A gate machine real-time control method and device

Abstract

This application discloses a real-time control method and device for turnstiles, relating to the field of intelligent traffic management and control technology. The method utilizes simulation on a traffic hub simulation platform with turnstiles to filter and obtain multiple candidate turnstile control schemes for each passenger flow pattern, forming a multi-scenario simulation strategy library. The turnstile control scheme is a hybrid one-way and two-way configuration scheme for the turnstiles. Real-time passenger flow data of the traffic hub is continuously acquired. Based on real-time passenger flow data acquired in historical periods, a long short-term memory network is used to predict passenger flow data for future periods. By integrating real-time passenger flow data, the passenger flow pattern for future periods is determined. Multiple candidate turnstile control schemes are matched in the multi-scenario simulation strategy library, and simulation is performed using the traffic hub simulation platform. The optimal turnstile control scheme is selected from the multiple candidate schemes, and the turnstiles are controlled to execute the optimal control scheme. The turnstiles of this application can intelligently and dynamically adapt to tidal passenger flow, improving resource utilization and traffic efficiency.

Description

Technical Field

[0001] This application relates to the field of intelligent traffic management and control technology, and in particular to a method and device for real-time control of turnstiles. Background Technology

[0002] In high-density passenger flow scenarios such as urban transportation hubs and large venues, the throughput efficiency of ticket gate systems is a key bottleneck restricting overall service capacity. Traditional solutions mainly fall into two categories: one is the widely adopted fixed-function one-way gate, where the entry and exit channels are fixed and separate. During peak hours, at most stations with significant tidal passenger flow, this model leads to severe overload of the entry gates while the utilization rate of the exit gates on the opposite side is extremely low, resulting in a structural mismatch of resources: "congested on one side, idle on the other." Furthermore, because the entrance and exit of one-way gates are usually set up separately, unnecessary detours occur during off-peak hours when passenger flow is low. These problems not only reduce throughput efficiency but also lead to a waste of space and equipment.

[0003] Another option is to deploy two-way turnstiles, which have the advantage of switching between traffic directions, providing flexibility and efficient use of space to some extent. However, in existing technologies, the control of two-way turnstiles largely relies on preset schedules or simple manual judgment, lacking a precise response to real-time passenger flow. During extreme peak periods with rapidly changing passenger flow, their inherent operating mode reveals limitations, making it difficult to promptly and accurately utilize idle resources in the opposite direction to cope with the instantaneous peak of one-way passenger flow. The lag and inaccuracy of the control limit the maximization of their effectiveness.

[0004] Therefore, the main problems in the existing technology are: the fixed one-way gate configuration cannot adapt to the spatiotemporal imbalance of passenger flow, and the existing two-way gate control method lacks support based on real-time perception and intelligent decision-making, resulting in low overall resource utilization of the gate system, the need to improve passage efficiency, and the inability to effectively cope with the impact of large passenger flow. Summary of the Invention

[0005] The purpose of this application is to provide a real-time control method and device for turnstiles, which can intelligently and dynamically adapt to tidal passenger flow, thereby improving resource utilization and passage efficiency.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a real-time control method for a turnstile, including: Establish a simulation platform for transportation hubs equipped with turnstiles; Using the aforementioned transportation hub simulation platform, all gate control schemes for different passenger flow patterns are simulated one by one, and multiple candidate gate control schemes for each passenger flow pattern are selected to form a multi-scenario simulation strategy library; the gate control scheme is a hybrid one-way and two-way configuration scheme for the gate. Continuously acquire real-time passenger flow data of the transportation hub; Based on real-time passenger flow data obtained from historical periods, long short-term memory networks are used to predict passenger flow data for future periods, and combined with real-time passenger flow data, passenger flow patterns for future periods are determined. Based on the passenger flow pattern in the future time period, multiple candidate gate control schemes are matched and obtained from the multi-scenario simulation strategy library, and the transportation hub simulation platform is used to perform simulations to select the optimal gate control scheme from the multiple candidate gate control schemes. In the future, the control gate will execute the optimal gate control scheme.

[0007] Secondly, this application provides a real-time gate control device, comprising: a gate and a cloud platform; the cloud platform controls the gate to execute the optimal gate control scheme by executing the above-described real-time gate control method.

[0008] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a real-time gate control method and device, which senses real-time passenger flow data of transportation hubs, combines passenger flow data predicted by long short-term memory networks to determine passenger flow patterns, and uses a transportation hub simulation platform to simulate in advance to select the optimal gate control scheme, thereby controlling the gates to execute the optimal gate control scheme, so that the gates can intelligently and dynamically adapt to tidal passenger flow, improve resource utilization and passage efficiency. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A schematic diagram illustrating the core advantages of the two-way turnstile provided in the embodiments of this application; Figure 2 A flowchart illustrating a real-time gate control method provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating the principle of a real-time gate control method provided in an embodiment of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0013] Compared to the fixed "one in, one out" configuration of one-way turnstiles, Figure 1 The bidirectional turnstiles shown are flexible. In rail transit stations where passenger flow exhibits significant tidal or sudden surges, and the number and proportion of passengers entering and exiting fluctuate over time, bidirectional turnstiles can be activated to simultaneously serve passengers entering and exiting the station, thereby improving equipment utilization, pedestrian space utilization, and passenger flow efficiency. However, the control logic of existing bidirectional turnstiles is relatively traditional and fixed, lacking the ability to accurately and dynamically respond to real-time passenger flow. This application provides a real-time control method for turnstiles suitable for time-varying passenger flow characteristics.

[0014] In one exemplary embodiment, such as Figure 2 As shown, the real-time control method for the gate includes the following steps 101 to 106.

[0015] Step 101: Establish a traffic hub simulation platform with turnstiles.

[0016] Step 102: Using the transportation hub simulation platform, simulate all gate control schemes for different passenger flow patterns one by one, and select multiple candidate gate control schemes for each passenger flow pattern to form a multi-scenario simulation strategy library; the gate control scheme is a hybrid configuration scheme of single and two-way gates.

[0017] Step 103: Continuously acquire real-time passenger flow data of the transportation hub.

[0018] Step 104: Based on the real-time passenger flow data obtained from historical periods, use the Long Short-Term Memory Network to predict the passenger flow data for future periods, and combine the real-time passenger flow data to determine the passenger flow pattern for future periods.

