Comprehensive transportation junction passenger flow intelligent analysis and dynamic early warning method

Abstract

The invention provides a comprehensive transportation junction passenger flow intelligent analysis and dynamic early warning method, the method is specially suitable for a comprehensive transportation junction scene of a multi-transportation-mode connection area, a three-dimensional channel and high-density passenger flow nodes, and the method comprises the following steps: arranging a multi-modal sensing network in a three-dimensional space of a comprehensive transportation junction, comprising the following steps: deploying a three-dimensional laser scanning array at a vertical traffic node, and synchronously capturing multi-layer people flow motion states of an escalator and stairs; a distributed inertial sensor matrix is embedded in the ground of the transfer hall, and vibration waveforms caused by the crowd movement trend are monitored; configuring a transmission type millimeter wave radar cluster in a security check area, and synchronously detecting personal belongings characteristics and a human body behavior mode; a hub exclusive space-time coupling field model is established, a hub space is divided into function-associated dynamics units, and each unit comprises a coupling effect relationship of a rail transit connection area, a commercial service area and a vertical commuting channel.

Description

Technical Field

[0001] The present invention belongs to the field of dynamic early warning, and in particular relates to a method for intelligent analysis and dynamic early warning of passenger flow in a comprehensive transportation hub. Background Art

[0002] Currently, passenger flow management at integrated transportation hubs generally relies on a traditional early warning model based on video surveillance and gate counting. This technology system has exposed a series of systemic flaws when dealing with the three-dimensional spatial structure of modern hubs, the integration of multiple transportation modes, and the sudden large passenger flow scenarios. Existing technologies primarily rely on two-dimensional video analysis algorithms to count pedestrian density in planar planes. This is unable to effectively capture the three-dimensional spatial dynamic characteristics of vertical transportation nodes such as escalators and corridors, making it difficult to promptly identify congestion risks caused by cross-level pedestrian interactions. At the perception level, single visible light cameras are susceptible to interference from factors such as ambient lighting changes and visual occlusion, resulting in a significant increase in false detection rates in rainy, snowy weather, or low-light conditions at night. Furthermore, solutions using millimeter-wave radar for auxiliary monitoring lack multi-physics field data fusion mechanisms, making it difficult to distinguish between abnormal individual behavior and normal luggage status in dense crowds, resulting in underreporting of safety hazards.

[0003] Data processing architectures often utilize a centralized cloud computing model, requiring raw video streams and sensor data to be transmitted over long distances to cloud servers. Network latency and bandwidth bottlenecks result in decision lags exceeding 30 seconds between data collection and the issuance of warning instructions, making them unable to meet the real-time response requirements in scenarios such as dense subway train arrivals and departures and sudden flight delays. Existing systems generally rely on fixed threshold rules based on historical statistical data, such as setting a maximum number of people per unit area to trigger an alert. This static model struggles to adapt to dynamic scenarios such as holiday passenger flow fluctuations and temporary traffic restrictions. Furthermore, it is unable to predict passenger flow phase changes caused by commercial promotions or emergencies, often leading to extreme cases of over-warning and under-response.

[0004] When it comes to multimodal data fusion, traditional methods employ simple weighted averages or logic gates, lacking in-depth exploration of spatiotemporal correlation features. For example, a dynamic correlation model cannot be established between crowd movement trends identified through video and cadence information detected by vibration sensors, resulting in an accuracy rate of less than 60% for early risk assessment. Regarding system adaptability, existing early warning algorithms are fixed once deployed and are unable to autonomously update their decision-making logic based on changes in hub operating models (such as new subway lines or commercial area renovations). Manual parameter recalibration is required, resulting in high maintenance costs and poor timeliness.

[0005] Furthermore, traditional technical solutions offer limited support for coordinated early warning across multiple modes of transport. Passenger flow data from systems like railways, aviation, and urban transportation exists in isolation within vertical business platforms, lacking information fusion and joint decision-making mechanisms across operating entities. When large-scale train delays trigger a chain reaction, it's impossible to quickly predict the impact on surrounding road networks, nor can it dynamically adjust the allocation of key resources like security checkpoints and connecting transport capacity. Digital twin technology has begun pilot applications in some advanced systems, but existing implementations are limited to three-dimensional visualization. There's a lack of two-way interaction between virtual models and physical entities, making it impossible to inject real-time sensor data into digital space for stress testing, or to reverse-deploy simulation optimization strategies to actual systems. Consequently, the predictive and early warning value of digital twins has been effectively unleashed.

