A subway traffic operation decision adjustment system and method based on a visual large screen

Abstract

The present application relates to the technical field of intelligent transportation, in particular to a subway traffic operation decision adjustment system and method based on a visual large screen, comprising: a perception data module, a data processing hub, a visual large screen interactive platform, an intelligent instruction decision engine, and a dynamic adjustment unit; wherein the perception data module collects train positioning data, platform monitoring video streams, gate passage data, and environmental sensor data in real time; the data processing hub integrates and analyzes the data, mines potential patterns, correlation relationships, and abnormal conditions in the data; the visual large screen interactive platform presents a line network congestion heat map, equipment fault positioning, and passenger flow prediction trends in real time; the intelligent instruction decision engine is used to dynamically generate train schedule adjustment strategies and emergency relief schemes; and the dynamic adjustment unit dynamically adjusts electronic guide screens and platform broadcasts and dynamically schedules trains. Thus, the problems of abnormal response lag, unreasonable resource allocation, and low efficiency in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation, in particular to a subway traffic operation decision adjustment system and method based on a visual large screen. BACKGROUND

[0002] Under the background of accelerating global urbanization and continuous expansion of rail transit network scale, as a large-capacity public transport tool, the subway plays a core role in urban traffic hubs. With the increasing diversification of passenger travel demand and the significant increase in train operation density, the subway operation system faces multiple challenges in balancing safety, efficiency and service quality. Under this situation, it is particularly important to build an intelligent operation decision adjustment system based on a visual large screen. This system can monitor the subway line network operation state in real time and conduct multi-dimensional data analysis, and rely on intelligent algorithms to realize rapid response to abnormal events and dynamic scheduling optimization, to meet the core needs of improving subway operation efficiency and reducing operation cost.

[0003] However, the current traditional subway operation decision mainly relies on manual experience or single data source analysis, which not only leads to delayed response to abnormal events due to low data processing efficiency, and cannot adapt to complex scenarios such as peak passenger flow mutation or equipment sudden failure, but also causes unreasonable allocation of operation resources and difficulty in guaranteeing train punctuality. At the same time, the traditional decision system is limited by the significant "data island" phenomenon, and it is difficult to link and integrate multi-source heterogeneous data such as train operation state, passenger flow distribution, equipment parameters and scheduling instruction execution time sequence, and conduct in-depth correlation analysis. In addition, the simplification of decision logic makes the system fall into a decision blind area in scenarios such as large passenger flow congestion, interval equipment failure or bad weather, resulting in low efficiency of delay event handling, and negative impact on passenger travel experience, operation safety and economic benefits. SUMMARY

[0004] The present application provides a subway traffic operation decision adjustment system and method based on a visual large screen to solve the problems of delayed abnormal response, unreasonable resource allocation and low efficiency in the prior art.

[0005] The first aspect embodiment of the application provides a subway traffic operation decision adjustment system based on a visual large screen, comprising: a perception data module, a data smelting hub, a visual large screen interactive platform, an intelligent instruction decision engine, and a dynamic adjustment unit; wherein the perception data module is configured to collect train positioning data, platform monitoring video stream, gate passage data, and environmental sensor data in real time; the data smelting hub is configured to integrate and analyze the collected data, and mine potential patterns, correlation relationships, and abnormal conditions in the data; the visual large screen interactive platform is configured to present a line network congestion heat map, device fault positioning, and passenger flow prediction trend in real time, and switch interfaces and highlight information according to different operation scenarios and needs; the intelligent instruction decision engine is configured to dynamically generate train schedule adjustment strategies and emergency evacuation schemes; and the dynamic adjustment unit is configured to dynamically adjust electronic guide screens and platform broadcasts according to decision instructions, and dynamically schedule trains.

[0006] Preferably, the perception data module comprises a sensor cluster and an abnormal data filtering unit, wherein the sensor cluster is configured to collect train position, passenger flow density, human body temperature, and obstacle distance data; and the abnormal data filtering unit is configured to use a sliding window outlier detection algorithm to remove invalid data caused by sensor drift, shielding, or communication interruption in real time.

[0007] Preferably, the data smelting hub comprises a streaming computing pipeline, a graph neural network analysis module, and a graph reasoning engine, wherein the streaming computing pipeline is configured to aggregate and convert video streams, sensor data, and operation logs in real time; the graph neural network analysis module is configured to construct a dynamic relationship topology and mine potential congestion transmission paths; and the graph reasoning engine is configured to fuse historical fault libraries and real-time working condition data, automatically generate a congestion cause reasoning tree and a risk probability matrix.

[0008] Preferably, the visual large screen interactive platform comprises a multi-level alarm projection module, an AR perspective rendering engine, and a scene adaptive engine, wherein the multi-level alarm projection module is configured to map device faults to a line network three-dimensional sand table according to influence ranges, and mark core node faults in red; the AR perspective rendering engine is configured to generate a virtual perspective view of real-time positioning of trains in tunnels, and superimpose a full load rate heat sign; and the scene adaptive engine is configured to automatically switch interface layouts of a commuting peak mode, an emergency evacuation mode, and a night maintenance mode through an operation scenario classifier.

[0009] Preferably, the intelligent instruction decision engine comprises a road network prediction model, a resource constraint solver and a contingency optimizer, wherein the road network prediction model is configured to fuse LSTM neural network and cellular automata algorithm to deduce the diffusion range and duration of a delay event in the road network; the resource constraint solver is configured to generate an optimal additional operation / stop operation scheme under multiple restrictions of train turnaround time, number of backup vehicles and crew configuration; and the contingency optimizer is configured to fuse an expert rule base to generate an emergency broadcast script and a diversion path of a hierarchical response.

[0010] Preferably, the dynamic adjustment unit comprises an interlocking controller, an optical flow guide matrix and a dispatching instruction compiler, wherein the interlocking controller is configured to trigger multi-language electronic voice and AR navigation arrows; the optical flow guide matrix is configured to generate an evacuation path light belt through programmable LED floor tiles; and the dispatching instruction compiler encodes the strategy into an executable instruction set of an ATO train automatic dispatching unit to dynamically adjust the ATO driving curve.

[0011] The second aspect embodiment of the present application provides a subway traffic operation decision adjustment method based on a visual large screen, comprising: acquiring train positioning data, platform monitoring video stream, gate passage data and environmental sensor data; constructing a real-time line network state model according to the train positioning data, platform monitoring video stream, gate passage data and environmental sensor data; based on the real-time line network state model, fusing LSTM neural network and cellular automata algorithm to identify abnormal events, and simultaneously dynamically rendering a line network congestion heat map, a device fault positioning marker and a passenger flow prediction trend curve on a visual large screen; according to the abnormal events and passenger flow prediction trend, combining a slime mold algorithm to dynamically generate a train schedule adjustment strategy and an emergency evacuation scheme; according to the adjustment strategy and the evacuation scheme, dynamically adjusting an ATO train automatic dispatching curve, simultaneously performing multi-language broadcast and updating electronic guide screen content.

[0012] The third aspect embodiment of the present application provides an electronic device, comprising: a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a subway traffic operation decision adjustment method based on a visual large screen as described in the above embodiments.

[0013] The fourth aspect embodiment of the present application provides a computer readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement a subway traffic operation decision adjustment method based on a visual large screen as described in the above embodiments.

[0014] The fifth aspect embodiment of the present application provides a computer program product comprising a computer program or instructions for implementing a subway traffic operation decision adjustment method based on a visual large screen as described in the above embodiments.