[0019] Step 105: Based on the passenger flow pattern in the future time period, multiple candidate gate control schemes are obtained by matching in the multi-scenario simulation strategy library, and the transportation hub simulation platform is used to perform simulation to select the optimal gate control scheme from the multiple candidate gate control schemes.

[0020] Step 106: In the future time period, control the turnstiles to execute the optimal turnstile control scheme.

[0021] Implementing steps 101 to 106 above fully leverages the advantages of bidirectional turnstiles, requiring real-time adjustment of the turnstile opening direction based on predicted and real-time passenger flow information.

[0022] In another exemplary embodiment of this application, step 101 described above may be replaced by steps 201 to 203.

[0023] Step 201: Establish the basic model of the transportation hub using the Anylogic multi-agent simulation platform.

[0024] Using the Anylogic multi-agent simulation platform, a general basic model is first established. The model integrates hardware elements such as station boundaries, obstacles, security facilities, and gate layout. Subsequently, different schemes can be switched by simply adjusting the gate parameters, reducing redundant modeling.

[0025] Step 202: Simulate the underlying logic of pedestrians using the social force model built into the Anylogic multi-agent simulation platform.

[0026] The underlying logic of the simulation adopts a social force model, with the following formula: 1. According to Newton's second law, the equation for the total force on an individual body (the core driving formula for simulation): ; : The two-dimensional acceleration vector of pedestrian i at time t; The total force acting on pedestrian i at time t; Self-driving force (subjective goal-oriented force); Pedestrian - Pedestrian repulsion force; Pedestrian-facility repulsion force; Random force; The weight of pedestrian i is assumed to be 75kg for males and 60kg for females. For children, the weight is calculated based on their height.

[0027] 2. Formulas for calculating each component force (1) Self-driving force (goal-oriented core force): ; Relaxation time (the time it takes for a pedestrian to adjust their speed to the desired speed, for adults) Children / Carriers with luggage ; : Dynamic expected velocity (simulation scenario settings: 1.2 m / s in free flow, reduced to 0.3~0.6 m / s in congested conditions); : Desired direction unit vector (pointing to the entrance / exit) ); : The real-time velocity vector of pedestrian i at time t.

[0028] (2) Pedestrian-pedestrian repulsion force (core force of crowding behavior): ; Repulsive force strength coefficient (adult) Luggage carrier ,child ); : Repulsive force range coefficient (taken as 0.2m, characterizing the rate at which the repulsive force decreases with distance); Virtual radius for pedestrians (0.3m for adults, 0.5m for those carrying luggage); The real-time Euclidean distance between pedestrian i and pedestrian j. ; : The unit vector pointing from pedestrian j to pedestrian i, .

[0029] (3) Pedestrian-facility repulsion force (convenience constraint core force): ; Real-time distance between pedestrian i and facility B (turnstile, fence, wall); : The unit vector pointing from facility B to pedestrian i.

[0030] (4) Random force (individual difference correction force): ; Random force intensity coefficient (taken as 0.5N, balancing simulation realism and stability); Two-dimensional Gaussian random vector (mean 0, variance 1, randomly generated at each time step).

[0031] 3. Simulation Iteration Formula (Motion State Update, Underlying Calculation Process) The simulation library calculates pedestrian speed and position frame by frame through time step iteration to achieve dynamic passenger flow simulation. Velocity update formula (from acceleration integral): ; Constraint: The upper speed limit shall not exceed the desired speed. To prevent pedestrians from speeding; Position update formula (from velocity integral): ;in, Let i be the two-dimensional centroid coordinates. .

[0032] 4. Simplified Formula for Group Behavior (Simulation Optimization for Large-Scale Passenger Flow) For scenarios with large passenger flows in subway stations (such as thousands of passengers during peak hours), it is unnecessary to calculate the force on each pedestrian. The simulation can be simplified by using the average force of the group, thus reducing the computational cost. Average collective force: ; in, To monitor the total number of people in the area, It represents the overall movement trend of the group (consistent direction = orderly flow, chaotic direction - crowding and hedging).

[0033] Group average speed: It can directly correlate the passenger flow density-speed relationship in the simulation scenario to verify the consistency with real subway station passenger flow data.

[0034] Step 203: Embed the Monte Carlo method into the Anylogic multi-agent simulation platform, and randomly set it to simulate passenger flow demand with different total passenger flow and entry / exit ratio, thereby building a transportation hub simulation platform for passenger flow prediction and gate control scheme selection.

[0035] To reflect real-time passenger flow data at subway stations, various passenger flow patterns, including "tidal" and "platform" patterns, are considered. Furthermore, for extreme passenger flows caused by special weather, holidays, or large-scale events, the Monte Carlo method is used to define random variables, including: 1. Passenger arrival distribution: inbound passenger flow (Poisson distribution), outbound passenger flow (Poisson distribution), passenger arrival interval (exponential distribution), and queuing patience time (normal distribution). 2. Pedestrian micro-parameters: card swiping time (normal distribution, mean 2.5s, standard deviation 0.8s), card swiping failure rate (binomial distribution, failure rate 3%), and walking speed to the turnstile (normal distribution, mean 1.1m / s).

[0036] The Monte Carlo method is embedded to handle random variables and achieve probability simulation. Built-in software functions are directly called to implement Monte Carlo random sampling, corresponding to the previously defined variable distribution.

[0037] Poisson distribution (passenger flow): poisson(λ), where λ is the mean; Normal distribution (card swipe time, walking speed): normal(μ,σ), where μ is the mean and σ is the standard deviation; Binomial distribution (failure rate, card swipe failure rate): binomial(p,n), where p is the probability and n is the number of times; Exponential distribution (passenger flow interval): exponential(β), where β is the mean interval.

[0038] In another exemplary embodiment of this application, step 102 described above may be replaced by steps 301 to 304.

[0039] Step 301: Using the transportation hub simulation platform, simulate all gate control schemes under different passenger flow ratios one by one to obtain the index data of each gate control scheme under each passenger flow ratio; the passenger flow ratio represents the passenger flow pattern; the index includes traffic efficiency index and safety congestion index.

[0040] ①Simulation of turnstile solutions, the core data source for the solution library.

[0041] Based on the basic model and Monte Carlo embedding, taking the four-gate channel as an example, we simulated all 34 gate schemes one by one, collected the index data of each scheme, and provided support for the scheme library.