[0006] In terms of energy efficiency, traditional solutions require the continuous operation of high-performance server clusters to maintain 24 / 7 video analysis. The average annual electricity consumption of a single medium-sized hub exceeds 100,000 kWh, significantly conflicting with the goal of green and low-carbon operations. These shortcomings have hindered the existing technology system from achieving breakthroughs in key metrics such as early warning accuracy, response timeliness, and scenario adaptability, making it unable to meet the urgent need for intelligent and safe operations in modern integrated transportation hubs. Summary of the Invention

[0007] The present invention proposes an intelligent analysis and dynamic early warning method for passenger flow in comprehensive transportation hubs. This technical solution solves the problem that traditional early warning systems are unable to effectively handle three-dimensional spatial structures, coordination of multiple transportation modes, high-density dynamic passenger flows and sudden abnormal events in comprehensive transportation hubs. Specifically, these problems include insufficient three-dimensional dynamic capture of vertical transportation nodes, delayed identification of cross-regional passenger flow coupling risks, poor adaptability of static threshold models, insufficient depth of multimodal data fusion, and lack of cross-departmental collaborative response mechanisms. These are the core bottlenecks.

[0008] The technical solution of the present invention is achieved by: a method for intelligent analysis and dynamic early warning of passenger flow in a comprehensive transportation hub, which is specifically applicable to comprehensive transportation hub scenarios such as multi-transportation connection areas, three-dimensional channels, and high-density passenger flow nodes. The method comprises the following steps:

[0009] A multimodal sensing network is deployed within the three-dimensional space of integrated transportation hubs. This includes deploying three-dimensional laser scanning arrays at vertical transportation nodes to simultaneously capture the multi-layered movement of people on escalators and stairs; embedding a distributed inertial sensor matrix on the floor of transfer halls to monitor vibration waveforms caused by crowd movement trends; and deploying penetrating millimeter-wave radar clusters in security checkpoints to simultaneously detect the characteristics of personal belongings and human behavior patterns.

[0010] A hub-specific spatiotemporal coupling field model is established, dividing the hub space into functionally related dynamic units. Each unit contains the coupling relationship between the rail transit connection area, the commercial service area, and the vertical commuter corridor. Based on the improved equation system, the passenger flow evolution equation under the coordinated action of multiple transportation modes is constructed, in which the electromagnetic field term represents the gravitational effect of cross-layer passenger flow, and the fluid field term describes the viscous resistance of the same-layer corridor.

[0011] Implement dynamic risk field quantitative analysis to calculate the potential energy gradient between functional units and identify congestion risks caused by cross-regional passenger flow attraction; detect abnormal passenger flow vortices at transfer nodes through vorticity tensor analysis; and integrate multi-mode timetable data to predict the impact of external input flow on the internal potential energy field.

[0012] When the potential energy of a vertical commuter corridor exceeds its limit, a hierarchical coordinated response strategy is generated to dynamically adjust the direction and speed of escalators to balance the pressure difference between floors. When abnormal vortexes are detected in the transfer area, adjacent gate groups are linked to implement fuse-type passenger flow interception. Based on the multi-modal coordination agreement, timetable adjustment suggestions are sent to related operators to mitigate external passenger flow shocks.

[0013] Through the digital twin system, the evolution of the entire hub situation is realized, and a virtual hub environment that includes rail transit arrival and departure and commercial service attractions is constructed; historical passenger flow data and extreme scenarios generated by adversarial means are injected to train the reinforcement learning strategy library to adapt to the complex working conditions of multiple transportation modes coupling; the mapping relationship between field equation parameters and response strategies is iteratively updated daily.

[0014] While existing technologies primarily rely on flat cameras or single sensors, this solution builds a fully 3D spatial perception system through the coordinated deployment of a 3D laser scanning array, a distributed inertial sensor matrix, and penetrating millimeter-wave radar. Laser scanning arrays at vertical traffic nodes capture multi-layered pedestrian movement on escalators and stairs in a dynamic orthogonal configuration, overcoming the limitations of traditional 2D vision, which cannot analyze vertical pedestrian interactions. Inertial sensors embedded in the transfer hall floor analyze vibration waveform spectra to predict group movement trends in advance. Compared to traditional image-based motion tracking technology, these sensors can operate stably in complete darkness or visually obstructed scenes. The millimeter-wave radar cluster in the security check area uses synthetic aperture technology to simultaneously analyze human behavior patterns and object geometric features, addressing the shortcomings of traditional security equipment, which can only detect single points and cannot associate behavior with objects.