[0015] Thus, the present application includes the following advantages:

[0016] The embodiment of the present application collects train position and passenger flow density key data in real time through the perception data module, and cooperates with the abnormal data filtering unit to remove invalid information, so as to ensure the accuracy and reliability of the data. The data processing hub uses the stream computing pipeline, the graph neural network analysis module and the graph reasoning engine to perform real-time aggregation, deep mining and intelligent reasoning on massive data, accurately locates potential congestion paths and fault causes, and uses the multi-level alarm projection, AR perspective rendering and scene adaptive engine to convert complex operation data into intuitive three-dimensional sand table and virtual perspective view, and automatically switches the interface layout according to different scenes, so as to help the operation personnel quickly master the overall situation. The intelligent instruction decision engine combines the LSTM neural network and the cellular automaton algorithm to deduce the influence of the late event diffusion, generates an optimal train scheduling scheme and a hierarchical emergency response plan under resource constraints, and quickly converts the decision instruction into multi-language broadcast, AR navigation and LED floor tile guidance, dynamically adjusts the ATO train scheduling curve, improves the subway operation efficiency and emergency disposal capacity, and reduces the operation risk and cost. Thus, the problems of abnormal response lag, unreasonable resource allocation and low efficiency in the prior art are solved.

[0017] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, including the accompanying drawings, wherein:

[0019] Figure 1 FIG. 1 is a structural schematic diagram of a subway traffic operation decision adjustment system based on a visual large screen according to an embodiment of the present application;

[0020] Figure 2 FIG. 2 is a schematic diagram of a perception data module according to an embodiment of the present application;

[0021] Figure 3 FIG. 3 is a schematic diagram of a subway early morning peak operation scene according to an embodiment of the present application;

[0022] Figure 4 FIG. 4 is a schematic diagram of a data processing hub according to an embodiment of the present application;

[0023] Figure 5 FIG. 5 is a schematic diagram of a visual large screen interactive platform according to an embodiment of the present application;

[0024] Figure 6A diagram of a weekday subway evening peak period according to an embodiment of the present application;

[0025] Figure 7 A diagram of an intelligent instruction decision engine according to an embodiment of the present application;

[0026] Figure 8 A diagram of a dynamic adjustment unit according to an embodiment of the present application;

[0027] Figure 9 A diagram of a subway traffic operation decision adjustment system based on a visual large screen according to an embodiment of the present application;

[0028] Figure 10 A flowchart of a subway traffic operation decision adjustment method based on a visual large screen according to an embodiment of the present application;

[0029] Figure 11 A diagram of a weekday subway morning peak period in a certain city according to an embodiment of the present application;

[0030] Figure 12 A diagram of a subway traffic operation decision adjustment method based on a visual large screen according to an embodiment of the present application;

[0031] Figure 13 A structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0033] With reference to the accompanying drawings, an embodiment of the present application is described below. A subway traffic operation decision adjustment system based on a visual large screen is provided. In view of the problem of abnormal response lag mentioned in the background art, the present application provides a subway traffic operation decision adjustment system based on a visual large screen. In the system, the perception data module collects train position data, passenger flow density key data in real time, cooperates with the abnormal data filtering unit to remove invalid information, and ensures the accuracy and reliability of the data. The data processing hub uses a stream computing pipeline, a graph neural network analysis module, and a graph reasoning engine to perform real-time aggregation, deep mining, and intelligent reasoning on massive data, accurately locates potential congestion paths and fault causes, and uses a multi-level alarm projection, AR perspective rendering, and a scene adaptive engine to convert complex operation data into intuitive three-dimensional sand tables and virtual perspective views, and automatically switches the interface layout according to different scenes to assist operation personnel in quickly grasping the overall situation. The intelligent instruction decision engine combines LSTM neural networks and cellular automata algorithms to deduce the influence of late events and generate optimal train scheduling schemes and hierarchical emergency response plans under resource constraints. The dynamic adjustment unit quickly converts decision instructions into multi-language broadcasts, AR navigation, and LED tile guidance, dynamically adjusts the ATO train scheduling curve, improves subway operation efficiency and emergency response capabilities, and reduces operation risks and costs. Thus, the problems of abnormal response lag, unreasonable resource allocation, and low efficiency in the prior art are solved.

[0034] Figure 1 A structure diagram of a subway traffic operation decision adjustment system based on a visual large screen provided by an embodiment of the present application is shown.

[0035] An embodiment of the present application provides a subway traffic operation decision adjustment system based on a visual large screen. The system 10 comprises:

[0036] A perception data module 100, a data processing hub 200, a visual large screen interactive platform 300, an intelligent instruction decision engine 400, and a dynamic adjustment unit 500.

[0037] The perception data module 100 is configured to collect train positioning data, platform monitoring video streams, gate passage data, and environmental sensor data in real time. The data processing hub 200 is configured to integrate and analyze the collected data, mine potential patterns, correlation relationships, and abnormal situations in the data. The visual large screen interactive platform 300 is configured to present a line network congestion heat map, device fault positioning, and passenger flow prediction trends in real time, and switch interfaces and highlight information according to different operation scenarios and needs. The intelligent instruction decision engine 400 is configured to dynamically generate train schedule adjustment strategies and emergency relief plans. The dynamic adjustment unit 500 is configured to dynamically adjust electronic guide screens and platform broadcasts according to decision instructions, and dynamically schedules trains.

[0038] It can be understood that, in the embodiments of the present application, the train position, passenger flow density key data are collected in real time by the perception data module, invalid information is removed by the abnormal data filtering unit to ensure the accuracy and reliability of the data; the data processing hub uses the streaming computing pipeline, the graph neural network analysis module and the graph reasoning engine to perform real-time aggregation, deep mining and intelligent reasoning on massive data, accurately locates potential congestion paths and fault causes; the visual large screen interaction platform uses multi-level alarm projection, AR perspective rendering and scene adaptive engine to convert complex operation data into intuitive three-dimensional sand table, virtual perspective view, and automatically switches the interface layout according to different scenes to help operation personnel quickly grasp the overall situation; the intelligent instruction decision engine combines LSTM neural network and cellular automata algorithm to deduce the influence of late event diffusion and generate the optimal train scheduling scheme and hierarchical emergency response plan under resource constraints; the dynamic adjustment unit quickly converts the decision instruction into multi-language broadcast, AR navigation, LED floor tile guidance, dynamically adjusts the ATO train scheduling curve, improves the subway operation efficiency and emergency disposal capacity, and reduces the operation risk and cost. Thus, the problems of abnormal response lag, unreasonable resource allocation and low efficiency in the prior art are solved.

[0039] In the embodiments of the present application, the perception data module 100 includes: as shown in Figure 2 a sensor cluster, an abnormal data filtering unit.

[0040] The sensor cluster is used to collect train position, passenger flow density, human body temperature and obstacle distance data; the abnormal data filtering unit uses a sliding window outlier detection algorithm to remove invalid data caused by sensor drift, shielding or communication interruption in real time.

[0041] It can be understood that, in the embodiments of the present application, the sensor cluster collects train position and passenger flow density data to provide rich raw information for subway operation decision and comprehensively perceive the operation environment; the abnormal data filtering unit uses a sliding window outlier detection algorithm to identify and remove invalid data caused by sensor drift, shielding or communication interruption in time, improve the accuracy and reliability of the data, avoid invalid data interference in subsequent analysis and decision, and improve the stability and decision accuracy of subway traffic operation.