[0042] ② Quantitative evaluation and construction of a gate system solution library.

[0043] The model simulates various hybrid configuration schemes of one-way and two-way turnstiles under different passenger flow ratios, and uses the average passenger queuing time and average queuing length as performance indicators as quantitative evaluation criteria. Quantitative evaluation index system for optimal gate scheme: Traffic efficiency indicators (60% weighting): Average queue length: The average number of pedestrians queuing in front of the turnstile, calculated as: the sum of queue lengths in each time period ÷ the number of time periods; Maximum queue length: The longest queue length that appears in front of the turnstile, i.e., the maximum queue length in all time periods; Single gate / All gates: The maximum number of people who can pass through in 1 hour, i.e., the total number of people who can pass through in 1 hour.

[0044] Safe congestion index (40%): Congestion probability at turnstiles: The probability of ≥3 people queuing at turnstiles, calculated as: number of congestion attempts ÷ total number of simulations × 100%; Passenger flow conflict rate: Conflicts such as pedestrians cutting in line and routes in front of the turnstile are calculated as: total number of conflicts ÷ total number of people passing through × 100%.

[0045] Step 302: Weighted summation of the passage efficiency index data and safety congestion index data of each gate control scheme under each passenger flow ratio to obtain the quantitative evaluation index data of each gate control scheme under each passenger flow ratio.

[0046] Step 303: Determine all gate control schemes with the largest quantitative evaluation index data under each passenger flow ratio, and use them as multiple candidate gate control schemes under each passenger flow ratio.

[0047] Step 304: Establish a multi-scenario simulation strategy library based on the one-to-one correspondence between passenger flow ratio, candidate gate control schemes and quantitative evaluation index data.

[0048] The simulation data of all optimal solutions are aggregated and categorized according to "scenario-solution-indicator" to form a queryable and callable solution library with a clear structure for convenient subsequent application. The optimal gate configurations and control strategies for various typical passenger flow patterns are stored to form the multi-scenario simulation strategy library.

[0049] In another exemplary embodiment of this application, step 103 above provides a method for acquiring pedestrian flow data through multi-source sensing. Step 103 can be replaced by steps 401 to 407.

[0050] Step 401: Monitor the three-dimensional coordinate point cloud data of pedestrians in the upstream guidance area using lidar.

[0051] Real-time passenger flow data is obtained based on lidar: lidar equipment is deployed in the entrance area and elevator area of ​​each subway station for remote data monitoring. LiDAR can directly output the three-dimensional coordinate point cloud data of pedestrians in the monitoring area.

[0052] Step 402: Determine the number of pedestrians in the upstream guidance area based on the pedestrian's 3D coordinate point cloud data.

[0053] Step 403: Based on the pedestrian's three-dimensional coordinate point cloud data, the inter-frame difference method is used to determine the pedestrian's movement speed, and together with the number of pedestrians in the upstream guidance area, they constitute the real-time passenger flow data of the upstream guidance area.

[0054] Based on this three-dimensional coordinate data, the number of pedestrians in the area can be calculated in real time, and the local movement speed of pedestrians can be accurately obtained through the inter-frame difference method. At the same time, by analyzing the trend of continuous frame coordinate data, the flow direction and aggregation pattern of passenger flow can be predicted in advance, providing early warning information for subsequent gate scheduling and other scenarios, realizing the early perception of passenger flow demand, and serving as a data source for predicting and supplementing passenger flow / demand.

[0055] The specific formula for calculating pedestrian speed using the inter-frame difference method of lidar is as follows: The lidar sampling frequency is f (unit: Hz), and the time interval between two adjacent frames is: Such as 10Hz radar .

[0056] Target clustering was performed on the point clouds of frame k and frame (k+1), and the pedestrian point cloud cluster with the number i was extracted: , ,in, This represents the number of laser points in the point cloud cluster. For laser points Three-dimensional coordinates in the radar coordinate system.

[0057] pedestrians The motion within the interval is considered as uniform linear motion (an approximation within a short time interval).

[0058] 1. Calculation of pedestrian target centroid coordinates First, calculate the centroid of the point cloud cluster of the pedestrian in two adjacent frames. The centroid is the optimal representative point of the target location, which can cancel out the interference of single-point noise. ; Symbol explanation: Let be the 3D coordinates of the centroid of pedestrian i in frame k; m is the number of laser points in the point cloud cluster of this pedestrian in frame (k+1). In the subway station gate scene, pedestrians only move horizontally, so the z-axis (height) component can be ignored, and only the 2D centroid (x,y) is calculated.

[0059] 2. Calculation of centroid displacement between adjacent frames Calculate pedestrian time Two-dimensional displacement vector within : ; The scalar magnitude of the displacement (i.e., the Euclidean distance) is: .

[0060] 3. Pedestrian speed calculation (1) Scalar velocity (commonly used, reflecting the magnitude of velocity): ; The unit of scalar velocity is m / s.

[0061] (2) Vector velocity (including direction of motion, used for trajectory analysis): ; Velocity direction angle (angle with radar coordinate system axis) : .

[0062] 4. Supplementary Formulas for Engineering Optimization Simple inter-frame difference is easily affected by point cloud noise and occlusion. In practical applications, it is necessary to combine the Kalman filter to smooth the velocity results and predict the optimal velocity of the (k+1)th frame. ; This is the Kalman gain, used to balance the weights of the prediction speed and the measurement speed.

[0063] By remotely monitoring passenger flow, the demand for entry gates can be predicted, and only the number of people is collected to protect personal identification information. This step can detect large waves of passengers about to arrive at the gate area in advance, providing valuable early warning time for the dynamic switching of gate modes, and realizing a leap from "passive response" to "proactive prediction".

[0064] Step 404: Collect video data of the target monitoring area of ​​the gate through a binocular camera; the binocular camera and the lidar are connected to the same external timing network.

[0065] Step 405: Determine the three-dimensional spatial information of the gate target monitoring area based on the video data of the gate target monitoring area.