[0015] Traditional methods use statistics or simple mechanical models to describe passenger flow, but this solution innovatively introduces a modified Maxwell equations and fluid dynamics model to construct a hub-specific space-time coupling field model. The electromagnetic field term quantifies the gravitational effect of cross-layer passenger flow (such as the attraction of commercial areas to rail transit passengers), and the fluid field term characterizes the viscous resistance of channels on the same level (such as the friction effect of people moving in corridors). This is the first time to achieve a deep integration of physical field theory and passenger flow dynamics. The model divides the hub space into functionally related dynamic units such as rail transit docking areas and commercial service areas. It describes the complex coupling relationship between multiple transportation modes through non-Euclidean geometric topological mapping. Compared with the traditional uniform grid division method, it is more in line with the actual operation scenario of the hub.

[0016] Existing technologies rely on a single density threshold to trigger early warnings. This solution proposes a risk quantification system that integrates potential energy gradient, vorticity tensor, and external disturbances. Potential energy gradient analysis is used to identify congestion risks caused by cross-regional passenger flow attraction (such as the impact of subway passenger flow on commercial areas), and vorticity tensors are used to detect abnormal passenger flow vortices at transfer nodes (such as queues at security checkpoints). Furthermore, the system integrates multi-mode timetable data to predict the disturbance of external input flow on the internal potential energy field (such as the chain reaction caused by flight delays). Compared with traditional static models, this analysis system can dynamically capture the phase change process of passenger flow and achieve accurate attribution of risk causes.

[0017] Unlike traditional single-point response models, this solution incorporates a coordinated response mechanism across facilities and levels. When potential energy in a vertical commuter corridor exceeds its limit, escalator speeds are dynamically adjusted to balance inter-level pressure differences, rather than simply shutting down the equipment. When abnormal vortexes are detected in transfer areas, adjacent gate groups are linked to implement fuse-type flow cutoffs to prevent the spread of localized congestion. Based on a multi-modal collaborative agreement, schedule adjustment recommendations (such as delaying subway departure intervals) are sent to associated operators to mitigate external passenger flow shocks at the source. This multi-level coordinated strategy overcomes the limitations of traditional systems, which can only control local facilities.

[0018] Existing digital twin technologies are mostly limited to visual display. This solution constructs a heterogeneous event-driven virtual hub environment: a discrete event engine is used to simulate rail transit arrivals and departures, a continuous Poisson process is used to model commercial appeal, and extreme scenarios (such as terrorist attacks and extreme weather) generated by adversarial forces are injected for stress testing. The reinforcement learning policy library iteratively updates the mapping relationship between field equation parameters and response strategies daily, enabling the system to adapt to changes in scenarios such as hub renovations and new line openings, while traditional systems require manual parameter recalibration. In addition, the policy evolution graph extracts key decision paths through topological persistent homology analysis, supporting the derivative optimization of cross-transportation mode collaborative strategies. This is a closed-loop evolution capability that cannot be achieved with existing technologies.

[0019] As a preferred embodiment, the deployment of the multimodal perception network includes setting up a multispectral LiDAR array on each transition platform of the vertical transportation node, with its scanning plane dynamically orthogonal to the escalator running trajectory, and synchronously acquiring the three-dimensional skeleton characteristics and thermal radiation distribution of pedestrians through wavelength division multiplexing technology;

[0020] The distributed inertial sensor matrix is ​​embedded in the transfer hall ground substrate in a non-uniform grid format. Each node is equipped with a triaxial MEMS accelerometer and a piezoelectric vibration sensor. The time-frequency domain coherence analysis is used to decouple individual gait vibration from group movement trends.

[0021] The millimeter-wave radar cluster uses a MIMO antenna architecture to form a synthetic aperture, performs multi-perspective penetrating scans of security inspection channels, and combines a polarization scattering feature classification algorithm to distinguish the geometric configuration and motion trajectory of metal objects, generating a threat assessment vector that integrates multiple physical quantities.

[0022] As a preferred embodiment, the establishment of the space-time coupling field model includes dividing the dynamic units into three types of interactive modules: rail transit-dominated, road passenger transport-coupled, and commercial service adsorption based on the topological connection strength of multiple transportation modes in the hub; the improved set of equations introduces non-Euclidean topological mapping, in which the electric field intensity component represents the passenger flow gravitational potential gradient in the waiting hall across transportation modes, and the magnetic flux density component describes the passenger flow curl constraint in the three-dimensional corridor; the fluid field term adopts a variable viscosity coefficient model to define the viscous resistance coefficient tensor of the interlayer channel, and its component values ​​are dynamically correlated with the escalator operation status and the real-time passenger flow density ratio.

