods in the following embodiments are conventional methods, and the materials and reagents used in the following embodiments are commercially available unless otherwise specified.

[0050] Example 1 Figure 1 A flowchart of an urban rail transit emergency command method based on digital twins is shown, including: collecting multi-source data and setting emergency triggering conditions; deploying a lightweight adaptive adjustment twin model; and distributing the obtained preferred solution in a hierarchical manner. The specific implementation steps are as follows:

[0051] Step 1: Collect data from multiple sources and set emergency trigger conditions;

[0052] Specifically, to collect multi-source data, edge sensing units with uninterrupted power supply and station-level edge gateways are deployed in the station hall, platforms, and passageways, and unified time synchronization using Network Time Protocol (NTP) and two-way authentication using Transport Layer Security (TLS) are implemented. The collected multi-source data includes: wireless signal monitoring data under normal fire-free conditions, pressure plate electrical signals, gate entry and exit data, train arrival and departure times monitored by the Automatic Train Supervision (ATS) system, gate control and platform screen door status, and camera quality. The collected multi-source data undergoes preprocessing, specifically denoising and normalization, followed by unified timestamps. To confirm camera clarity and enable the analysis of pedestrian flow during a fire to issue commands, a camera availability scoring formula is constructed. This score determines whether a camera's clarity is insufficient, preventing inaccuracies and delays in safety commands. The formula is as follows:

[0053] ,

[0054] It is a rating of camera usability. It is the weight of the smoke index. The smoke index for cameras is used to measure their visibility. It is the ambiguity weighting coefficient. It is a normalization operation, which is performed by converting... Transform to a dimensionless scale of [0,1], thereby making Comparable to other items on a weighted basis. Indicates the ambiguity term. Let L be the Laplace response variance. A larger Laplace response variance results in a clearer image, while a smaller variance results in heavier smoke. L is the Laplace operator. It is the weight of the camera confidence score. It refers to the confidence level of the camera image. It is the weight of the camera's frame drop rate. It is the camera's frame drop rate;

[0055] After constructing the camera availability scoring formula, a safety threshold for camera availability scoring is set, and the camera availability score is judged based on this safety threshold. The availability threshold is used to determine whether the camera can clearly observe the trend of pedestrian flow. In the event of a fire, if the camera's availability score is lower than the safety threshold, the camera can no longer track the crowd and accurately determine the direction of pedestrian flow. In this case, activation requests are sent to multiple operators to obtain de-identified aggregated data, which is then used to analyze pedestrian flow and trends. If there is no fire, but the camera's availability score is lower than the availability threshold, only staff are notified to repair and clean the camera. To enable the uploading of real-time multi-source data... According to the data, a 60-second time window is set based on the time within the urban rail station. Preprocessed multi-source data is placed within this window for time alignment, and a station grid-to-station topology mapping is established, mapping the sensing coverage grid to the topological edges to provide input for the subsequent construction of a comprehensive linear score. The camera availability score is normalized to the range [0,1] to ensure consistency with the preprocessed multi-source data. Domain experts determine the importance of each indicator based on their experience with urban rail transit accidents and assign weight coefficients. To ensure that the weights of each indicator are on the same order of magnitude, thus making the calculated comprehensive score comparable and reasonable, a comprehensive linear score formula is constructed. The formula is as follows:

[0056] ,

[0057] To determine the safety status of urban rail transit stations using the comprehensive linear score, N represents the total number of indicators involved in the comprehensive linear score. This indicates that the weights of each indicator are normalized. M represents the weight of the F-th indicator, where M is the weight of the indicator. It is a normalization operation. Let O be the score for the F-th indicator, O be the indicator score, and F be the index of the summation symbol used to iterate through all indicators.

[0058] To ensure that safety command is not disrupted during a fire, emergency trigger conditions are set. These conditions are as follows: if the camera is powered off or obscured by smoke after a fire, and its availability score falls below a safety threshold, an emergency alarm is triggered. Similarly, if the overall linear score falls below a set threshold, an emergency alarm is triggered. After an emergency alarm is triggered, in the absence of cameras, to confirm crowd movement, activation requests are sent to multiple operators, and de-identified aggregated data is acquired at a grid size of 10–25 meters and a granularity of 30–60 seconds. In normal operation, this data is not retrieved, thus complying with privacy and compliance requirements.