[0066] Real-time passenger flow data is obtained by deploying binocular cameras: A binocular camera is a visual perception device that simulates the imaging mechanism of human binocular vision. The core consists of two synchronously calibrated cameras and an image acquisition and processing unit. Its core advantage is that it calculates the depth information of the scene through the principle of binocular visual stereo matching, and inversely calculates the distance between the target and the camera through stereo matching algorithms (SGBM algorithm and BM algorithm). At the same time, it outputs the two-dimensional coordinates of the target on the image plane, and finally realizes the perception of three-dimensional spatial information of "two-dimensional coordinates + depth".

[0067] Binocular cameras are deployed in areas such as boarding areas and security checkpoints to acquire depth information and two-dimensional coordinate data of pedestrians using the principle of binocular vision, complementing LiDAR data. At the same time, the depth information acquired can further assist in judging the distance between pedestrians and surrounding facilities, providing supplementary evidence for congestion identification and supplementing the amount of pedestrian flow data.

[0068] Step 406: Based on the video data of the gate target monitoring area, the heterogeneity index of pedestrians is determined by the adjacent frame difference method, and together with the three-dimensional spatial information of the gate target monitoring area, it constitutes the real-time passenger flow data of the gate target monitoring area.

[0069] The video data acquired by the binocular camera is processed using the frame difference method. First, noise reduction and grayscale preprocessing are used to reduce interference from ambient light changes and equipment noise. Then, the pixel difference between two consecutive frames is calculated using the adjacent frame difference method to filter out the moving target area to separate pedestrians from the background. Finally, image feature extraction algorithms are combined to determine heterogeneous indicators such as pedestrian age and whether they are carrying large items. Classification and statistics are completed by analyzing the contour features and size ratio of the target area, providing data support for refined passenger flow management.

[0070] Step 407: Align and integrate the real-time passenger flow data of the upstream guidance area and the real-time passenger flow data of the gate target monitoring area to form structured passenger flow data, which serves as the real-time passenger flow data of the transportation hub.

[0071] For example, the process of forming structured passenger flow data is as follows: Real-time passenger flow data from both the upstream guide area coordinate system and the gate target monitoring area coordinate system are mapped to the station's global coordinate system through external parameter calibration; a virtual cross-section leading to the gate is set in the upstream guide area, and the average speed of pedestrians crossing the virtual cross-section per unit time is calculated based on the real-time passenger flow data of the upstream guide area in the station's global coordinate system; the time delay for pedestrians to reach the gate in the upstream guide area is determined based on the preset path length and the average pedestrian speed; the time delay is added to the time corresponding to the real-time passenger flow data of the upstream guide area in the station's global coordinate system to obtain the real-time passenger flow data of the upstream guide area under the gate target monitoring area. The system analyzes passenger flow data. Based on real-time passenger flow data from the upstream guidance area within the gate target monitoring area, it determines the passenger flow, passenger speed, and passenger density. It analyzes the movement direction of pedestrians within the passenger flow speed and performs trend fitting on the number of pedestrians, passenger density, and movement direction across multiple consecutive frames to determine pedestrian aggregation trends. It extracts pedestrian category distribution from video data within the gate target monitoring area. Based on the pedestrian movement speed in the real-time passenger flow data from the upstream guidance area within the gate target monitoring area, and the video data from the gate target monitoring area, it determines abnormal pedestrian behavior. Finally, it combines the passenger flow, passenger speed, passenger density, pedestrian aggregation trends, pedestrian category distribution, and abnormal pedestrian behavior into structured passenger flow data.

[0072] To achieve spatiotemporal alignment of multi-source data and construction of scenes within the same region, the more detailed implementation process is as follows: 1) Unified time synchronization: Connect the lidar and binocular camera to the same external time synchronization network (NTP) and use millisecond-level timestamps to ensure that data from different devices can be aligned within the same time window ΔT.

[0073] 2) Unified spatial coordinates and area identification: Assign area_id to the preset monitoring areas within the station (such as "gate target monitoring area", "upstream guidance area", "station hall open area", etc.) and map the coordinate system of each device to the global coordinate system within the station through external parameter calibration; then complete spatial mapping and aggregation statistics based on area_id.

[0074] 3) Projection and time delay compensation at different physical locations: When the lidar is deployed in the upstream guidance area and the binocular camera is deployed in the gate target monitoring area, the upstream observation is projected onto the gate target monitoring area using the "virtual boundary + arrival time delay" method: A virtual cross-section / boundary line leading to the gate is set in the upstream area, and the passenger flow and average speed crossing the cross-section per unit time are counted, and the calculation is based on the preset passage path length L_path and average speed. Estimate the time delay to reach the gate The passenger flow and speed at upstream time t are mapped to the gate area time t+ΔT_delay, so that the outputs of devices at different locations form a consistent passenger flow status description under the unified scenario of "gate target monitoring area".

[0075] 4) Structured passenger flow data output: After alignment, the data is integrated to form structured passenger flow data, which includes at least {timestamp, station ID, area_id, passenger flow N_in / N_out, total number of people in the area N, average speed}. The following parameters are used for subsequent simulation, prediction, and control: velocity variance, passenger flow density ρ, area module M, mainstream direction proportion, pedestrian category distribution, and abnormal behavior identifier.

[0076] After alignment, the data is directly integrated to form structured passenger flow data and output. The final output includes real-time basic passenger flow information such as passenger flow in and out of the station, passenger flow speed, passenger flow density, pedestrian gathering trend, pedestrian category distribution, and intuitive abnormal behaviors such as sudden stagnation and rapid running.

[0077] 1. Pedestrian gathering trend 1) LiDAR clusters targets in continuous frame point clouds to count the number of pedestrians in the monitored area (entrance, elevator area, etc.) in real time. Combined with the pedestrian velocity vector (including the direction angle θ) calculated by the inter-frame difference method, it analyzes the consistency of pedestrian movement direction (such as most pedestrians facing the same gate area).

[0078] 2) Perform trend fitting on the number of pedestrians, density (number of people / monitored area area), and movement direction for multiple consecutive frames (e.g., 10-20 frames). If the trend shows "the number of people continues to increase + the density exceeds the threshold (e.g., ≥0.04 people / ft in the station hall area)", the trend will be determined. 2 The characteristics of "+concentrated movement direction" are used to determine "pedestrian gathering trend".

[0079] 2. Pedestrian Category Distribution 1) The video captured by the binocular camera is first denoised and grayscale preprocessed, and then the pedestrians and background are separated by the frame difference method. The contour features (such as height and body proportion) and size parameters (such as contour area) of each pedestrian target are extracted.