[0023] As a preferred implementation, the digital twin system constructs a virtual hub environment driven by heterogeneous events, in which rail transit arrival and departure events are simulated using a discrete event-driven engine, and commercial service attraction is modeled as a continuous space-time Poisson process; the extreme scenarios generated by the adversarial method include two core stress test cases: timetable conflict chain reaction and commercial area oversaturation adsorption, and introduces stochastic differential equations to describe the passenger flow phase change process under extreme weather conditions; by establishing a dynamic strategy evolution map, the key decision paths in the reinforcement learning strategy library are identified through topological persistent homology analysis, and a derived strategy set for cross-transportation mode collaborative response is synthesized based on conditional generative adversarial networks.

[0024] As a preferred implementation, the dynamic risk field quantitative analysis expands the analysis scope of passenger flow attraction potential to a multi-dimensional space that includes schedule changes by constructing a comprehensive data integration module, and determines the critical conditions for cross-floor passenger flow attraction through stability criteria; an adaptive curl analysis method is used at transfer nodes to decompose the rotational component and diffusion component of the passenger flow vortex; a disturbance prediction model based on state transition is designed, and by analyzing the correlation between transportation vehicle arrival events and commercial activities, multi-scenario probability prediction results of external passenger flow input are generated.

[0025] After adopting the above technical solution, the beneficial effects of the present invention are as follows: This technical solution realizes full coverage monitoring of the three-dimensional space of the integrated transportation hub through a multimodal stereoscopic perception network, solving the problem of traditional two-dimensional visual systems monitoring blind spots in vertical commuting nodes, and improving the detection rate of abnormal events; the spatiotemporal coupled field model introduces electromagnetic field and fluid mechanics theory into passenger flow dynamics analysis for the first time, and can accurately quantify multi-physics field effects such as cross-layer gravity and channel viscous resistance, thereby improving the risk prediction accuracy compared with traditional statistical models;

[0026] The multi-dimensional quantification system for dynamic risk fields uses potential energy gradients, vorticity tensors, and external disturbance analysis to identify potential congestion points 10-15 minutes in advance, offering significant advantages over the delayed alarm modes of existing technologies. A hierarchical collaborative response strategy overcomes the limitations of single-point control. Through multi-level interventions such as dynamic escalator speed regulation, gate fuse interruption, and cross-operator timetable coordination, it improves the efficiency of handling high-passenger flow scenarios while minimizing disruption to normal operations. The self-evolving system driven by digital twins uses daily adversarial training and strategy iteration to shorten the deployment cycle for adapting to changing scenarios such as hub renovations and new line openings from the traditional weeks to within 24 hours. Furthermore, the coverage of extreme scenarios generated by the virtual environment is increased, significantly enhancing system robustness.

[0027] The multi-transportation coordination protocol breaks down data barriers between rail, aviation, and urban transportation. In emergencies like train delays, it dynamically adjusts resource allocation, including security checkpoints and connecting transport capacity, mitigating the risk of cross-system chain reactions. Regarding energy efficiency, the edge-fog-cloud collaborative computing architecture rationally distributes data processing loads, reducing power consumption compared to traditional centralized cloud computing models and aligning with the goal of building a green hub. The overall solution forms a complete technical loop of "perception-modeling-warning-response-evolution," providing a next-generation solution for the intelligent operation of integrated transportation hubs. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0031] Example:

[0032] like Figure 1 As shown, the method for intelligent analysis and dynamic early warning of passenger flow in integrated transportation hubs is specifically applicable to integrated transportation hub scenarios such as multi-transportation connection areas, three-dimensional channels, and high-density passenger flow nodes. The method includes the following steps:

[0033] A multimodal sensing network is deployed within the three-dimensional space of integrated transportation hubs. This includes deploying three-dimensional laser scanning arrays at vertical transportation nodes to simultaneously capture the multi-layered movement of people on escalators and stairs; embedding a distributed inertial sensor matrix on the floor of transfer halls to monitor vibration waveforms caused by crowd movement trends; and deploying penetrating millimeter-wave radar clusters in security checkpoints to simultaneously detect the characteristics of personal belongings and human behavior patterns.