[0059] Specifically, the ATS (Automatic Train Protection System) collects and aggregates train arrival and departure times, operational status, track occupancy information, and passenger boarding and alighting statistics in real time, providing data support for evacuation plans. Establishing a grid-to-station topology mapping involves dividing the station floor plan into fixed-size grids, recording the spatial coordinates, area, and corresponding topological affiliation of each grid, and aligning and mapping this data to the corresponding topological edges based on sensor coverage, the overlap between the operator's grid, and the camera's field of view, thus establishing a one-to-one mapping relationship between the grid and the station topology. Topological affiliations include nodes such as the concourse, platform, passageways, entrances / exits, and staircases. The operator's grid, after an emergency is triggered, divides the mobile operator's wireless coverage area into several grids at the station scale, aggregating de-identified statistical indicators within each grid, thereby enabling... This system can map operator observations into spatially granular crowd density and flow clues without exposing personal identification information, successfully providing a basis for evacuation decisions when there is no line of sight or cameras are unavailable. It aggregates de-identified statistical indicators including: number of active terminals, number of access cells, signal strength distribution, and handover counts. The sensing coverage grid is the area covered by sensor coverage, with sensor types including: Bluetooth eavesdroppers, vibration sensors, trackside electromagnetic probes, and cameras. The importance and weighting of each indicator are based on historical traffic accident experience, assigning weights according to the indicator's impact on traffic accidents. For example, a fire might be caused by aging circuits, primarily due to aging circuits. Experience with urban rail transit accidents specifically includes: station fires, equipment or signal failures, and accidents caused by extreme weather.

[0060] Step 2: Deploy a lightweight adaptive tuning twin model;

[0061] Specifically, after data acquisition and emergency triggering conditions are set, the station-level edge simulator executes the emergency trigger. The specific steps are as follows: the edge simulator projects grid observations onto the nodes and edges of the station's physical topology based on the provided grid-topology mapping relationship; the edge simulator calculates the estimated number of people in each grid according to the number of active terminals in each mapped grid and a pre-calibrated coefficient C. Based on the estimated number of people With grid area The formula for data density estimation is as follows:

[0062] ,

[0063] The estimated number of people per grid cell g, Estimate the area of ​​the population in each grid cell g. For density estimation of raster g, where g is the index of the raster cell;

[0064] After constructing the data-side density estimation formula, in order to further observe the trend of pedestrian flow and the distribution of the crowd, the simulator calculates the discrete flux based on the intersection of adjacent grid cells. And construct the grid-level flow vector formula, the formula is:

[0065] ,

[0066] It is the data flow direction estimation vector of grid g. The estimated number of people in each raster cell, where h is the index of the neighboring raster cells of raster cell g. Let W be the set of neighbors of grid g, representing the set of all grids that are spatially connected and accessible to g. This represents the flux from grid g to grid h. Let g be the unit direction vector pointing from grid g to grid h, where g is the index of a cell in the grid.

[0067] To provide effective data for the edge simulator and to prevent data quality degradation or unavailability when mobile operators aggregate data, the edge simulator determines the retention confidence of each topology node by parsing data from line-of-sight probes. and scaling factor used for density correction A correction update formula is constructed to correct the formula for data-side density, thereby reflecting signs of being trapped or stuck. The correction formula is as follows:

[0068] ,

[0069] It is a data-side density estimation of node v. It is the density correction factor for node v. It is an assigned symbol, where v represents a spatial unit with physical meaning in the station topology;

[0070] After constructing the correction formula, when the data side is unreliable (meaning the corrected data is below the data side's safety threshold), in order to provide effective data for the edge simulation layer, the edge simulator performs physical side estimation by constructing a population conservation formula. The population conservation formula is as follows:

[0071] ,

[0072] Indicates the time step. The estimated number of people in the next node v, where P is the estimated number of people. Let TI be the discrete time step, and TI be the current time in the discrete time step. For time TI node The estimated number of people within the station, where v represents a spatially meaningful physical unit in the station topology. The inflow term represents the time step. Within, the cumulative number of people flowing into node v via all channels G pointing to node v. Let G be the set of all incoming edges pointing to node v, G be the directed edges of node v, and in be the symbol in graph theory representing the set of all directed edges terminating at v. This represents a specific edge in the set of edges belonging to node v. Indicates the time step The total number of people flowing out through all channels indicated by node v. Let v represent the set of outgoing edges of all edges pointing to node v. `out` represents a specific edge in the set, and `out` is a notation in graph theory representing the set of all directed edges originating from node `v`. This represents the flow rate of people on edge G at time TI. For the flow rate, This represents the net local source item at time TI node v, which represents the number of people added or removed within that node, and s is the net local source item.

[0073] After constructing the population conservation formula, in order to combine the data-side density estimation formula and the population conservation formula, thereby maintaining the physical rationality and robustness of the simulation when observations are unreliable, and providing a reliable and executable situational estimate for evacuation commands, the crowd density and crowd flow rate obtained from the data-side density estimation formula and the population conservation formula are normalized to the [0,1] interval, thus eliminating dimensions. Then, a weighted fusion formula is constructed, which is:

[0074] ,

[0075] The weights after fusion Here, represents the weights on the data side, and Data represents the normalized density estimate of the data side. This is a normalized physical density estimate;

[0076] After constructing the formula for the weighted fusion weights, the edge simulator reads the final density of each topology node, the flow of adjacent edges, and the time-varying density of each node from the output of the formula. The edge simulator then performs a directed combination of the final density, flow of adjacent edges, and time-varying density of each topology node according to pre-configured calibration coefficients. The specific steps of this combination are as follows: density and flow are considered as positive risk contributions, while density rise rate, detected reverse flow, and capacity overruns are considered as additional penalties. The positive risk contribution and additional penalties are combined to form a risk scale value. The risk scale value is then normalized to obtain a risk score for each node. An activation threshold is set, and the obtained risk score is compared with the activation threshold. A node is considered an activation seed when its value is greater than or equal to the activation threshold. The local simulation domain is then expanded around the activation seed. The edge simulator generates several candidate emergency control schemes within the local simulation domain. Each candidate scheme consists of several atomic actions, including: gate timing adjustment, evacuation route switching, station broadcast text dissemination, fan speed adjustment, and one-way lockdown measures. The edge simulator performs lightweight, rapid simulation on each candidate scheme within the local domain and evaluates key performance indicators within a time window to ensure the scheme's feasibility. The edge simulator performs executability verification on candidate schemes that have undergone lightweight simulation evaluation. A multi-objective evaluation function is constructed to score the emergency schemes generated in the local simulation domain. The multi-objective evaluation function formula is:

[0077] ,

[0078] The solutions are scored based on their overall merits; higher scores indicate that the solution will be given priority for selection. To restore the time weighting coefficient, Recovery time represents the time required for the congested area to return to an acceptable state after the implementation of this plan. The peak weighting coefficient for the number of stranded people. The peak number of stranded individuals represents the maximum number of stranded individuals or the highest instantaneous number of stranded individuals caused by this plan within the forecast window. The reverse flow penalty weight coefficient, The reverse flow penalty reflects the severity of the potential for reversal caused by the scheme. For safety margin weighting coefficients, For safety margin, it represents the positive term of the execution plan in the safety dimension;