[0080] 2) Based on image feature extraction algorithms, combined with preset rules for classification: Age classification: "Adult / Child" is distinguished by silhouette height (below 1.2m is considered a child, 1.2m-1.8m is considered an adult) and head-to-body ratio. Determining whether someone is carrying large items: Distinguishing between "person carrying luggage" and "person without luggage" based on the irregularity and volume of the target's outline (combined with lidar point cloud cluster volume data). 3. Sudden stay 1) The lidar calculates the real-time scalar velocity (v=Δd / Δt) of each pedestrian using the inter-frame difference method, and combines it with Kalman filtering to smooth the velocity results and avoid noise interference.

[0081] 2) Set the judgment threshold: If the speed v of a pedestrian is less than 0.1 m / s and the displacement Δd is less than 0.03 m for 3 consecutive frames (0.3 seconds when the sampling frequency is 10 Hz), it is initially judged as "stationary".

[0082] 3) Verification using trajectory data from binocular cameras: If the pedestrian's two-dimensional coordinates do not change significantly within multiple consecutive frames, and the surrounding pedestrian density does not become extremely crowded (excluding "passive stagnation"), then it is determined to be "sudden stagnation".

[0083] 4. Running fast 1) The lidar calculates the scalar velocity of pedestrians in real time, and combines it with the optimal velocity value after Kalman filtering correction to set a judgment threshold: if the pedestrian speed v≥2.0m / s (far exceeds the normal walking speed of 1.1m / s), and maintains this speed level for 2 consecutive frames (0.2 seconds), it is initially judged as "fast movement".

[0084] 2) Verification using trajectory data from binocular cameras: If the pedestrian's movement trajectory is a straight line or approximately a straight line, and there is no frequent change in direction (excluding normal behaviors such as "quick avoidance"), then it is determined to be "running fast".

[0085] This process utilizes complementary integration of multi-source data from LiDAR, binocular cameras, and frame difference methods to overcome the limitations of data from a single device. At the same time, it strictly adheres to privacy protection regulations, extracting only the macroscopic attributes and behavioral characteristics of pedestrians and refraining from collecting sensitive information such as biometrics, thus ensuring the security of personal information.

[0086] In another exemplary embodiment of this application, step 104 described above may be replaced by steps 501 to 505.

[0087] Step 501: Take real-time passenger flow data from the same period over the past 3 months, real-time passenger flow data from the past hour, and external characteristics affecting passenger flow as input data.

[0088] Step 502: Based on the input data, use a Long Short-Term Memory (LSTM) network to predict passenger flow data for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes; the LSM network includes two hidden layers and employs an attention mechanism; the passenger flow data includes the real-time number of people entering the station and the real-time number of people exiting the station.

[0089] Step 503: Based on the ratio of real-time number of people entering the station to real-time number of people exiting the station in the passenger flow data for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes, obtain the passenger flow ratio for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes.

[0090] Step 504: Obtain the real-time passenger flow ratio based on real-time passenger flow data.

[0091] Step 505: The real-time passenger flow ratio is weighted and fused with the passenger flow ratios for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes to obtain the fused passenger flow ratios for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes; the passenger flow ratio represents the passenger flow pattern.

[0092] The collaborative implementation of long-term prediction using Long Short-Term Memory (LSTM) networks and ultra-real-time data simulation aims to: Firstly, use LSTM to predict passenger flow trends 15 minutes to 2 hours in the future, generating multi-period passenger flow evolution data. Then, use this data to drive ultra-real-time simulations using Anylogic, verifying the effectiveness of different gate control strategies in advance and providing forward-looking support for strategy library loading and gate pre-switching. Specifically: The LSTM long-term prediction module focuses on four key time nodes: 15 minutes, 30 minutes, 60 minutes, and 120 minutes in the future. It outputs core targets such as total passenger flow in and out of the station, one-way peak flow, flow direction ratio, and peak hours, as well as auxiliary targets such as pedestrian category distribution and density peak areas. The model inputs include historical passenger flow data from the past three months, real-time rolling passenger flow data from the past hour (including the number of people entering and exiting the station, passenger flow speed, passenger flow density, and passenger composition), and external features such as holidays, weather, and surrounding activities. A network structure with two LSTM hidden layers and an attention mechanism is used. The prediction results must have a historical error rate ≤10%. Subsequently, based on this prediction data and combined with the subway station physical model parameters (number of turnstiles, passage width, etc.) from the Anylogic simulation strategy library, the model is further optimized. Real-time calibration data from LiDAR and binocular cameras is used to drive Anylogic simulations with a super real-time speedup ratio of 1:5 to 1:10, selecting the optimal candidate strategy. Simultaneously, every 2 minutes, real-time passenger flow data (including passenger arrival characteristics and personnel composition characteristics) is compared with simulation parameters (Agent speed, quantity, arrival distribution, etc.), calculating the deviation rate (deviation rate = |simulation value - actual value| / actual value). If the deviation rate exceeds a preset threshold (e.g., 10%), the simulation parameters are adjusted proportionally. The pre-control execution time is dynamically adjusted by subtracting the gate scheme execution time (switching, crowd control, etc.) from the passenger arrival time predicted by LSTM, ensuring pre-control preparation is completed before the peak passenger flow arrives, and providing data support for subsequent gate control command generation and strategy library optimization.

[0093] The process of selecting the turnstile scheme based on the ratio of passenger flow entering and exiting the station is as follows: First, the core evaluation indicator is defined as the ratio of inbound to outbound passenger flow = real-time number of inbound passengers / real-time number of outbound passengers. Then, passenger flow data in the gate area is continuously acquired through LiDAR and binocular cameras. The system uses a 30-second sliding statistical time window to count the number of inbound and outbound passengers within the time window to obtain the real-time passenger flow ratio, and outputs the real-time passenger flow ratio in a rolling update cycle of 2 seconds. Combining the passenger flow trend curve for the next 15 minutes to 2 hours output by the LSTM long-term prediction model, a weighted fusion method with a weight of 0.6 for real-time data and 0.4 for predicted data is used to complete data calibration and avoid single data bias. Then, based on the fused passenger flow ratio, candidate solutions for the corresponding scenario are matched from the Anylogic simulation strategy library. Subsequently, 1 to 3 candidate solutions are imported into Anylogic and simulated with a social force model at a super real-time speedup ratio of 1:5 to 1:10. The optimal feasible solution is selected by weighted scoring based on indicators such as gate load (30% weight) and passage efficiency (25% weight). Then, control commands such as direction switching timing are sent to the gate control system, and personnel evacuation commands are pushed simultaneously. During execution, the passenger flow ratio is continuously monitored. After returning to equilibrium, bidirectional operation is gradually restored. If the imbalance continues, the optimal solution is executed again.