[0034] A hub-specific spatiotemporal coupling field model is established, dividing the hub space into functionally related dynamic units. Each unit contains the coupling relationship between the rail transit connection area, the commercial service area, and the vertical commuter corridor. Based on the improved equation system, the passenger flow evolution equation under the coordinated action of multiple transportation modes is constructed, in which the electromagnetic field term represents the gravitational effect of cross-layer passenger flow, and the fluid field term describes the viscous resistance of the same-layer corridor.

[0035] Implement dynamic risk field quantitative analysis to calculate the potential energy gradient between functional units and identify congestion risks caused by cross-regional passenger flow attraction; detect abnormal passenger flow vortices at transfer nodes through vorticity tensor analysis; and integrate multi-mode timetable data to predict the impact of external input flow on the internal potential energy field.

[0036] When the potential energy of a vertical commuter corridor exceeds its limit, a hierarchical coordinated response strategy is generated to dynamically adjust the direction and speed of escalators to balance the pressure difference between floors. When abnormal vortexes are detected in the transfer area, adjacent gate groups are linked to implement fuse-type passenger flow interception. Based on the multi-modal coordination agreement, timetable adjustment suggestions are sent to related operators to mitigate external passenger flow shocks.

[0037] Through the digital twin system, the evolution of the entire hub situation is realized, and a virtual hub environment that includes rail transit arrival and departure and commercial service attractions is constructed; historical passenger flow data and extreme scenarios generated by adversarial means are injected to train the reinforcement learning strategy library to adapt to the complex working conditions of multiple transportation modes coupling; the mapping relationship between field equation parameters and response strategies is iteratively updated daily.

[0038] This application document lays out a heterogeneous sensing layer within the hub's three-dimensional space: Three-dimensional laser scanning arrays are deployed at vertical transportation nodes, using dynamic beamforming technology to capture the multi-layered motion vector field of pedestrian flows along escalators and stairs. A distributed inertial sensor matrix is ​​embedded in the transfer hall floor, decoupling individual gait from group motion modalities based on time-frequency analysis of vibration waveforms. A millimeter-wave radar cluster is deployed in the security checkpoint area, using polarization interferometry to simultaneously analyze human behavior patterns and object scattering characteristics, forming a multi-physics fusion threat assessment matrix. After spatiotemporal alignment of the sensor data at the edge computing node, a multidimensional feature tensor stream with a timestamp is generated.

[0039] The hub space is divided into functionally related dynamic units, such as rail transit connection cores and commercial attraction rings, and asymmetric topological connections are constructed. A cross-layer passenger flow gravitational potential model is established based on a modified Maxwell equation system, in which the commercial service attraction is modeled as a time-varying electric potential source. Non-Newtonian fluid constitutive equations are introduced to characterize the viscous resistance of intra-layer channels, and escalator operating parameters are used as shear rate adjustment terms. Hamiltonian Monte Carlo sampling is used to calculate the potential energy gradient between units in real time, and the perturbation parameters of multi-mode transport schedules are integrated to predict the impact of external input flow on the internal field.

[0040] The vorticity transport equation is used to analyze the spiral intensity and diffusion rate of passenger vortices at transfer nodes, and the Lyapunov stability criterion is combined to identify critical states of cross-regional passenger flow attraction. A multi-level risk decision tree is constructed: exceeding the vertical commuter potential energy limit triggers an escalator group control strategy, balancing inter-level pressure differences through dynamic speed regulation; when abnormal vortexes reach a threshold, a fuse mechanism is activated, linking adjacent gates to implement graded flow interception; a three-dimensional holographic guidance system is simultaneously activated to isolate risky paths, and encrypted coordinated control instructions are sent to related operators.

[0041] A heterogeneous event-driven virtual hub environment was constructed, with a discrete event engine simulating rail transit arrival and departure disturbances and a continuous Poisson process modeling the attraction effect of commercial passenger flow. Stress testing was conducted by injecting extreme scenarios generated through adversarial means, and the control strategy was iteratively optimized using a deep deterministic policy gradient algorithm. A dynamic policy evolution graph was established, and topological persistent homology analysis was used to extract key decision paths. Combined with a conditional generative adversarial network, a coordinated response plan across transport modes was synthesized, achieving adaptive updates of the mapping between field equation parameters and policies.

