Airport service process optimization method and system based on artificial intelligence

Abstract

The invention discloses an airport service process optimization method based on artificial intelligence, which comprises the following steps: firstly, fusing multi-modal data such as positioning, vision and business systems to construct a global coverage behavior feature library, and breaking an information barrier; key behavior features are extracted through space-time calibration and dynamic feature screening, a feature library is continuously updated, and the behavior characterization accuracy is improved; based on a feature library, in combination with network modeling, time sequence prediction and a graph neural network, passenger flow fluctuation is pre-judged, a node dependency relationship is analyzed, early recognition of queuing abnormity and equipment faults is realized, and service logic is dynamically adjusted by a rule engine to enhance fault tolerance; and finally, pre-verifying strategy feasibility through a digital twin framework, synchronously generating resource scheduling, moving line optimization and information guide instructions by utilizing reinforcement learning, deploying edge calculation to guarantee real-time processing, and forming a continuous optimization cycle from perception to decision by combining three-dimensional thermodynamic diagram visualization and multi-modal instruction closed-loop feedback. And the airport service efficiency and the passenger experience are comprehensively improved.

Description

Technical Field

[0001] This invention relates to the field of aviation services, and more particularly to an artificial intelligence-based method and system for optimizing airport service processes. Background Technology

[0002] Traditional airport service process management relies heavily on manual experience and static systems, making it difficult to cope with dynamic passenger flow changes and sudden anomalies. Existing technologies often use independent subsystems to process location, monitoring, and business data, resulting in significant information silos and hindering comprehensive collaborative decision-making. Passenger behavior analysis is limited to single data sources and lacks multimodal fusion mechanisms, leading to one-sided behavioral feature extraction. Service node resource allocation depends on historical average predictions, failing to respond in real-time to peak passenger flow or equipment failures, often resulting in both queuing congestion and resource idleness. Anomaly detection is mostly based on threshold alarms, lacking process-level correlation analysis models, resulting in significant warning lag. Optimization strategies are often generated through offline simulations, creating a temporal and spatial gap with real-world scenarios and posing high execution risks. The lack of a data closed-loop system makes it difficult to feedback strategy effects to the decision-making end, restricting the system's ability to continuously evolve. Summary of the Invention

[0003] This invention proposes an artificial intelligence-based method for optimizing airport service processes, including: S1. Collect multimodal data, including positioning data, visual data, and business system data, and construct a behavioral feature library through fusion processing; S2. Perform spatiotemporal calibration and feature filtering on the multimodal data, extract key behavioral features, and update the behavioral feature library; S3. Based on the key behavioral characteristics, identify process anomalies in the service process; S4. Generate dynamic optimization strategies to optimize service processes.

[0004] The location data includes data obtained through Bluetooth beacons, mobile application indoor positioning, and baggage RFID tags; The visual data includes camera monitoring data based on target detection algorithms; The business system data includes departure system data and equipment management system data that are connected through application programming interfaces.

[0005] Step S2 includes: Spatiotemporal calibration of multimodal data is performed using filtering algorithms; Key behavioral features are extracted using feature filtering algorithms; The feature selection algorithm integrates an automatic feature extraction network to generate and update the behavioral feature library.

[0006] Step S3 includes: Based on the network model, a service process model is constructed, defining service nodes and transfer rules; Predict passenger flow fluctuations using time series forecasting models; The system combines graph neural network models to analyze service node dependencies and detects queuing time anomalies and equipment failures in real time.

[0007] The service process model integrates a rule engine, which dynamically updates the service node transfer rules based on real-time data.

[0008] Step S4 includes: Digital twin simulation based on a simulation model framework; A policy is generated using a reinforcement learning algorithm, which simultaneously performs the following operations: Dynamically adjust the number of security checkpoints and check-in counters; optimize passenger flow planning; and provide information guidance through mobile devices by pushing route planning information.

[0009] The simulation model framework deploys edge computing nodes, which are used to perform high-concurrency data processing in steps S2 and S3. The dynamic adjustment of resource quantities specifically includes: adjusting security checkpoints and check-in counters; and optimizing passenger flow planning specifically includes reducing detour distances.

[0010] The visualization and interactive operation performed after step S4 includes: Heatmaps are generated using a 3D visualization engine. Output multimodal instructions, including application programming interface control instructions, mobile optimization suggestions, and multilingual reports; This forms a closed-loop workflow from data collection to strategy execution.