[0094] In another exemplary embodiment of this application, a method for evaluating the safety and comfort of a transportation hub (LOS) is provided, comprising: obtaining pedestrian density values ​​for different functional areas of the transportation hub; establishing a hierarchical correspondence between the pedestrian space service level score and the pedestrian density value for each functional area of ​​the transportation hub; determining the pedestrian space service level score for each functional area of ​​the transportation hub by searching the hierarchical correspondence based on the pedestrian density values ​​for different functional areas of the transportation hub; determining congested functional areas based on the pedestrian space service level scores for each functional area of ​​the transportation hub; and implementing flow restriction control by adjusting the gate direction for gates entering the congested functional areas to alleviate congestion.

[0095] A more detailed evaluation process is as follows: Step 1: Obtain the data source based on the first aspect. Includes trajectory data (X / Y coordinates, area labels); video counts (database time period, area, number of people, average speed); sensor data (turnstile throughput, escalator passenger flow, real-time density values). Step 2: Data Preprocessing 1) Outlier cleaning: Remove trajectories that have been stationary for more than 5 minutes, coordinate drift points, and duplicate counts; 2) Calculate the area density: Divide the station space into different areas and use sensors deployed in the station to measure the pedestrian density (unit: person / square meter) in each area. 3) Area matching: Associate the trajectory and counting data with the preset subway station functional areas (such as "Entrance / Exit 1" and "Platform 2 waiting area").

[0096] Step 3: Quantify the service level of the station's pedestrian space based on the density value.

[0097] Service level ratings are assigned to different scenarios based on the Fruit LOS threshold for different areas of the subway station: LOS is assigned in a graded manner (A=100, B=80, C=60, D=40, E=20, F=0).

[0098] Table 1. Correspondence between hierarchical levels

[0099] By integrating the Level of Service (LOS) score, the control results are fed back to the simulation library to form a closed-loop management system, with regular manual optimization. After identifying congested areas, flow control based on gate direction adjustment is implemented on the turnstiles entering those areas. Simulations verify the congestion mitigation effect of the turnstile scheme, thereby reducing crowd gathering in high-density areas and improving the LOS score. Specifically, if the score of a local area (such as narrow passages, corners, or pedestrian congestion zones) reaches a certain level... The score, or the station's overall average score is less than [number] points. The score, or the overall score standard deviation of the station is greater than The LOS density range for the area can be fine-tuned based on feedback data. The three threshold values ​​are determined by the managers of specific stations, and the specific values ​​can be set as intermediate values ​​after referring to the local and overall density values ​​for congested and non-congested periods.

[0100] In another exemplary embodiment of this application, reference is made to... Figure 3 Optimization of the simulation strategy library: 1) Iterate on existing scheme parameters: For existing scenario schemes in the strategy library, modify their core parameters (optimal gate direction ratio, switching sequence, and coordination of evacuation instructions) to improve the passage efficiency and safety of the scheme in actual application. 2) Eliminate inefficient solutions: If an existing solution exceeds the critical density index in multiple actual adjustments, it will be removed from the strategy library for this management to avoid subsequent calls.

[0101] This invention addresses the technical problem of existing turnstile systems' inability to intelligently adapt to tidal passenger flow, leading to both resource scarcity and waste. By integrating multi-source sensing (binocular vision, millimeter-wave radar, and AI algorithms) with Anylogic-based intelligent simulation decision-making, real-time, precise, and coordinated control of unidirectional and bidirectional turnstiles is achieved. This invention significantly improves turnstile resource utilization, effectively reduces passenger queue length and waiting time, enhances the overall operational efficiency of transportation hubs and passenger experience, and provides an innovative solution for intelligent traffic management.

[0102] Based on the same inventive concept, this application also provides a real-time gate control device that applies the aforementioned real-time gate control method. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the real-time gate control device provided below can be found in the limitations of the real-time gate control method described above, and will not be repeated here.

[0103] In one exemplary embodiment, a real-time gate control device is provided, comprising: a gate and a cloud; the cloud controls the gate to execute the optimal gate control scheme by performing the above-described real-time gate control method.

[0104] As an optional implementation, the cloud platform provides standardized interactive interfaces for data acquisition, data query, strategy invocation, and command issuance and status feedback. The standardized data acquisition interface is used to upload real-time passenger flow data from transportation hubs equipped with turnstiles; the standardized data query interface is used to query the real-time passenger flow data and future passenger flow data of the transportation hub; the standardized strategy invocation interface is used to invoke the optimal turnstile control scheme; and the standardized command issuance and status feedback interface is used to issue turnstile direction control commands and query the turnstile operating status.

[0105] The standardized API design for the turnstile system is as follows: Step 1: Preparatory Steps for API Design 1. Determine the API technical architecture It adopts a RESTful architecture, featuring statelessness, caching, and standardized request methods, making it suitable for multi-terminal and multi-system integration scenarios.

[0106] 2. Define the transmission protocol specifications (1) Sensitive data (such as gate control commands and equipment authentication information) are transmitted using HTTPS protocol to ensure data security; (2) Non-sensitive data (such as publicly available passenger flow query data) supports HTTP protocol transmission, taking into account both compatibility and low cost requirements.

[0107] 3. Unify data formats and coding standards (1) Data format: All interactive data adopts JSON format, the key name adopts the underscore naming method, and the field type is clearly defined (numeric, character, etc.). (2) Encoding standard: UTF-8 character encoding is adopted to avoid Chinese character garbled problems during cross-device interaction.

[0108] 4. Standardize timestamp format Use millisecond-level timestamps, based on the UTC+8 time zone, to ensure consistency with the spatiotemporal alignment requirements of passenger flow data.

[0109] 5. Set up security mechanisms (1) Use APIKey as the system interface credential; (2) Use an MD5 encryption signature mechanism with "APIKey + timestamp + request parameters" to prevent unauthorized access and data tampering; (3) The APIKey is set to be valid for 1 year and is updated regularly to enhance security.