[0042] The system breaks through the traditional single-point response model through a closed-loop system that integrates real-time data injection at the perception layer, parallel solution of field equations at the fog computing layer, dynamic assessment of risk decision trees, coordinated control of multi-level actuators, and reverse optimization of digital twins. In complex coupled scenarios such as high-speed rail delays and commercial promotions, it enables full-process intelligent management and control, encompassing three-dimensional spatial situational awareness, dynamic deduction of multiple physical fields, cross-level resource optimization, and generalized strategy for unknown scenarios, significantly enhancing hub operational resilience.

[0043] Taking a specific practical scenario as an example: In the scenario of a chain of passenger flow backlogs caused by sudden multiple flight delays at a large international airport terminal, this method uses a three-dimensional laser scanning array deployed in the check-in hall to capture the cross-layer passenger flow density distribution of escalators and corridors on each floor in real time. Combined with the ground-embedded inertial sensor matrix, it analyzes the vibration spectrum characteristics of the group movement trend. Simultaneously, a millimeter-wave radar cluster penetration scan is used in the security check area to identify the micro-motion behavior patterns of carriers of abnormal stranded items. After spatiotemporal registration of the multimodal perception data at the edge nodes, a six-dimensional spatiotemporal feature tensor is constructed. Based on the improved spatiotemporal coupling field model of the Maxwell equations, the terminal is divided into three major dynamic units: the check-in core, the commercial ring, and the boarding corridor. The flight delay event is modeled as a time-varying electric potential disturbance source. The gravitational potential field of the commercial area and the viscous resistance field of the security check area form a dynamic equilibrium equation. The passenger flow potential energy gradient between the check-in core and the boarding corridor is calculated through Hamiltonian Monte Carlo sampling to predict the risk of cross-layer gravitational imbalance caused by the delay. When the vorticity tensor analysis detects that the spiral intensity exceeds the limit in the security check area, the terminal is divided into three dynamic units: the check-in core, the commercial ring, and the boarding corridor. When a vortex of passenger flow occurs, the system triggers a three-level fuse mechanism. The first response dynamically reduces the operating speed of adjacent escalator groups to reduce the pressure difference between passenger flows between floors. The second response activates electronic fences to guide some passengers to detour through the commercial buffer zone. The third response sends a request for increased connecting capacity to the city transportation hub through the air-rail intermodal transport agreement. At the same time, the trajectory of the carrier of abnormal items identified by millimeter-wave radar is matched with encrypted feature vectors, and then linked to the three-dimensional holographic guidance system to generate an isolation path and synchronize it with the security robot cluster. The digital twin system loads real-time delay data and adversarially generated blizzard extreme weather scenarios, optimizes the dynamic allocation strategy of check-in counters through a deep deterministic policy gradient algorithm, and extracts key decision paths for coordinated scheduling across transport modes based on topological persistent homology analysis. It reversely updates the commercial adsorption coefficient and security check efficiency parameters in the field equation, forming a full-link closed-loop disposal process from multi-physical field perception, cross-layer situation deduction, multi-level coordinated control, to digital evolution feedback, ultimately achieving the dynamic resolution of the three-dimensional passenger flow crisis caused by flight delays before it causes large-scale congestion.

[0044] On each transition platform of the vertical transportation nodes of the high-speed rail hub station, a multispectral LiDAR array scans the escalator running trajectory in a dynamic orthogonal configuration, and synchronously captures the three-dimensional skeletal characteristics and thermal radiation distribution of pedestrians through wavelength division multiplexing technology, accurately identifying areas of abnormal cross-floor passenger density; a distributed inertial sensor matrix embedded in a non-uniform grid on the ground of the transfer hall uses three-axis MEMS accelerometers and piezoelectric vibration sensors to collaboratively collect vibration waveforms, and uses time-frequency domain coherence analysis to separate individual high-frequency gait signals from group low-frequency motion trends, predicting the risk of large passenger flow gathering in advance; the millimeter-wave radar cluster deployed in the security check channel uses a MIMO antenna architecture to form a synthetic aperture, scanning human bodies and object targets from multiple angles, and combines a polarization scattering feature classification algorithm to distinguish between the normal metal structure of luggage and the geometric configuration of dangerous objects, generating a multi-physical quantity threat assessment vector that integrates motion trajectory, polarization characteristics, and energy reflection intensity, to achieve early warning of abnormal object carrying behavior.