[0011] The present invention also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned artificial intelligence-based airport service process optimization method.

[0012] This invention also proposes an airport service process optimization system based on artificial intelligence, comprising: The multimodal data acquisition module collects positioning data, visual data, and business system data. The intelligent data processing module uses a filtering algorithm for spatiotemporal calibration and generates a behavioral feature library by integrating an automatic feature extraction network and a feature selection algorithm. The behavior analysis module constructs a service process model containing service nodes and transfer rules based on a network model, and combines a time series prediction model and a graph neural network model to detect process anomalies. The dynamic decision-making module uses a simulation model framework to perform digital twin simulations and employs reinforcement learning algorithms to generate resource adjustment, traffic flow optimization, and mobile terminal information guidance strategies. The visualization module uses a 3D visualization engine to display heat maps, output multimodal commands, and form a closed-loop process.

[0013] This invention constructs a comprehensive behavioral feature library through multimodal data fusion, eliminating information barriers and providing a highly complete data foundation for process optimization. A spatiotemporal calibration mechanism addresses the asynchronous nature of multi-source sensing, and the dynamic feature library continuously adapts to scenario evolution, improving the accuracy of behavioral representation. Service process modeling based on a network model, combined with time-series prediction and graph neural networks, enables prediction of passenger flow fluctuations and analysis of node dependencies, allowing for early identification of queuing anomalies and equipment failures. A rule engine dynamically adjusts transfer logic, enhancing system fault tolerance. A digital twin framework pre-verifies the feasibility of optimization strategies, and reinforcement learning algorithms synchronously generate resource scheduling, traffic flow planning, and information guidance instructions, reducing implementation risks. Edge computing nodes ensure high-concurrency real-time processing, a 3D visualization engine intuitively presents heatmaps, and multimodal instructions are fed back to the execution end in a closed loop, forming a continuous optimization cycle from perception to decision-making, comprehensively improving airport service efficiency and passenger experience. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an artificial intelligence-based airport service process optimization method proposed in this invention. Detailed Implementation

[0015] refer to Figure 1 This invention proposes an artificial intelligence-based method for optimizing airport service processes, including: S1. Collect multimodal data, including location data, visual data, and business system data, and construct a behavioral feature library through fusion processing.

[0016] The location data includes data acquired via Bluetooth beacons, indoor positioning via mobile applications, and baggage RFID tags; the visual data includes camera monitoring data based on target detection algorithms; and the business system data includes data from the departure system and equipment management system connected via application programming interfaces (APIs). The Bluetooth beacon specifically refers to a low-power Bluetooth broadcast device, which calculates distance using a signal strength attenuation model; the target detection algorithm specifically includes the YOLO series neural network architecture, capable of simultaneously outputting human body location boxes and action classification labels.

[0017] Specifically, location data is obtained by capturing the signal strength of Bluetooth beacons deployed in the airport area and combining it with the indoor positioning function built into the mobile application to calculate the real-time location of passengers. Baggage RFID tags record the movement trajectory of baggage through RFID readers. Visual data is collected by a high-definition camera array and identifies passenger posture, movement direction and interaction behavior based on target detection algorithms. Business system data is exchanged with the departure system in real time with flight check-in status through application programming interfaces. At the same time, it connects with the equipment management system to obtain the operating parameters of equipment such as security screening machines and automatic doors. The three types of data are timestamped and then input into the fusion processor to generate spatiotemporally correlated original behavioral features.

[0018] Specific implementation examples are as follows: In the check-in hall scenario, Bluetooth beacons are deployed on the ceiling at 10-meter intervals. After receiving the beacon signal, the passenger's mobile application calculates its own coordinates; the baggage handling area reader scans the RFID tag to record the baggage sorting path; the departure system interface synchronizes flight delay data every five minutes; the camera group covers the queuing area, and the target detection algorithm identifies whether the passenger is stationary or moving.

[0019] In summary, multi-source heterogeneous data achieves full-domain coverage, and fusion processing eliminates data silos, providing highly complete input for behavioral modeling.

[0020] S2. Perform spatiotemporal calibration and feature filtering on the multimodal data, extract key behavioral features, and update the behavioral feature library.