[0110] Step 2: Develop data acquisition APIs and interface with multi-source sensing devices. 1. Development of LiDAR data upload interface (1) Define request parameters: including device ID, site ID, collection timestamp, Base64 encoded point cloud data compressed package, and device operating status; (2) Upload rules: Batch upload is supported, with a maximum of 10 frames of data that can be uploaded at a time. The request timeout is set to 30 seconds. (3) Response parameters: return data upload result, data verification status, and error message (such as data format error, device unauthorized, etc.).

[0111] 2. Development of data upload interface for binocular cameras (1) Define request parameters: including device ID, site ID, collection timestamp, depth image data, pedestrian 2D coordinate data, and pedestrian category data (standard fields: `age_group` (values ​​are adult / child), `luggage` (values ​​are yes / no)); (2) Response parameters: return the data entry status and data integrity verification results. If data is missing, return a supplementary collection suggestion simultaneously.

[0112] Step 3: Develop data query APIs (integrate with third-party platforms / operation systems) 1. Development of a real-time passenger flow status query interface (1) Parameter requirements: The required parameters are stationId (site ID), apiKey (connection certificate), and timestamp (time stamp); the optional parameters are areaIds (area ID) and indicators (query indicators). (2) Indicator filtering logic: Supports filtering returned fields by "number of people, speed, density, flow direction, LOS level, and security risk"; (3) Data processing requirements: Spatiotemporal alignment and outlier cleaning must be completed before returning data to ensure data consistency; (4) Response format: Data is returned in groups according to the monitoring area, including the core indicator values ​​and status descriptions.

[0113] 2. Development of a long-term passenger flow forecast query interface (1) Prediction time node limitation: Only passenger flow data at four fixed time nodes of 15 minutes, 30 minutes, 60 minutes and 120 minutes are supported; (2) Data source: Connect to the LSTM long-term prediction module to return prediction results and confidence scores (calculated based on historical errors); (3) Anomaly handling mechanism: When the prediction confidence is lower than 70%, an early warning message is returned and it is recommended to refer to real-time passenger flow data.

[0114] Step 4: Develop strategy invocation API (interfacing with cloud-based decision-making systems) 1. Request parameter validation: Verify the validity of the gate group ID; verify the correctness of the current gate operation strategy format.

[0115] 2. Strategy matching and filtering logic First, based on the passenger flow ratio after real-time and prediction fusion, candidate solutions for the corresponding scenario are matched from the Anylogic simulation strategy library; 1-3 candidate solutions are imported into the Anylogic simulation platform and simulation is carried out with a super real-time acceleration ratio of 1:5 to 1:10 combined with the social force model; weighted scoring is performed according to indicators such as gate load (30% weight) and passage efficiency (25% weight) to select the optimal feasible solution.

[0116] 3. Return Content Guidelines Includes details of the optimal strategy (single gate direction configuration), simulation verification indicators, and execution timing suggestions; if the simulation times out (exceeds 5 seconds), the default optimal solution will be returned, and background asynchronous simulation optimization will be triggered.

[0117] Step 5: Development of APIs for instruction issuance and status feedback (interfacing with gate terminals) 1. Development of the gate direction control command interface (1) Command verification: Verify the gate direction switching delay (≥3 seconds to avoid the risk of people being trapped) and the legality of the target direction; (2) Security verification: Control commands must be verified by a combination of "apiKey + timestamp + command content" to prevent illegal commands; (3) Response logic: Return instruction ID (for subsequent status query) and real-time execution status (processing / success / failed).

[0118] 2. Development of a gate operation status query interface (1) Status synchronization: Real-time data retrieval from the gate terminal, including the current direction of the gate, switch status, fault information, and passage count; (2) Filtering logic: Supports filtering and querying by gate ID and instruction ID, and returns the running status of all gates by default; (3) Fault handling: When the gate status is “fault”, return the fault type and preliminary handling suggestions.

[0119] The real-time control method and device for turnstiles in this application are applicable to transportation hubs such as subway stations and high-speed rail stations, and can dynamically adjust the passage direction of one-way and two-way turnstiles.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0121] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A real-time control method for a turnstile, characterized in that, include: Establish a simulation platform for transportation hubs equipped with turnstiles; Using the aforementioned transportation hub simulation platform, all gate control schemes for different passenger flow patterns are simulated one by one, and multiple candidate gate control schemes for each passenger flow pattern are selected to form a multi-scenario simulation strategy library. The gate control scheme is a hybrid single- and two-way configuration scheme for the gate; Continuously acquire real-time passenger flow data of the transportation hub; Based on real-time passenger flow data obtained from historical periods, long short-term memory networks are used to predict passenger flow data for future periods, and combined with real-time passenger flow data, passenger flow patterns for future periods are determined. Based on the passenger flow pattern in the future time period, multiple candidate gate control schemes are matched and obtained from the multi-scenario simulation strategy library, and the transportation hub simulation platform is used to perform simulations to select the optimal gate control scheme from the multiple candidate gate control schemes. In the future, the control gate will execute the optimal gate control scheme.

2. The real-time control method for turnstiles according to claim 1, characterized in that, Establish a transportation hub simulation platform with turnstiles, specifically including: A basic model of the transportation hub was established using the Anylogic multi-agent simulation platform. The underlying logic of pedestrians was simulated using the social force model built into the Anylogic multi-agent simulation platform. The Monte Carlo method is embedded in the Anylogic multi-agent simulation platform, and random settings are used to simulate passenger flow demand with different total passenger volume and entry / exit ratio, thereby building a transportation hub simulation platform for passenger flow prediction and gate control scheme selection.

3. The real-time control method for turnstiles according to claim 1, characterized in that, Using the aforementioned transportation hub simulation platform, all gate control schemes for different passenger flow patterns are simulated one by one. Multiple candidate gate control schemes for each passenger flow pattern are then selected, forming a multi-scenario simulation strategy library, specifically including: Using the aforementioned transportation hub simulation platform, all gate control schemes under different passenger flow ratios are simulated one by one to obtain the index data of each gate control scheme under each passenger flow ratio; the passenger flow ratio represents the passenger flow pattern; the index includes traffic efficiency index and safety congestion index. The passage efficiency index data and safety congestion index data of each gate control scheme under each passenger flow ratio are weighted and summed to obtain the quantitative evaluation index data of each gate control scheme under each passenger flow ratio. Identify all gate control schemes with the largest quantitative evaluation index data under each passenger flow ratio, and use them as multiple candidate gate control schemes under each passenger flow ratio; Based on the one-to-one correspondence between passenger flow ratio, candidate gate control schemes, and quantitative evaluation index data, a multi-scenario simulation strategy library is established.