[0079] Specifically, the station-level edge simulator is a local execution unit deployed within the urban rail transit system and running on edge computing devices. It receives aligned multi-source sensing data output from step one, performs data fusion simulation, risk assessment, and executability verification to select the optimal solution. Finally, it outputs atomic instructions for command and control unit execution. Simulating in a virtual environment, it experiences minimal interference from the surrounding environment and operates stably. The number of active terminals is the number of mobile phone users within a grid obtained from multiple mobile operators after an emergency trigger condition is met. The pre-calibration coefficient is determined through offline regression based on historical labeled data, by... The original observations, such as the number of active terminals and wireless signal characteristics confirmed by the business, are converted into calibration coefficients used to estimate the number of people, density, or flow rate. This allows the edge simulator to stably map the original observations to actual crowd parameters. Historical labeled data consists of manually labeled and multi-source verified data of the actual population distribution at known past time points and scenarios, such as on-site manual counting, gate measurement, and emergency drill records. The observations are then timestamped and recorded into a dataset. Offline regression uses statistical regression methods on the labeled historical dataset to output a set of calibrated transformation coefficients and bias terms, such as devices allocated by sensor type and grid, to determine the population density or flow rate. The system identifies the population coefficient and density correction factor to obtain the most robust parameters and confidence intervals, and determines the population conversion coefficient and related correction factors. To prevent degradation or unavailability of mobile operator aggregated data, the edge simulator monitors the quality score of the aggregated mobile operator data in real time. When the quality score falls below a preset threshold, the fusion weight of the operator data is reduced while the weight of the line-of-sight probes is increased, thereby maintaining the robustness of node-level density and flow estimation and ensuring the executability of emergency command decisions. The data from the line-of-sight probes represents data obtained through wireless signal monitoring. The activated seed serves as the center point of high-risk areas, and high-risk areas are used as… The local simulation domain simulates the surrounding conditions of an active seed by connecting other active seeds outward from the active seed as the center, and then generates an emergency control plan based on the surrounding conditions. The lightweight fast simulation executes several candidate emergency control plans generated by the local simulation domain, scores the executed plans, and selects the best plan based on their feasibility. Key performance indicators include: recovery time, peak number of stranded people, and backflow penalty. Backflow penalty indicates that after the emergency evacuation order is issued, the crowd moves towards the fire source instead of the evacuation route specified in the evacuation order. The penalty coefficient indicates that the evacuation plan is not feasible.

[0080] Step 3: Distribute the selected preferred solutions in a tiered manner;

[0081] Specifically, after constructing the multi-objective function evaluation formula, the edge simulator decomposes the preferred solution obtained by executing the multi-objective function evaluation formula into a standardized set of atomic instructions. Each atomic instruction includes the target device, action, triggering condition, survival time, priority, and fallback measures. The edge simulator publishes the atomic instruction set, candidate solutions, and solution scores as structured outputs to the station-level message bus for downstream command and control units to receive and execute. The control unit determines the distribution level of each atomic instruction based on its priority and target device type: the first level is wired priority within the station; the second level is wireless priority within the station; and the third level is multi-hop relay for vehicle-mounted relays and operator private networks.

[0082] Specifically, the first layer prioritizes wired connections within the station, directly issuing commands when the wired equipment link is uncongested. This involves using the station-level edge controller to directly control lower-level devices, which receive commands and directly drive them. Lower-level devices include ventilation and smoke exhaust controllers, platform screen door controllers, and access control controllers. The second layer uses onboard repeaters, issuing commands when the wired equipment link is congested or unavailable. This utilizes the station's wireless infrastructure to send commands or broadcast information to terminals for safe evacuation. Wireless terminals include displays and intercoms within the station. The third layer... When both the carrier relay and the operator's private network multi-hop relay are unavailable at the first and second layers, the train is used as a mobile relay station. Multi-hop forwarding is performed through the operator's private network as a last resort, serving as both a means to communicate evacuation routes to passengers on the train and an inter-station forwarding path. Multi-hop forwarding forwards data through multiple relay nodes, enabling communication to cross unreachable areas and thus improving message delivery rate and robustness. A relay station is a device or system that acts as an intermediate node in the communication link, receiving and forwarding data to extend coverage, cross unreachable areas, or provide backup transmission paths for terminals.

[0083] like Figure 2 The image shows the effect of an emergency command method for urban rail transit based on digital twins. The black bars represent the effect after issuing an evacuation order according to the patent cited in the background, while the gray bars represent the effect after issuing an evacuation order according to the present invention. The images show the following effects after issuing evacuation orders according to the present invention and the patent cited in the background: the peak density of the gray bars is lower than that of the black bars; the maximum congestion in the gray bars is lower than that in the black bars; the number of trapped people in the gray bars is lower than that in the black bars; and the evacuation time in the gray bars is shorter than that in the black bars. Therefore, the present invention effectively reduces the risk of congestion and congestion in the short term.