[0045] In view of the coupling characteristics of multiple transportation modes such as rail transit, road passenger transport and commercial services in the integrated hub, the space is divided into three types of interactive modules based on the strength of topological connections: rail transit-dominated (such as subway connection core), road passenger transport-coupled (such as long-distance bus transfer area) and commercial service-attracted (such as catering and shopping area); non-Euclidean geometric topological mapping is used to construct an improved set of field equations, in which the electric field intensity component represents the passenger flow gravitational potential gradient in the waiting hall across transportation modes (such as the nonlinear variation of the attractiveness of the high-speed railway station hall to the commercial district with the intensity of promotional activities), and the magnetic flux density component describes the curl constraint of the passenger flow in the three-dimensional corridor (such as the spiral motion restriction caused by the height difference in the escalator area); the fluid field term introduces a variable viscosity coefficient model to define the viscous resistance coefficient tensor of the inter-layer channel. The component values ​​are dynamically correlated with the real-time operating speed, inclination angle and passenger flow density ratio of the escalator, accurately characterizing the traffic blockage effect of the same-layer channel under different operating conditions, providing a physical field quantitative basis for dynamic regulation.

[0046] A heterogeneous event-driven virtual hub environment is constructed. Rail transit arrival and departure events are simulated through a discrete event-driven engine to simulate sudden disturbances such as train delays and additional trains. The attractiveness of commercial services is modeled as a continuous spatiotemporal Poisson process to reflect the time-varying adsorption effect of promotional activities on passenger flow. Extreme scenarios generated by adversarial means include chain reactions of schedule conflicts (such as the cumulative impact of passenger flow caused by consecutive train delays) and oversaturated adsorption in commercial areas (such as supersaturated passenger flow caused by holiday promotions). Stochastic differential equations are introduced to describe phase transition behaviors such as the attenuation of passenger flow movement speed and sudden changes in path selection preferences under blizzard weather. Topologically persistent homology analysis is performed on the reinforcement learning policy library through a dynamic strategy evolution graph to identify the topological invariance characteristics of key decision paths such as hub evacuation path optimization and cross-level escalator group control. Based on the conditional generative adversarial network, a set of derived cross-transportation mode collaborative response strategies is synthesized (such as linking the subway to extend the operating hours to relieve stranded high-speed rail passengers), realizing two-way strategy iterative optimization between the digital twin environment and the physical system.

[0047] The comprehensive data integration module extends traditional passenger flow attraction potential analysis to a multidimensional phase space that includes schedule disturbances and weather influences. The Lyapunov stability criterion is used to determine the critical conditions for cross-floor passenger flow attraction (such as the gravitational threshold of the commercial area's maximum carrying capacity on subway connecting passenger flow). An adaptive curl analysis method is used at transfer nodes. Based on the Helmholtz decomposition principle, the vortex field of passenger flow is separated into a rotational component (reflecting the circulation intensity in the queuing area) and a diffusion component (representing the tendency of congestion to spread to surrounding areas), accurately locating high-risk nodes such as security checkpoints and ticket gates. A disturbance prediction model based on a hidden Markov chain is designed. By using the state transition probability matrix of transportation vehicles to events and commercial activities (such as a flight delay triggering a jump in the probability of passenger flow detention in the dining area), multi-scenario probability prediction results for external passenger flow input are generated to support the early deployment of a graded warning strategy. The above technical solution, through the deep collaboration of physical field quantification, digital twin deduction, and multi-source data fusion, has built a complete technology chain from three-dimensional perception, dynamic modeling, to intelligent control.

[0048] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent analysis and dynamic early warning of passenger flow in integrated transportation hubs. The method is specifically applicable to integrated transportation hub scenarios such as multi-transportation connection areas, three-dimensional channels, and high-density passenger flow nodes. It is characterized by: The method comprises the following steps: A multimodal sensing network is deployed within the three-dimensional space of integrated transportation hubs. This includes deploying three-dimensional laser scanning arrays at vertical transportation nodes to simultaneously capture the multi-layered movement of people on escalators and stairs; embedding a distributed inertial sensor matrix on the floor of transfer halls to monitor vibration waveforms caused by crowd movement trends; and deploying penetrating millimeter-wave radar clusters in security checkpoints to simultaneously detect the characteristics of personal belongings and human behavior patterns. A hub-specific spatiotemporal coupling field model is established, dividing the hub space into functionally related dynamic units. Each unit contains the coupling relationship between the rail transit connection area, the commercial service area, and the vertical commuter corridor. Based on the improved equation system, the passenger flow evolution equation under the coordinated action of multiple transportation modes is constructed, in which the electromagnetic field term represents the gravitational effect of cross-layer passenger flow, and the fluid field term describes the viscous resistance of the same-layer corridor. Implement dynamic risk field quantitative analysis to calculate the potential energy gradient between functional units and identify congestion risks caused by cross-regional passenger flow attraction; Detect abnormal passenger flow vortices at transfer nodes through vorticity tensor analysis; Fusion of multi-mode transport schedule data to predict the disturbance of external input flow to the internal potential energy field; When the potential energy of the vertical commuter channel exceeds the limit, a hierarchical coordinated response strategy is generated to dynamically adjust the escalator's direction and speed to balance the pressure difference between floors. When abnormal vortexes are detected in the transfer area, adjacent gate groups are linked to implement fuse-type passenger flow interception. Based on the multi-modal coordination agreement, schedule adjustment suggestions are sent to related operators to mitigate external passenger flow shocks; Through the digital twin system, the evolution of the entire hub situation is realized, and a virtual hub environment that includes rail transit arrival and departure and commercial service attractions is constructed; historical passenger flow data and extreme scenarios generated by adversarial means are injected to train the reinforcement learning strategy library to adapt to the complex working conditions of multiple transportation modes coupling; the mapping relationship between field equation parameters and response strategies is iteratively updated daily.