[0042] It should be noted that the sliding window outlier detection algorithm is a real-time data anomaly recognition algorithm based on a sliding data window of a set size, which calculates the statistical quantity in the window and determines whether the current data is an outlier according to the statistical quantity. The formula of the sliding window outlier detection algorithm is:

[0043]

[0044]

[0045]

[0046] in, Let be the set of sliding window data at time t; To adjust the sliding window size; The data value at time t; This represents the average of the data within the window. Let be the sliding window data set at time t. Standard deviation of internal data; Let be the i-th observation in the time series.

[0047] For example, such as Figure 3 As shown, in a subway operation scenario, taking a busy station during the morning rush hour as an example, the sensor cluster collected 360 sets of passenger flow density data between 8:00 and 8:30. The outlier filtering unit used a sliding window outlier detection algorithm, setting the sliding window size to 30 sets of data (i.e., covering the data from the previous 150 seconds). As time progressed, the window continuously slid and updated. At 8:10, the algorithm calculated the average passenger flow density within the sliding window at that moment to be 5 people per square meter, with a standard deviation of 0.8 people. At this time, newly collected data showed a passenger flow density of 9 people per square meter, a deviation from the average that far exceeded the set threshold (usually set to mean ± 2 times standard deviation). The algorithm quickly determined this data to be outlier data, i.e., invalid data caused by temporary sensor obstruction or malfunction, and removed it. Statistics show that during this morning rush hour, the algorithm accurately removed approximately 5% of the outlier data in this way, ensuring the accuracy of subsequent passenger flow analysis, providing a solid data foundation for operational decisions, and contributing to the efficient and stable operation of the subway.

[0048] In this embodiment of the application, the data processing hub 200 includes: Figure 4 As shown, the streaming computing pipeline, graph neural network analysis module, and graph inference engine are included.

[0049] Among them, the streaming computing pipeline is used to aggregate and transform video streams, sensor data and operation logs in real time; the graph neural network analysis module is used to build dynamic relationship topology and explore potential congestion transmission paths; the graph inference engine is used to integrate historical fault databases and real-time operating condition data to automatically generate congestion cause inference trees and risk probability matrices.

[0050] It can be understood that the embodiments of the application aggregate and convert video streams, sensor data and operation logs in real time through a streaming computing pipeline, quickly integrate multi-source heterogeneous data, and provide a unified and standardized data format; the graph neural network analysis module deeply mines the potential relationship among stations, trains and passenger flow by constructing a dynamic relationship topology, accurately predicts the congestion transmission path, and identifies operation risks in advance; the graph reasoning engine fuses the historical fault library with real-time working condition data, automatically generates a congestion cause reasoning tree and a risk probability matrix, clearly presents the fault logical relationship and risk possibility, improves the processing efficiency and analysis depth of metro operation data, and assists operation personnel in quickly locating the problem source, predicting risks, enhancing the safety and stability of metro operation.

[0051] It should be noted that real-time aggregation and conversion of video streams, sensor data and operation logs first access video streams through the RTSP protocol, collect sensor data using the MQTT protocol, pull operation logs using Flume, clean the data using a sliding window outlier detection algorithm, and convert them into a PROTOBUF / JSON format, complete standardization, and then associate and aggregate the data based on the space-time dimension, and convert the aggregated data into structured data that can be directly used for analysis.

[0052] The RTSP protocol is an application layer protocol used to control the real-time transmission of audio and video streams; the MQTT protocol is a lightweight instant messaging protocol; Flume is a distributed log collection system used to collect, aggregate and transmit log data from multiple data sources to a storage system.

[0053] To build a dynamic relationship topology and mine potential congestion transmission paths, first, the subway system is abstracted into a basic graph structure composed of stations (nodes) and lines (edges), and real-time attributes of passenger flow and train operation status are assigned to nodes and edges; then, through a space-time association algorithm, real-time data in different time windows are mapped into the graph structure to form a dynamic graph that evolves over time; then, through graph embedding, high-dimensional features of nodes and edges are compressed into low-dimensional vectors to capture the inherent association; the dynamic graph is iteratively updated and reasoned to identify hidden dependency relationships between nodes, and a relationship topology that reflects dynamic characteristics such as passenger flow propagation and fault diffusion is constructed.

[0054] The space-time association algorithm formula is:

[0055]

[0056] wherein, is the association weight between station i and station j at time t; , is the weight coefficient; is the spatial association term, which measures the association degree of nodes i and j in space; For time-related items, it reflects the association of nodes i and j in the time dimension at time t.

[0057] Fusion history fault library and real-time working condition data, automatically generate congestion cause reasoning tree and risk probability matrix, through the fusion of historical fault library and real-time working condition data, build knowledge graph containing entities and relationships, then use causal chain reasoning rules, from abnormal events to forward and backward reasoning, combined with the weight of evidence calculation to generate congestion cause reasoning tree, and according to the historical statistics and real-time data to quantify the occurrence probability of each cause and influence, form risk probability matrix.

[0058] Causal chain reasoning rule is a kind of logical deduction criterion based on event causal relationship, through defining the association rule of "if a certain condition is established, then the subsequent results are triggered", from the initial abnormal point to deduce its possible influence and root cause, to build the reasoning logic of fault propagation chain.

[0059] The formula of evidence weight calculation is:

[0060]

[0061] Among them, The weight of evidence E to hypothesis H; H is not true; The probability of evidence E appearing when H is true; The probability of evidence E appearing when H is not true; Logarithmic function.

[0062] Risk probability matrix formula:

[0063]

[0064] Among them, The historical use weight ratio; The environmental impact weight ratio; , , The animal / plant / human garbage influence coefficient; The comprehensive risk index; , , The sub-item weight coefficient.

[0065] For example, in the subway operation scenario, when there is a passenger flow congestion at a certain transfer station during the morning peak, the graph reasoning engine quickly retrieves the historical fault library and matches the records of the station that once caused congestion due to escalator failure and insufficient security channel. At the same time, real-time working condition data is fused, including the current security pass rate, escalator operation status, and the increase in the rate of incoming passenger flow. Through correlation analysis, a congestion cause reasoning tree is automatically generated, with the root cause pointing to the superimposed effect of the increase in commuter passenger flow of surrounding office buildings and the diversion of passenger flow at B station. The first branch includes direct factors such as reduced security efficiency and decreased transfer channel capacity, and the second branch is further refined to specific equipment failures and construction information. At the same time, the engine outputs a risk probability matrix. It is shown that within the next 15 minutes, the risk probability of the A station platform congestion reaching 50% overstaffing is 80%; the risk probability of the gate failure caused by passenger backlog is 30%; and the risk probability of causing a chain congestion at surrounding stations (such as C station) is 60%. Operation personnel immediately start the emergency plan according to these accurate data, deploy 5 standby security personnel to A station, open a temporary security channel, and coordinate with the construction party to suspend the transfer channel construction for 2 hours to restore the capacity, successfully alleviating the congestion risk.

[0066] In the embodiment of the present application, the visual large-screen interactive platform 300 includes, as shown in Figure 5 a multi-level alarm projection module, an AR perspective rendering engine, and a scene adaptive engine.

[0067] The multi-level alarm projection module is used to map device failures to the line network three-dimensional sand table according to the influence range, with core node failures marked in red.

[0068] It can be understood that, in the embodiment of the present application, the multi-level alarm projection module classifies device failures according to the influence range, with core node failures marked in red, intuitively presenting the influence degree of the fault on the subway line network, facilitating the operation and maintenance personnel to quickly lock the key problems; the AR perspective rendering engine generates a virtual perspective view of the real-time positioning of the train in the tunnel, with the full load rate of the carriages superimposed, helping the dispatchers to accurately grasp the train operation and passenger distribution, and reasonably arrange the transport capacity; the scene adaptive engine automatically switches the adaptive interface layout according to different modes such as commuter peak, emergency evacuation, and night maintenance through the operation scene classifier, so that the key information is highlighted, improving the operation efficiency and decision accuracy of the staff in various scenarios. The visual degree, response speed, and intelligent level of subway operation management are enhanced, and the safe and efficient operation of the subway is ensured.