[0084] Example 2, based on Example 1, presents a digital twin-based emergency command method for urban rail transit. The specific scheme is as follows:

[0085] Step one involves multi-source data acquisition and preprocessing. Specifically, the edge sensing simulator acquires aggregated data from station gates, line-of-sight probes, and gridded operator data. The edge sensing simulator then performs noise reduction, normalization, and spatiotemporal alignment on the data, outputting standardized event packets for subsequent use.

[0086] Step two involves calculating camera availability scores and power health. When a fire indication is met, aggregated data from the operator is retrieved to obtain statistical indicators for grid aggregation and de-identification. This condition is only met when a fire indication is triggered, and a request is sent. The edge simulator activates the local simulation domain using the trigger node as a seed. Based on the projection observations output in step one, the node density and edge flux are estimated using preset calibration coefficients. Data-physical confidence weighted fusion is performed according to quality scores, and a small number of candidate control schemes are generated. Three schemes are selected, and the candidate schemes are subjected to short-step rapid simulation and executability verification within a time window to select the preferred scheme. The preferred scheme is then sent to the control unit for execution.

[0087] Step 3 involves using a layered distribution and closed-loop verification approach. The specific steps are as follows: After receiving the preferred solution, the control unit distributes the solution in a layered manner. If the local link is unreachable, it is downgraded to station wireless, train-mounted relay, or operator relay. The control unit compares the receipts and observations in a scrolling window and automatically switches to the backup solution or triggers manual intervention when the deviation exceeds the threshold. After the event ends, the operator data request is stopped and the log is archived.

[0088] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A digital twin-based urban rail transit emergency command method, characterized in that: multiple source data are collected and preprocessed; a camera availability score formula is constructed to determine whether the camera is clear; a camera availability score safety threshold is set, and the camera availability score is judged according to the safety threshold; a comprehensive linear scoring formula is constructed; an emergency trigger judgment condition is set; after the emergency trigger, a station-level edge emulator is executed, including: constructing a data-side density estimation formula and a correction formula; the edge emulator estimates the physical side by constructing a population conservation formula; the crowd density and crowd flow rate obtained by the data-side density estimation formula and the population conservation formula are normalized; a weighted fusion formula is constructed; by setting an activation threshold, the risk score and the activation threshold are compared to obtain an activation seed; the risk seed is expanded to a local simulation domain; an emergency plan is generated from the local simulation domain, and the emergency plan generated by the local simulation domain is scored by constructing a multi-objective function evaluation function; the edge emulator decomposes the preferred plan obtained by executing the multi-objective function evaluation function formula into a standardized atomic instruction set; the control unit determines the delivery level according to the priority of each atomic instruction and the target device type.

2. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of constructing the camera availability score formula to determine whether the camera is clear are: By constructing the camera availability score formula, it is determined whether the camera is clear according to the score, which leads to inaccurate and delayed safety command, the camera availability score formula is: ， is a score of camera usability, is a weight of the smoke index, is a smoke index for the camera, is a blur weight coefficient, is a normalization operation, denotes a blur term, is a Laplacian response variance, L is a Laplacian operator, is a weight of camera confidence, is a camera frame confidence, is a weight of camera frame drop rate, is a camera frame drop rate.

3. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of constructing the comprehensive linear scoring formula are: By constructing the comprehensive linear scoring formula, the comprehensive linear scoring formula is: ， to express the safety state of the urban rail station, N is the total number of indexes participating in the comprehensive linear score, representing the weight of the Fth index, M is the weight of the index, is a normalization operation, is the score of the Fth index, O is the index score, and F is the index of the summation symbol.

4. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of setting the emergency trigger judgment condition are: The emergency trigger judgment condition is: after a fire occurs and the camera is powered off or the camera is blocked by smoke, the camera availability score is lower than the safety threshold, then the emergency alarm is triggered; If the comprehensive linear score is lower than the set threshold, the emergency alarm is triggered; after the emergency alarm is triggered, in the case of no camera, in order to confirm the flow direction of the crowd, an activation request is sent to multiple operators to obtain de-identified aggregated data with a 10-25m grid and a 30-60s granularity; In normal mode, the data of multiple operators is not pulled.

5. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of constructing the data-side density estimation formula and the correction formula are: According to the estimated number of people With the grid area The data side density estimation formula is constructed, and the formula is: ， the estimated number of people for each grid g, the area of the estimated number of people for each grid g, the density estimate for grid g, g is the index of the cell of the grid; The edge emulator determines a retention confidence for each topology node by analyzing data from the non-line-of-sight probes and a scaling factor for density correction and constructs a correction update formula to correct the formula for the data side density, the correction formula being: ， is the data-side density estimate for node v, is the density correction factor for node v, is a symbol assigned, v denotes a spatial unit with physical meaning in the station topology.

6. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of the edge emulator estimating the physical side by constructing the population conservation formula are: In order to provide effective data for the edge simulation layer, the edge emulator estimates the physical side by constructing the population conservation formula, and the population conservation formula is: ， denotes the predicted number of people in node v after a time step P is the predicted number of people, is the discrete time step, TI is the current time of the discrete time, is the estimated number of people in node v at time TI, v denotes a space element with physical meaning in the station topology, is the inflow term denoting the cumulative number of people flowing into node v via all channels G pointing to node v in time step is the cumulative number of people flowing into node v via all channels G pointing to node v in time step denotes the set of all incoming edges pointing to node v, G is the directed edge of node v, in is a symbol in graph theory denoting the set of all directed edges with terminal point v, denotes a certain edge in the set of edges of node v, is the cumulative number of people flowing out of node v via all channels pointing from node v in time step is the cumulative number of people flowing out of node v via all channels pointing from node v in time step denotes the set of all outgoing edges of the edges pointing to node v, denotes a certain specific edge in the set, out is a symbol in graph theory denoting the set of all directed edges with node v as starting point, denotes the flow rate of people on edge G at time TI, is the flow rate of people, denotes the net local source term of node v at time TI, denotes the number of people added or removed in the node, s is the net local source term.

7. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of constructing the weighted fusion formula are: The calculation formula of weighted fusion is: ， is the fused weight, is the data-side weight, Data is the normalized data-side density estimate, is the normalized physical-side density estimate.

8. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps of generating an emergency plan from the local simulation domain are: The edge simulator generates several candidate emergency control schemes in the local simulation domain, each candidate scheme being composed of several atomic actions, including: adjustment of the machine beat, switching of the evacuation indicator screen route, issuance of station broadcast text, adjustment of the fan gear, and one-way containment measures; the edge simulator performs lightweight and rapid simulation on each candidate scheme in the local domain, and evaluates key performance indicators within a time window, thereby ensuring the feasibility of the scheme.

9. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps for scoring the emergency scheme generated by the local simulation domain through the construction of the multi-objective function evaluation function are: The edge simulator performs executability verification on the candidate scheme that has undergone lightweight simulation evaluation, scores the emergency scheme generated by the local simulation domain through the construction of a multi-objective evaluation function, and the formula of the multi-objective function evaluation function is: ， for a comprehensive merit of the scheme, for a recovery time weight coefficient, for a recovery time, for a peak value of the number of people staying weight coefficient, for a peak value of the number of people staying, for a reverse flow penalty weight coefficient, for a reverse flow penalty, for a safety margin weight coefficient, for a safety margin.

10. The urban rail transit emergency command method based on digital twinning according to claim 1, characterized in that: The specific steps for the control unit to determine the delivery level according to the priority of each atomic instruction and the target device type are: The first layer is wired within the station, and the wired device link is directly delivered when there is no congestion; the second layer is a vehicle-mounted relay, which is delivered when the wired device link is congested or unavailable; the third layer is a multi-hop relay between the vehicle-mounted relay and the operator's private network, which is delivered when the first and second layers are unavailable.

Images