[0021] Specifically, this includes: using filtering algorithms to perform spatiotemporal calibration on multimodal data; extracting key behavioral features through feature selection algorithms; wherein the feature selection algorithm integrates an automatic feature extraction network to generate and update the behavioral feature library.

[0022] Specifically, the Kalman filter algorithm is used to correct the time deviation between the positioning data and the visual data, that is, to align the camera frame timestamp with the Bluetooth beacon reporting time to compensate for network transmission delay. The fused data is processed by a feature filtering algorithm composed of a convolutional autoencoder. The encoder compresses the data dimension to extract the spatial distribution pattern, the decoder reconstructs the key feature vector, and the automatic feature extraction network learns the behavior sequence pattern through a gated recurrent unit. The output features are incrementally updated after being compared with the historical behavior feature library.

[0023] Specifically, the gate control loop unit is a variant of a recurrent neural network, which controls the transmission of historical information by updating and resetting the gate; the behavior feature database is a distributed graph database, which connects behavior feature nodes with passenger movement lines as edges.

[0024] The specific implementation example is as follows: When a passenger is captured by both a camera and a Bluetooth beacon, the filtering algorithm aligns the two data sources with a window of XX milliseconds; the feature filtering algorithm extracts twenty-dimensional features such as "wandering index" and "acceleration frequency" from the passenger's movement trajectory; when the automatic feature extraction network detects a new baggage delay pattern, it creates a corresponding feature classification in the behavior feature library.

[0025] In summary, spatiotemporal calibration eliminates positioning drift caused by multi-sensor time differences, and the dynamic feature library enables behavioral representations to continuously adapt to changes in the airport scene.

[0026] S3. Based on the key behavioral characteristics, identify process anomalies in the service process.

[0027] Step S3 specifically includes: constructing a service flow model based on a network model, defining service nodes and transfer rules; predicting passenger flow fluctuations using a time series prediction model; analyzing service node dependencies using a graph neural network model; detecting queuing time anomalies and equipment failures in real time; furthermore, the service flow model integrates a rule engine, which dynamically updates the service node transfer rules based on real-time data.

[0028] Specifically, a service process model is constructed based on complex network theory, defining security checkpoints and check-in counters as service nodes, passenger movement paths as transfer rules, and a time series prediction model using a seasonal autoregressive algorithm to analyze historical passenger flow data and predict peak passenger flow in the next fifteen minutes. A graph neural network model aggregates the states of adjacent nodes, triggering an abnormal alarm when the queue length of a node exceeds a threshold. The rule engine receives prediction results and equipment status in real time, and dynamically redirects passengers to backup nodes if a check-in counter malfunctions.

[0029] Specifically, the rule engine is a decision-making system based on the Drools framework. It performs logical inference through a condition-action rule chain, and the service node dependency relationship specifically refers to the process constraint that passengers must transfer to the security checkpoint after completing check-in.

[0030] Specific implementation examples are as follows: During the morning peak hours, the prediction model outputs that the check-in area will experience passenger overload. The graph neural network detects that the adjacent security checkpoint nodes have sufficient capacity. The rule engine immediately generates an "open fast track" instruction. When the temperature sensor of the baggage sorting equipment alarms, the service node is automatically marked as unavailable.

[0031] In summary, multi-dimensional models work together to achieve early warning of anomalies, and dynamic rule mechanisms improve the system's fault tolerance.

[0032] S4. Generate dynamic optimization strategies to optimize service processes.

[0033] Step S4 includes: performing digital twin simulation based on the simulation model framework; generating a strategy through a reinforcement learning algorithm, which simultaneously performs the following operations: The number of security checkpoints and check-in counters is dynamically adjusted; passenger flow planning is optimized; and information guidance is achieved through push route planning via mobile devices. Specifically, the simulation model framework deploys edge computing nodes, which are used to perform high-concurrency data processing in steps S2 and S3. Dynamically adjusting the number of resources specifically involves adjusting the number of security checkpoints and check-in counters; optimizing passenger flow planning specifically involves reducing detour distances.

[0034] After step S4, visualization and interactive operations are performed, including: generating a heat map based on a 3D visualization engine; outputting multimodal instructions, including application programming interface control instructions, mobile optimization suggestions, and multilingual reports; forming a closed-loop workflow from data collection to strategy execution.