[0069] It should be noted that equipment failures are categorized and mapped onto the network's 3D sandbox according to their impact. Level 1 alarms (emergency level, marked in red) address core node failures or large-scale outages, such as main substation tripping; Level 2 alarms (critical level, marked in orange) handle anomalies of important equipment affecting localized areas, such as single-station critical equipment failures; Level 3 alarms (information level, marked in yellow) focus on equipment failures or warnings with minor operational impacts, such as non-critical equipment anomalies; Level 4 alarms (information level, marked in blue / gray) record auxiliary information such as equipment status changes. Fault information is mapped to the network's 3D sandbox using color and hierarchy, helping maintenance personnel quickly identify high-priority faults and efficiently carry out emergency response.

[0070] For example, such as Figure 6 As shown, during the evening rush hour on a weekday, a sudden malfunction occurred at a key transfer station on Metro Line 5. The multi-level alarm projection module was quickly activated, and through rapid analysis of the fault data, the malfunction was determined to be a Level 1 alarm. On the 3D model of the metro network, the node representing this transfer station was prominently marked in red, and the surrounding affected lines were also presented in a semi-transparent red state. Statistics showed that the malfunction caused train delays on Line 5 within a 10-kilometer radius before and after the fault point, with an average delay time of 20 minutes, affecting 15 trains and approximately 3,000 passengers. Simultaneously, passenger congestion occurred at three adjacent stations, with approximately 200-300 passengers at each station, and these areas were marked as Level 2 alarm zones. Based on the different alarm levels displayed on the 3D model, relevant staff quickly prioritized their work, focusing on repairing the malfunction at the key transfer station, while simultaneously deploying personnel to the congested stations to manage passenger flow, minimizing the impact of the malfunction on metro operations.

[0071] In this embodiment of the application, the intelligent instruction decision engine 400 includes, as follows: Figure 7 As shown, the road network prediction model, resource constraint solver, and contingency plan optimizer are presented.

[0072] Among them, the road network prediction model is used to integrate LSTM neural network and cellular automata algorithm to deduce the spread range and duration of delay events in the road network; the resource constraint solver is used to generate the optimal additional / cancelled service plan under multiple constraints such as train turnaround time, number of spare cars and crew configuration; and the contingency plan optimizer is used to integrate expert rule base to generate emergency broadcast scripts and traffic diversion paths for graded responses.

[0073] It can be understood that the embodiments of the application fuse LSTM neural network and cellular automata algorithm through a road network prediction model, accurately deduce the diffusion range and duration of the late event in the road network, and master the potential influence area in advance; the resource constraint solver comprehensively considers multiple restriction conditions such as train turnaround time, number of standby vehicles and crew configuration, quickly generates the optimal start-up / shut-down scheme, reasonably allocates resources, and reduces the loss of delay; the pre-plan optimizer generates the emergency broadcast language and diversion path of hierarchical response combined with the expert rule base, ensures accurate information transmission and efficient guidance. The decision efficiency and scientificity of the subway in response to the sudden situation of delay are improved, the operation order is ensured, and the passenger travel experience is enhanced.

[0074] It should be noted that the LSTM neural network formula is:

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] wherein, is the output of the forgetting gate; is the time step; is the input vector; is the hidden state at the previous time; is the forgetting gate weight matrix; is the forgetting gate bias vector; is the Sigmoid activation function; is the output of the input gate; is the input gate weight matrix; is the input gate bias vector; is the candidate cell state; is the hyperbolic tangent activation function; is the candidate state weight matrix; is the candidate state bias vector; is the cell state; is the cell state at the previous time; is the output of the output gate; is the output gate weight matrix; is the output gate bias vector; is the hidden state.

[0082] The formula of the cellular automata algorithm is:

[0083]

[0084] wherein, is the state of cell c at time t+1; is each basic unit in the road network; is the set of all cells; is the state of cell c at time t; is a joint computing function combining cellular automata and long short-term memory networks; is a fusion rule; is the current state set; is the updated state set.

[0085] For example, during a weekday morning rush hour, the first train of subway line 3 is delayed by 10 minutes due to signal equipment failure. The road network prediction model is immediately started, and through the LSTM neural network analysis of similar fault data in the past 30 days, it is found that similar delay events in history will spread to adjacent 5 stations within 40 minutes; at the same time, the cellular automata algorithm divides the line into 20 road segment cells, and according to the connection relationship and operation rules of each road segment, simulates that the delay will spread along the upward direction at a speed of affecting 2 cells every 10 minutes. After model deduction, it is predicted that the affected range will cover the whole line of line 3 and the transfer station with line 5 after 1 hour, causing about 8000 person-times of delay. The operation department adjusts the subsequent train departure plan in advance, adds 3 standby trains in the affected area, and guides passengers to staggered travel through broadcast, effectively reducing the influence range and duration of the delay event.

[0086] In the embodiments of the present application, the dynamic adjustment unit 500 includes, as shown in Figure 8 a broadcast linkage controller, an optical flow guide matrix, and a scheduling instruction compiler.

[0087] The broadcast linkage controller is used to trigger multi-language electronic voice and AR navigation arrows; the optical flow guide matrix is used to generate evacuation path light bands through programmable LED floor tiles; the scheduling instruction compiler encodes the strategy into an executable instruction set of the ATO train automatic scheduling unit, and dynamically adjusts the ATO driving curve.

[0088] It can be understood that the embodiments of the application trigger multi-language electronic voice and AR navigation arrows through the broadcast linkage controller to provide clear and multi-lingual guidance for passengers; the light flow guide matrix generates evacuation path light bands by means of programmable LED floor tiles to intuitively guide passenger flow evacuation; the dispatch instruction compiler encodes the strategy into an executable instruction set of the ATO train automatic dispatch unit to dynamically adjust the driving curve. From passenger guidance to train dispatch, the cooperation improves the emergency response efficiency, optimizes the passenger evacuation experience, ensures the train operation to adapt to real-time strategies, and enhances the dynamic regulation and service level of subway operation.

[0089] For example, during the morning peak of the subway, the transfer station is suddenly congested with passenger flow due to equipment failure, and the light flow guide matrix responds quickly. According to the evacuation strategy output by the intelligent instruction decision engine, the programmable LED floor tiles in the station drive to generate clear evacuation path light bands in the station hall and platform area. From the transfer channel to the temporary security check area, the light band guides the 2000+ passengers in the backlog to be orderly distributed in a continuous flow of light, and cooperates with the multi-language voice and AR navigation arrows of the broadcast linkage controller to increase the congestion area passenger flow dissipation rate to 80% within 5 minutes, effectively relieving the passenger flow pressure and ensuring the safety and order of passenger traffic.