4. The real-time control method for turnstiles according to claim 3, characterized in that, The traffic efficiency indicators include: average queue length, maximum queue length, and total number of people passing through the turnstile in one hour; The safety congestion indicators include: the probability of congestion in front of the turnstiles and the passenger flow conflict rate.

5. The real-time control method for turnstiles according to claim 1, characterized in that, Continuously acquire real-time passenger flow data of the transportation hub, specifically including: The three-dimensional coordinate point cloud data of pedestrians in the upstream guidance area is monitored using lidar. The number of pedestrians in the upstream guidance area is determined based on the pedestrian's three-dimensional coordinate point cloud data. Based on the three-dimensional coordinate point cloud data of pedestrians, the inter-frame difference method is used to determine the movement speed of pedestrians, and together with the number of pedestrians in the upstream guidance area, they constitute the real-time passenger flow data of the upstream guidance area. Video data of the target monitoring area of ​​the gate is collected by a binocular camera; the binocular camera and the lidar are connected to the same external timing network. Based on the video data of the gate target monitoring area, determine the three-dimensional spatial information of the gate target monitoring area; Based on the video data of the gate target monitoring area, the adjacent frame difference method is used to determine the heterogeneity index of pedestrians, and together with the three-dimensional spatial information of the gate target monitoring area, they constitute the real-time passenger flow data of the gate target monitoring area. The real-time passenger flow data of the upstream guidance area and the real-time passenger flow data of the gate target monitoring area are aligned and integrated to form structured passenger flow data, which serves as the real-time passenger flow data of the transportation hub.

6. The real-time control method for a turnstile according to claim 5, characterized in that, The real-time passenger flow data from the upstream guidance area and the real-time passenger flow data from the gate target monitoring area are aligned and integrated to form structured passenger flow data, specifically including: By calibrating external parameters, the real-time passenger flow data of the upstream guide area coordinate system and the real-time passenger flow data of the gate target monitoring area coordinate system are both mapped to the station global coordinate system. A virtual cross-section leading to the gate is set in the upstream guidance area, and the average speed of pedestrians crossing the virtual cross-section is calculated per unit time based on the real-time passenger flow data of the upstream guidance area in the global coordinate system within the station. Based on the preset passage path length and the average speed of pedestrians, the time delay for pedestrians to reach the gate in the upstream guidance area is determined; The real-time passenger flow data of the upstream guide area in the station's global coordinate system is obtained by adding the time delay to the real-time passenger flow data of the upstream guide area in the gate target monitoring area. Based on the real-time passenger flow data of the upstream guidance area under the gate target monitoring area, the passenger flow, passenger flow speed and passenger flow density in and out of the station are determined; The movement direction of pedestrians in the passenger flow speed is analyzed, and the trend of pedestrian number, passenger flow density and movement direction in multiple consecutive frames is fitted to determine the pedestrian aggregation trend. Based on the video data of the target monitoring area of ​​the turnstile, extract the pedestrian category distribution; Based on the pedestrian movement speed in the real-time passenger flow data of the upstream guidance area under the gate target monitoring area, and the video data of the gate target monitoring area, abnormal pedestrian behavior is determined; The data on passenger flow, passenger speed, passenger density, pedestrian aggregation trends, pedestrian category distribution, and abnormal pedestrian behavior are combined to form structured passenger flow data.

7. The real-time control method for turnstiles according to claim 1, characterized in that, Based on real-time passenger flow data acquired from historical periods, long short-term memory networks are used to predict passenger flow data for future periods. By integrating real-time passenger flow data, future passenger flow patterns are determined, specifically including: The input data includes real-time passenger flow data from the same period over the past three months, real-time passenger flow data from the past hour, and external characteristics that affect passenger flow. Based on the input data, a Long Short-Term Memory (LSTM) network is used to predict passenger flow data for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes. The LSM network includes two hidden layers and employs an attention mechanism. The passenger flow data includes the real-time number of people entering and exiting the station. Based on the ratio of real-time number of people entering the station to real-time number of people exiting the station in the passenger flow data for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes, the passenger flow ratio for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes can be obtained. Based on real-time passenger flow data, obtain the real-time passenger flow ratio; The real-time passenger flow ratio is weighted and integrated with the passenger flow ratios for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes to obtain the integrated passenger flow ratios for the next 15 minutes, 30 minutes, 60 minutes, and 120 minutes; the passenger flow ratio represents the passenger flow pattern.

8. The real-time control method for turnstiles according to claim 1, characterized in that, Also includes: Obtain pedestrian density values ​​for different functional areas of the transportation hub; Establish a hierarchical correspondence between the pedestrian space service level score and the pedestrian density value for each functional area of ​​a transportation hub; Based on the pedestrian density values ​​of different functional areas of the transportation hub, the pedestrian space service level score of each functional area of ​​the transportation hub is determined by finding the corresponding hierarchical relationship. Based on the pedestrian space service level score of each functional area of ​​the transportation hub, the congested functional areas are determined; Flow control is implemented by adjusting the direction of the turnstiles entering the congested functional area to alleviate congestion.

9. A real-time control device for a turnstile, characterized in that, include: turnstiles and the cloud; The cloud platform controls the gate to execute the optimal gate control scheme by implementing the gate real-time control method according to any one of claims 1-8.

10. The real-time control device for the turnstile according to claim 9, characterized in that, The cloud platform is equipped with standardized interactive interfaces for data collection, data query, strategy invocation, and command issuance and status feedback. Standardized interactive interfaces for data collection are used to upload real-time passenger flow data from transportation hubs equipped with turnstiles; The standardized interactive interface for data query is used to query the real-time passenger flow data and future passenger flow data of the transportation hub. The standardized interaction interface for strategy invocation is used to invoke the optimal gate control scheme; The standardized interactive interface for issuing instructions and providing status feedback is used to issue gate direction control instructions and query the gate's operating status.

Images