2. The method for intelligent analysis and dynamic early warning of passenger flow in a comprehensive transportation hub according to claim 1, characterized in that: The deployment of the multimodal perception network includes setting up a multispectral LiDAR array at each transition platform of the vertical transportation node, with its scanning plane dynamically orthogonal to the escalator's running trajectory, and synchronously acquiring the pedestrian's three-dimensional skeleton characteristics and thermal radiation distribution through wavelength division multiplexing technology; The distributed inertial sensor matrix is ​​embedded in the transfer hall ground substrate in a non-uniform grid format. Each node is equipped with a triaxial MEMS accelerometer and a piezoelectric vibration sensor. The time-frequency domain coherence analysis is used to decouple individual gait vibration from group movement trends. The millimeter-wave radar cluster uses a MIMO antenna architecture to form a synthetic aperture, performs multi-perspective penetrating scans of security inspection channels, and combines a polarization scattering feature classification algorithm to distinguish the geometric configuration and motion trajectory of metal objects, generating a threat assessment vector that integrates multiple physical quantities.

3. The method for intelligent analysis and dynamic early warning of passenger flow in a comprehensive transportation hub according to claim 1, characterized in that: The establishment of the space-time coupling field model includes dividing the dynamic units into three types of interactive modules: rail transit-dominated, road passenger transport-coupled, and commercial service-adsorbed based on the topological connection strength of multiple transportation modes in the hub; the improved set of equations introduces non-Euclidean topological mapping, in which the electric field intensity component represents the passenger flow gravitational potential gradient in the waiting hall across transportation modes, and the magnetic flux density component describes the passenger flow curl constraint in the three-dimensional corridor; the fluid field term adopts a variable viscosity coefficient model to define the viscous resistance coefficient tensor of the interlayer channel, and its component values ​​are dynamically correlated with the escalator operation status and the real-time passenger flow density ratio.

4. The method for intelligent passenger flow analysis and dynamic early warning in a comprehensive transportation hub according to claim 1, characterized in that: The digital twin system constructs a virtual hub environment driven by heterogeneous events, in which rail transit arrival and departure events are simulated using a discrete event-driven engine, and commercial service attraction is modeled as a continuous space-time Poisson process; the extreme scenarios generated by adversarial means include two core stress test cases: timetable conflict chain reaction and commercial area oversaturation adsorption, and introduces stochastic differential equations to describe the passenger flow phase change process under extreme weather conditions; by establishing a dynamic strategy evolution map, the key decision paths in the reinforcement learning strategy library are identified through topological persistent homology analysis, and a derived strategy set for cross-transportation mode collaborative response is synthesized based on a conditional generative adversarial network.

5. The method for intelligent analysis and dynamic early warning of passenger flow in a comprehensive transportation hub according to claim 1, characterized in that: The dynamic risk field quantitative analysis expands the analysis scope of passenger flow attraction potential to a multi-dimensional space that includes schedule changes by constructing a comprehensive data integration module, and determines the critical conditions for cross-floor passenger flow attraction through stability criteria. Adaptive curl analysis method is used at transfer nodes to decompose the rotational and diffusion components of the pedestrian vortex. A disturbance prediction model based on state transition is designed to generate multi-scenario probability prediction results of external passenger flow input by analyzing the correlation between transportation arrival and departure events and commercial activities.

Images