[0090] The subway traffic operation decision adjustment system based on the visual large screen proposed in the embodiments of the application collects train position and passenger flow density key data in real time through the perception data module, cooperates with the abnormal data filtering unit to remove invalid information, and ensures the accuracy and reliability of the data; the data processing hub uses the stream computing pipeline, the graph neural network analysis module and the graph reasoning engine to perform real-time aggregation, deep mining and intelligent reasoning on massive data, accurately locates potential congestion paths and failure causes; the visual large screen interactive platform uses multi-level alarm projection, AR perspective rendering and scene adaptive engine to convert complex operation data into intuitive and visual three-dimensional sand table and virtual perspective view, and automatically switches the interface layout according to different scenes to help operation personnel quickly grasp the overall situation; the intelligent instruction decision engine integrates LSTM neural network and cellular automata algorithm to deduce the diffusion influence of the late event and generate an optimal train dispatching scheme and a hierarchical emergency response plan under resource constraints; the dynamic adjustment unit quickly converts the decision instruction into multi-language broadcast, AR navigation and LED floor tile guidance, dynamically adjusts the ATO train dispatching curve, improves the subway operation efficiency and emergency disposal capacity, and reduces the operation risk and cost. Thus, the problems of existing technologies such as abnormal response lag, unreasonable resource allocation and low efficiency are solved.

[0091] A subway traffic operation decision adjustment system based on a visual large screen will be described below through a specific embodiment, as shown in Figure 9 , comprising:

[0092] The system is deployed in a subway operation control center, covering 5 lines and 80 stations. The sensor cluster of the perception data module (including 2000 train positioning sensors, 1500 platform monitoring cameras, 3000 gate sensors and 800 environmental sensors) is distributed in the line, station and train. The data processing hub relies on 10 high-performance servers to build a streaming computing and graph neural network analysis environment. The visualization screen uses a 12m x 4m ultra-high-definition splicing screen. The intelligent instruction decision engine and dynamic adjustment unit are interconnected with each station and train subsystem through the subway special communication network.

[0093] The train positioning sensor collects data once every second, with an accuracy of 0.5 meters. The platform monitoring camera collects video stream at 25 frames per second. The gate sensor records each passage data (including passenger ID and passage time). The environmental sensor collects temperature (accuracy ±0.3°C) and obstacle distance (accuracy ±0.1 meters) data every 5 minutes. The abnormal data filtering unit uses a sliding window outlier detection algorithm with a window size of 10 seconds. For example, if the positioning drift exceeds 2 meters (beyond the normal error range) in 3 consecutive windows in the train positioning data, it is determined as invalid data and removed. During the morning peak of a working day, due to sensor communication interruption, the filtering unit removed more than 300 invalid gate data within 1 hour, ensuring data quality.

[0094] The streaming computing pipeline aggregates and converts video stream, sensor data and operation logs by seconds. For example, integrating train positioning and gate passage data, the real-time passenger flow change curve of each station is generated, and data correlation is completed within 1 second. The graph neural network analysis module constructs a dynamic relationship topology with stations as nodes and lines as edges, and mines potential congestion transmission paths. Through analysis, it is found that during the morning peak, the passenger flow congestion at station A is easily transmitted to stations B and C through transfer stations, with a transmission delay of about 15 minutes, providing a basis for pre-planning. The graph reasoning engine integrates historical fault library (including 2000 fault records in 5 years) and real-time working condition data to automatically generate congestion cause reasoning tree and risk probability matrix. When passenger flow congestion occurs at station D, the reasoning engine finds out the causes such as equipment failure (e.g. gate failure) and large passenger flow (e.g. large event dispersal), with corresponding risk probabilities of 30% and 70% respectively, assisting operation decision-making.

[0095] ​The multi-level alarm projection module classifies device faults according to the influence range (core node fault, general node fault) and maps them to the line network three-dimensional sand table. On a certain day, the signal device of station E fails (core node fault), and the large screen is marked in red, while displaying 3 lines and 5 stations affected by the fault, which helps the operation personnel quickly locate. The AR perspective rendering engine generates a virtual perspective view of the real-time positioning of the train in the tunnel, and superimposes the full load rate thermal identifier (green for full load rate <60%, yellow for 60%-80%, and red for >80%). The operation personnel can directly see the train position and car congestion, and the car full load rate of a certain train reaches 90% (red mark), which triggers the passenger flow diversion instruction in time. The scene adaptive engine automatically switches the mode through the operation scene classifier (trained based on passenger flow, time period, etc.). During the morning peak (7:00-9:00), it switches to the commuter peak mode, highlighting the passenger flow heat map and train punctuality rate; during emergency evacuation, it switches to the emergency mode, focusing on the evacuation path and emergency resource distribution; during night maintenance (23:00-5:00), it switches to the maintenance mode, showing the device maintenance progress and skylight operation area.

[0096] Road network prediction model integrates LSTM neural network and cellular automata algorithm to deduce the spread of late events. When a train is late by 5 minutes at station F, the model deduces that it will spread to 2 adjacent stations after 10 minutes, with a continuous impact of about 20 minutes, providing time prediction for train scheduling. The resource constraint solver generates the optimal additional / stop operation scheme under the constraints of train turnaround time (minimum 12 minutes), number of standby cars (10), and crew configuration (20 groups available). During the morning peak, the solver calculates that 2 standby cars should be added and 3 crew shifts should be adjusted to increase the capacity of the line by 20%. The contingency optimizer integrates an expert rule base (including 100 emergency handling rules) to generate emergency broadcast scripts and diversion paths for hierarchical response. When a large passenger flow occurs at station G, the system generates multilingual broadcast scripts such as "Please go to the transfer channel, passengers on Line 2 from Exit B" and diversion paths along the right side of the platform to the transfer channel to Line 2, effectively diverting passenger flow.

[0097] The broadcast linkage controller triggers multi-language electronic voice and AR navigation arrows. When there is a large passenger flow at station H, the AR navigation arrows are projected on the platform ground and wall surface to guide passengers to the transfer channel, and the multi-language broadcast guidance in Chinese, English and Japanese is played synchronously to guide 500 passenger times in 10 minutes. The light flow guide matrix generates evacuation path light strips through programmable LED floor tiles. When equipment failure occurs at station I and evacuation is needed, the LED floor tiles are lit up in 3 seconds to form a green light strip from the failure area to the safe exit to guide passengers to evacuate, and the evacuation efficiency is improved by 30%. The scheduling instruction compiler encodes the strategy into an executable instruction set of the ATO train automatic scheduling unit to dynamically adjust the ATO driving curve.

[0098] To sum up, the embodiment of the application guarantees data reliability through high-frequency and high-precision acquisition and abnormal filtering of perception data modules to lay a foundation for decision-making; the data processing hub aggregates multi-source data at a second level and mines potential risks through graph neural networks and graph reasoning, such as predicting congestion transmission paths 15 minutes in advance and accurately locating the causes of 70% large passenger flow congestion; the visual large screen shortens the fault positioning time by 30% and improves the passenger flow state perception efficiency by 40% through three-dimensional sand table hierarchical alarm, AR thermal perspective and scene adaptive switching; the intelligent instruction decision engine generates an optimal scheme by fusing algorithms and rules, such as predicting an error of ≤5 minutes in late diffusion and increasing train operation to improve the transport capacity by 20%; the dynamic adjustment unit realizes large passenger flow evacuation of 500 passenger times in 10 minutes and improves the evacuation efficiency by 30% through multi-language broadcast, light flow guidance and train driving curve optimization. The early morning peak punctuality rate of the line network is improved, the large passenger flow evacuation time is shortened, and the fault handling response speed is improved.

[0099] Secondly, a subway traffic operation decision adjustment method based on a visual large screen is described with reference to the accompanying drawings according to an embodiment of the application.

[0100] As shown in the figure, the subway traffic operation decision adjustment method based on the visual large screen includes the following steps: Figure 10

[0101] In step S101, train positioning data, platform monitoring video stream, gate passage data and environmental sensor data are acquired.

[0102] It can be understood that the embodiment of the application can accurately master the train running track and position by acquiring train positioning data to ensure train safety and punctuality; the platform monitoring video intuitively presents the platform passenger flow distribution and abnormal conditions to facilitate timely discovery of emergencies; the gate passage data quantifies the passenger flow in and out of the station to provide a basis for transport capacity allocation; and the environmental sensor data monitors temperature, humidity, obstacle distance and environmental parameters to ensure the safety of the operating environment. ​

[0103] In step S102, a real-time line network state model is constructed according to train positioning data, platform monitoring video stream, gate passage data and environmental sensor data.

[0104] The real-time line network state model is a dynamic digital model of the subway line network constructed by train running trajectory, passenger flow, and real-time equipment working condition data after algorithm processing.

[0105] It can be understood that the embodiments of the present application dynamically process the dispersed train positioning, monitoring video, gate passage and environmental sensor data using evidence reasoning algorithm, and convert them into visual and analyzable digital models, which accurately present the train running trajectory, passenger flow change and equipment working condition in the subway line network. The operation personnel can intuitively master the real-time dynamics of the line network, quickly identify potential problems such as train delay, passenger congestion and equipment abnormality, and also provide basic data for intelligent instruction decision engine, helping to develop scheduling strategy in advance, optimize transport capacity allocation, efficiently cope with emergency situations, and improve overall operation efficiency and safety.

[0106] It should be noted that the evidence reasoning algorithm formula is:

[0107]

[0108]

[0109] wherein, is the trust degree of the fused uncertain state; is the trust degree of the specific state A after fusion; is the recognition framework; is any subset of is the total number of data sources; is the weight of the ith evidence; is the correlation coefficient of the ith evidence; is the mass function of the ith evidence itself; , is the subscript index; is the weight of the jth evidence; is the correlation coefficient of the jth evidence; is the mass function of the jth evidence itself.

[0110] In step S103, based on the real-time line network state model, the LSTM neural network and the cellular automaton algorithm are fused to identify abnormal events, and at the same time, the line network congestion heat map, equipment failure positioning mark and passenger flow prediction trend curve are dynamically rendered on the visual large screen. ​​​​

[0111] The passenger flow prediction trend curve is a visual curve for showing the change trend and rule of the passenger flow quantity in the future period, which is fitted by analyzing historical passenger flow data and influencing factors (such as time, activities, traffic hub dynamics, etc.) and using a mathematical model or algorithm.

[0112] It can be understood that the embodiments of the present application intuitively present the future passenger flow change trend by mining historical data and multiple influencing factors, assist the operation party in predicting the peak period and passenger flow dense area in advance, flexibly allocate transport capacity, add temporary channels or adjust train frequency, and relieve congestion pressure; help the traffic hub optimize resource allocation and reasonably arrange the number of service windows such as security check and ticket sales; at the same time, it can also provide travel reference for passengers, stagger peak planning routes, and improve travel efficiency and experience.

[0113] For example, during a large holiday, the subway operation department fits the passenger flow prediction trend curve of the next three days by using the LSTM neural network algorithm based on historical passenger flow data, combined with the holiday travel rule, city traffic hub dynamics and surrounding large activity arrangement. It is found from the curve that there will be two passenger flow peaks from 10 am to 2 pm and from 5 pm to 8 pm, and the passenger flow of the line to the business district and scenic spot increases by more than 3 times the regular rate. The operation party increases the train frequency in the peak period in advance, increases the number of guide personnel at key stations, and pushes the peak period and crowded line prompts to passengers through the official APP, effectively relieving the passenger flow pressure, improving the passenger travel experience, and ensuring the safe and efficient operation of the subway network.

[0114] In step S104, a train frequency adjustment strategy and an emergency relief scheme are dynamically generated according to the abnormal event and the passenger flow prediction trend, combined with the slime mold algorithm.

[0115] The slime mold algorithm is an intelligent optimization algorithm that simulates the formation of an efficient foraging network by cytoplasm flow in Physarum polycephalum.

[0116] It can be understood that the embodiments of the present application simulate the dynamic flow characteristics of cytoplasm when Physarum polycephalum forages, quickly and flexibly adjust the search strategy according to the abnormal event and the passenger flow prediction trend. In the train dispatching scenario, the global and local search are balanced, the calculation amount is greatly reduced, and the real-time requirement is met; at the same time, with the adaptive weight and global guide item, the complex constraints such as track capacity and safety interval are efficiently processed, in the case of sudden large passenger flow or train failure, the optimal train frequency adjustment strategy and emergency relief scheme are quickly generated, which takes into account the operation cost, passenger waiting time and emergency response speed, and the dynamic adaptability and robustness of the dispatching system are improved.

[0117] It should be noted that the formula of the slime mold algorithm is:

[0118]

[0119] in, For flight schedule strategy; For a new generation of scheduling strategies; This is the historically optimal strategy; External disturbance coefficient; This is the data weight matrix; A random candidate strategy; It is a random number; Adjust the threshold for the strategy; For random exploration factors; This represents the current scheduling strategy.

[0120] For example, such as Figure 11 As shown, in the subway operation of a first-tier city, during the morning rush hour on weekdays, several core stations in the city center, such as Financial Street Station and Guomao Station, can see an average hourly passenger flow of 50,000 to 80,000 people, often causing congestion due to the excessive concentration of commuters. Meanwhile, equipment failures, extreme weather, and other abnormal events occur approximately 20-30 times annually. To cope with these complex situations, the operations department introduced a "slime mold" algorithm. They used key data such as real-time passenger flow, train speed, and line capacity at each station as "food source" information. When the algorithm detected that Financial Street Station was experiencing a sudden equipment failure, and the passenger flow was expected to accumulate to 100,000 people within one hour, far exceeding the station's capacity limit of 70,000 people per hour, it acted like slime mold sensing a high concentration of food source, quickly activating its response mechanism. Based on past data and real-time information, the algorithm, through multiple rounds of iterative calculations, completed strategy generation within just 5 minutes, deciding to urgently add 10 trains within one hour at three adjacent stations around the faulty station to guide passengers to evacuate from these stations. Meanwhile, the train schedules of 15 lines were adjusted, with trains originally destined for the faulty station being rerouted to alternative stations 1-2 kilometers away, effectively avoiding the fault location. Actual verification showed that after applying this algorithm, in similar abnormal events, the average passenger evacuation time was reduced from 30 minutes to 15 minutes, and operating costs were reduced by approximately 20% when dealing with abnormal events, significantly improving the subway's ability to respond to complex and dynamic environments and its operational efficiency.

[0121] In step S105, the ATO automatic train dispatch curve is dynamically adjusted according to the adjustment strategy and diversion plan. At the same time, multilingual broadcasts are conducted and the content of the electronic guidance screen is updated.

[0122] Among them, the ATO (Automatic Train Operation) curve refers to the speed-distance or time-distance curve automatically generated by the Automatic Train Operation (ATO) system based on line parameters, train performance, timetable requirements, and energy-saving objectives, used to control the train's traction, inertia, braking, and other operating states.

[0123] It can be understood that the embodiments of the application dynamically optimize the train operation state by real-time adjustment of the strategy and the diversion scheme. In response to sudden large passenger flow, by extending the inter-station running time and increasing the stop time, the platform congestion is relieved; in the case of equipment failure, the train operation path and speed are adjusted to ensure the safety of train operation. The transportation efficiency is improved, the train delay is reduced, the energy consumption is reduced, the green operation is realized, and at the same time, the multi-language broadcast and the electronic guide screen are linked to timely transmit accurate information to passengers, improve the passenger travel experience, and ensure the safe, efficient and stable operation of the rail transit system under complex scenes.

[0124] For example, during the international marathon event held in a certain city subway, it is expected that 30,000 return passengers will rush to the station near the finish line within 1 hour after the race. The ATO train automatic dispatching curve dynamic adjustment mechanism is enabled in the subway dispatching system: for the trains heading to the terminal, based on the passenger flow prediction data and the line parameters, the interval running curve is adjusted to the "acceleration section extension + braking section advance" mode, the maximum speed is increased from 80 km / h to 85 km / h on the flat line, the inter-station running time is shortened by 15%, the stop time at the terminal station is extended from 30 seconds to 45 seconds, and the train stopping accuracy is controlled within ±0.5 meters through curve optimization, which facilitates passengers to quickly get on and off the train. In addition, the system dynamically adjusts the tracking curve according to the position of the subsequent train, and the departure interval is compressed from 4 minutes to 3 minutes. Finally, through the precise adjustment of the ATO curve, the line transports 28,000 passenger trips within 1 hour, the train punctuality rate reaches 99%, the transportation capacity is increased by 30% compared with the conventional dispatching mode, the energy consumption is reduced by 12%, and efficient operation is realized in the large passenger flow scenario.

[0125] According to the subway traffic operation decision adjustment method based on the visual large screen provided in the embodiment of the application, the train position and passenger flow density key data are collected in real time through the perception data module, invalid information is removed through the abnormal data filtering unit, and the accuracy and reliability of the data are ensured; the data processing hub aggregates, deeply mines and intelligently reasons the massive data in real time by means of the stream computing pipeline, the graph neural network analysis module and the graph reasoning engine, accurately locates the potential congestion path and the fault cause; the visual large screen interactive platform converts the complex operation data into the intuitive three-dimensional sand table and virtual perspective view by means of the multi-level alarm projection, AR perspective rendering and scene adaptive engine, automatically switches the interface layout according to different scenes, and assists the operation personnel in quickly grasping the overall situation; the intelligent instruction decision engine combines the LSTM neural network and the cellular automaton algorithm, deduces the influence of the late event diffusion, generates the optimal train scheduling scheme and the hierarchical emergency response plan under the resource constraint, and quickly converts the decision instruction into the multi-language broadcast, AR navigation and LED floor tile guidance, dynamically adjusts the ATO train scheduling curve, improves the subway operation efficiency and emergency disposal capacity, and reduces the operation risk and cost. Thus, the problems of the prior art such as abnormal response lag, unreasonable resource allocation and low efficiency are solved.

[0126] A subway traffic operation decision adjustment method based on a visual large screen will be described below through a specific embodiment, as shown in Figure 12 , comprising:

[0127] A certain urban subway line 3 is the main artery of urban traffic, and the daily passenger flow breaks through 500,000 person-times during the morning peak period (7:30-9:00). The line passes through multiple core commercial centers and large residential areas, and the commuting and shopping passenger flows are highly superimposed, so that the daily operation faces great pressure. At 8:00 during the morning peak period on a certain day, the system monitoring network sensitively captures the abnormal fluctuation of the key section data, and triggers the subway traffic operation decision adjustment process based on the visual large screen.

[0128] Train positioning data is collected by the vehicle-mounted GPS module and the trackside communication base station at a high frequency of 1 time per second to obtain real-time train position, running speed and head / tail coordinates. For example, train T1 is accurately positioned at the 15th signal section of Line 3 at 8:00, with coordinates (X1, Y1) and a speed of 40 km / h. The high-definition cameras at each platform record the platform dynamics in real time by returning the monitoring video stream through a private network at a resolution of 1920x1080 and a frame rate of 25 frames per second. The gate system accurately records the entry and exit information of each passenger, including the passage time, gate number and passenger type. For example, during the period from 8:00 to 8:01, the A station entry gate G1 records 32 people passing through. At the same time, environmental sensors such as temperature, humidity, smoke and vibration are deployed in the tunnels and platforms to provide environmental data support for operational safety.

[0129] The acquired multi-source heterogeneous data is immediately transmitted to the data processing server cluster for cleaning, conversion and fusion. Invalid information such as abnormal data points with GPS positioning errors exceeding the threshold is removed to ensure data quality. Subsequently, a real-time line network state model is constructed using the graph database Neo4j. The stations, sections, switches and signal machines in the subway line are abstracted as nodes, and the train running path and passenger flow direction are abstracted as edges. The real-time running state of the subway line network is visually presented in a graph structure to provide a clear data model foundation for subsequent decision-making.

[0130] In the abnormal event identification section, the system combines LSTM neural networks and cellular automata algorithms to deeply analyze the real-time line network state model data. The LSTM neural network learns from historical passenger flow and train operation time series data to mine time series characteristics and rules and accurately predict normal operation trends. The cellular automata algorithm divides the subway line network into multiple cells and simulates the dynamic evolution process based on the train state and passenger flow density within the cells. At 8:00, the system analyzes and finds that the train running speed in the C-D station section has dropped sharply and subsequent trains are queuing. Combined with the passenger gathering situation in the platform monitoring video, it is determined that a device failure has occurred in the section and an alarm is triggered. In terms of visual display, the line section passenger flow and platform passenger flow are dynamically rendered in red (congestion), yellow (busy) and green (smooth); the device failure location is accurately marked using GIS map technology, and the type and location coordinates of the faulty device are displayed in detail; based on historical and real-time monitoring data, the LSTM prediction results are superimposed to show the passenger flow trend of each station in the next 30 minutes, and the large screen is updated in real time to provide a visual basis for operational decision-making.

[0131] When an abnormal event is identified (e.g., a sudden large passenger flow at a station, with passenger flow exceeding 80% of the station's maximum capacity within 15 minutes) and combined with the passenger flow prediction trend (future passenger flow will continue to rise for the next 30 minutes), the slime mold algorithm is used to find the optimal path for passenger evacuation. The adjustment strategy and the evacuation scheme are generated. The slime mold algorithm parameters are set, the population size is 50, the maximum number of iterations is 100, the decay coefficient is 0.5, the train departure frequency and the marshalling quantity of each period are taken as the decision variables, and the objective function of minimizing the operation cost (weight 0.4), the average waiting time of passengers (weight 0.3) and the average delay time (weight 0.3) is constructed; through the iteration of the core formula, the final adjustment strategy is generated, 8 trains are added within 30 minutes in the 3 adjacent stations around the affected station, the departure interval of 5 related lines is adjusted from 5 minutes to 3 minutes, and the emergency evacuation scheme is developed, the temporary guide signs are set, the passengers are guided to the adjacent stations through multi-language broadcast, and the delay and alternative route information is updated in real time on the electronic guide screen.

[0132] In the adjustment stage, the train schedule adjustment strategy is quickly transmitted to the automatic train operation system (ATO), and the ATO dynamically adjusts the automatic train scheduling curve accordingly. The running path and speed curve of the added train in the interval are re-planned to ensure the safety interval with the existing train; for the train with adjusted departure interval, the traction, braking time and speed threshold of each interval are finely modified, for example, the train station running speed is reduced by 10% for the train with extended departure interval, to ensure the operation safety and punctuality. In the information release link, the control center sends multi-language (Chinese, English, Japanese, etc.) broadcast instructions to each station through the broadcast system, and broadcasts the equipment failure, train adjustment and transfer suggestion in detail; meanwhile, the content of the electronic guide screen of each station is updated remotely, for example, the electronic guide screen of B station displays in real time that "the equipment in the C-D station interval is faulty, and it is recommended to take X bus to D station", to ensure that passengers can obtain accurate information in time.

[0133] In summary, the embodiment of the application realizes the accurate perception and dynamic modeling of the subway operation state through real-time data collection and deep fusion, quickly and accurately identifies abnormal events by using the LSTM neural network and the cellular automaton algorithm, and intuitively presents the operation situation by using the visual large screen, thereby providing clear basis for decision-making. On this basis, the slime mold algorithm is used to scientifically optimize the train scheduling strategy, and the operation cost and passenger experience are effectively balanced. In the scheme execution stage, the automatic adjustment curve and multi-channel information release ensure the efficient landing of the strategy and the timely awareness of passengers, greatly improve the response speed and disposal efficiency of the subway in the face of complex operation scenes and sudden conditions, reduce the operation risk, and enhance the safety, convenience and satisfaction of passenger travel, thereby significantly improving the intelligent operation level and comprehensive service capability of the urban subway transportation system.

[0134] Figure 13 The structure schematic diagram of the electronic device provided by the embodiment of the application is shown. The electronic device can include:

[0135] The memory 1301, the processor 1302, and the computer program stored in the memory 1301 and executable on the processor 1302.

[0136] The processor 1302 implements the method for adjusting subway traffic operation decision based on a large visual screen in the above embodiments when executing a program.

[0137] Further, the electronic device further comprises:

[0138] The communication interface 1303 is configured to communicate between the memory 1301 and the processor 1302.

[0139] The memory 1301 is configured to store a computer program executable on the processor 1302.

[0140] The memory 1301 can include a high-speed RAM memory, and can further include a non-volatile memory, for example, at least one disk memory.

[0141] If the memory 1301, the processor 1302 and the communication interface 1303 are independently implemented, the communication interface 1303, the memory 1301 and the processor 1302 can be connected to each other through a bus and complete communication between each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 13 Only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0142] Optionally, in a specific implementation, if the memory 1301, the processor 1302 and the communication interface 1303 are integrated on a chip, the memory 1301, the processor 1302 and the communication interface 1303 can complete communication between each other through an internal interface.

[0143] The processor 1302 can be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0144] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the method for adjusting subway traffic operation decision based on a large visual screen.

[0145] Furthermore, the embodiment of the present application also provides a computer program product, comprising a computer program or instructions, which, when executed, implement the above-mentioned subway traffic operation decision adjustment method based on a visual large screen.

[0146] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms is not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0147] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "multiple" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0148] Any process or method descriptions in flow charts or described elsewhere herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing the specified logical functions or processes, and the preferred embodiments of the present application also include additional implementation involving other processes or methods. It will be understood by those skilled in the art that the functions shown in the flow charts or otherwise described herein can be performed in an order different than that shown, in substantially simultaneous fashion, or in reverse order, as appropriate, and that the described embodiments can include additional functions not necessarily shown in the figures or described herein.

[0149] It should be understood that parts of the present application can be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. As in another embodiment, if implemented in hardware, any of the following technologies known in the art or their combinations can be used: discrete logic circuit with logic gate circuit for implementing logical functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA), etc.

[0150] Those skilled in the art can understand that all or part of the steps of the method carried out by the above-mentioned embodiments can be completed by programs instructing relevant hardware, and the programs can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0151] Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application. Those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A subway traffic operation decision adjustment system based on a visual large screen, characterized in that, The system comprises a perception data module, a data processing hub, a visual large-screen interactive platform, an intelligent instruction decision engine, and a dynamic adjustment unit. The perception data module is configured to collect train positioning data, platform monitoring video stream, gate passage data, and environmental sensor data in real time. The data processing hub is configured to integrate and analyze the collected data, mine potential patterns, correlation, and abnormal conditions in the data, and comprises a streaming computing pipeline, a graph neural network analysis module, and a graph reasoning engine. The visual large-screen interactive platform is configured to present a line network congestion heat map, device fault positioning, and passenger flow prediction trends in real time, and switch interfaces and highlight information according to different operation scenarios and needs. The intelligent instruction decision engine is configured to dynamically generate train schedule adjustment strategies and emergency evacuation plans, and comprises a road network prediction model, a resource constraint solver, and a pre-plan optimizer. The dynamic adjustment unit is configured to dynamically adjust electronic guide screens and platform broadcasts according to decision instructions, and dynamically schedule trains. The perception data module comprises a sensor cluster and an abnormal data filtering unit.

2. The subway traffic operation decision adjustment system based on the large screen visualization according to claim 1, characterized in that, The visual large-screen interactive platform comprises a multi-level alarm projection module, an AR perspective rendering engine, and a scene adaptive engine.

3. The subway traffic operation decision adjustment system based on the large screen visualization according to claim 1, characterized in that, The multi-level alarm projection module is configured to map device faults to a line network three-dimensional sand table according to the influence range, and mark core node faults in red. The AR perspective rendering engine is configured to generate a virtual perspective view of real-time positioning of trains in tunnels, and superimpose a full load rate heat map. The scene adaptive engine is configured to automatically switch interface layouts of a commuting peak mode, an emergency evacuation mode, and a night maintenance mode through an operation scenario classifier.

4. The subway traffic operation decision adjustment system based on the large screen visualization according to claim 1, characterized in that, The dynamic adjustment unit comprises a broadcast linkage controller, an optical flow guide matrix and a dispatch instruction compiler, wherein the broadcast linkage controller is used for triggering multi-language electronic voice and AR navigation arrows; the optical flow guide matrix is used for generating evacuation path light bands through programmable LED floor tiles; and the dispatch instruction compiler encodes strategies into an executable instruction set of an ATO train automatic dispatch unit to dynamically adjust an ATO driving curve.

5. A subway traffic operation decision adjustment method based on a visual large screen, characterized in that, Comprise: acquire train positioning data, platform monitoring video stream, gate passing data and environmental sensor data; construct a real-time line network state model according to the train positioning data, platform monitoring video stream, gate passing data and environmental sensor data, wherein the dispersed train positioning, monitoring video, gate passing and environmental sensor data are dynamically processed by using an evidence reasoning algorithm to be converted into a visualized and analyzable digital model, wherein the evidence reasoning algorithm formula is: in, The degree of trust in the uncertain state after fusion; The trust level of the specific state A after fusion; To identify the frame; for any subset of; Total number of data sources; Let be the weight of the i-th piece of evidence; Let be the correlation coefficient of the i-th piece of evidence; Let mass be the mass function of the i-th piece of evidence itself; , Subscript index; For the first The weight of each piece of evidence; For the first The correlation coefficient of each piece of evidence; For the first The mass function of the evidence itself; based on the real-time line network state model, fuse an LSTM neural network and a cellular automaton algorithm to identify abnormal events, and simultaneously dynamically render a line network congestion heat map, device fault positioning mark and passenger flow prediction trend curve on a visualized large screen; according to the abnormal events and passenger flow prediction trend, combine a slime mold algorithm to dynamically generate a train schedule adjustment strategy and an emergency evacuation scheme; according to the adjustment strategy and evacuation scheme, dynamically adjust an ATO train automatic dispatch curve, simultaneously perform multi-language broadcast and update electronic guide screen content.

6. An electronic device, comprising: Comprise: a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize the metro traffic operation decision adjustment method based on a visualized large screen according to claim 5.

7. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions are executed to realize the metro traffic operation decision adjustment method based on a visualized large screen according to claim 5.

8. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions are executed to realize the metro traffic operation decision adjustment method based on a visualized large screen according to claim 5.

Images