[0035] Specifically, a 3D simulation environment for the airport is constructed based on a digital twin framework. A real-time service process model is loaded for stress testing. A reinforcement learning algorithm adopts a near-end policy optimization framework. In the simulation environment, the agent is trained to perform three-order actions: adjusting the number of open security checkpoints to balance queue length, planning the shortest detour path to reduce passenger walking distance, and pushing personalized navigation routes through passenger mobile applications. Edge computing nodes are deployed in various areas of the terminal to locally execute data filtering and anomaly detection algorithms to reduce cloud load. The 3D engine renders a heat map of the entire area, highlighting congested nodes in red. The application programming interface sends flight change instructions to the boarding gate display screen. The mobile terminal generates multilingual optimization suggestions, and the strategy execution results are fed back to the data collection terminal to form a closed loop.

[0036] Digital twin simulation specifically refers to establishing a virtual mapping model of the physical airport and driving simulation operation through real-time data. Multimodal commands specifically include three interactive forms: equipment control protocols, text messages, voice messages, and broadcasts.

[0037] A specific implementation example is as follows: When congestion occurs in the international departure hall, the reinforcement learning agent simulates the effect of adding three security checkpoints in the digital twin. After confirming that it is feasible, it sends an activation command to the on-site equipment. The edge nodes filter the original video streams from the cameras and only upload fragments of abnormal behavior to the central server.

[0038] In summary: digital twin implementation strategy pre-verification reduces implementation risks, while edge computing ensures the real-time performance of high-concurrency data processing.

[0039] This invention also proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the artificial intelligence-based airport service process optimization method as described in claim 1. When the program instructions stored in the storage medium are executed by the processor, the entire process of generating the aforementioned multimodal data acquisition behavior feature library construction process anomaly identification dynamic optimization strategy is fully implemented.

[0040] The present invention also proposes an airport service process optimization system based on artificial intelligence, including: a multimodal data acquisition module, an intelligent data processing module, a behavior analysis module, a dynamic decision-making module, and a visualization module.

[0041] The multimodal data acquisition module collects positioning data, visual data, and business system data. The multimodal data acquisition module forms a positioning data acquisition network through a physical layer Bluetooth beacon array mobile application data interface RFID reader / writer. The visual acquisition unit uses a combination of wide-angle and telephoto cameras to cover monitoring needs at different distances. The business system interface is compatible with the International Air Transport Association standard data protocol.

[0042] The intelligent data processing module uses a filtering algorithm for spatiotemporal calibration and generates a behavioral feature library by using a feature selection algorithm that integrates an automatic feature extraction network. The spatiotemporal calibration unit uses an extended Kalman filter to fuse multi-source clock signals, and the feature selection unit controls the feature dimension compression rate through a convolutional neural network kernel. The behavioral feature library is divided into data partitions according to passenger nationality and flight type.

[0043] The behavior analysis module constructs a service process model containing service nodes and transfer rules based on a network model. It combines a time series prediction model and a graph neural network model to detect process anomalies. The service process modeling unit defines four levels of nodes: check-in, security check, border inspection, and boarding. The time series prediction unit uses a long short-term memory network to process non-stationary passenger flow data. The graph neural network unit quantifies the influence weight of nodes through a graph attention mechanism.

[0044] The dynamic decision-making module performs digital twin simulation through a simulation model framework, and uses reinforcement learning algorithms to generate resource adjustment, traffic flow optimization, and mobile terminal information guidance strategies. The digital twin engine loads the building information model to generate simulation scenarios, and the reinforcement learning agent uses a deep deterministic strategy gradient algorithm to output resource scheduling instructions. Edge computing nodes are equipped with field-programmable gate arrays to accelerate matrix operations.

[0045] The visualization module displays heatmaps based on a 3D visualization engine, outputs multimodal commands, and forms a closed-loop process. The 3D visualization engine is preferably developed based on the Unity framework. The heatmap congestion index is updated, and the multimodal command generator is compatible with multiple languages.

[0046] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based airport service process optimization method, characterized by, The method comprises the following steps: S1, collecting multi-modal data including positioning data, visual data and business system data, and constructing a behavior feature library through fusion processing; S2, performing time-space calibration and feature screening on the multi-modal data, extracting key behavior features and updating the behavior feature library; S3, identifying process abnormalities in the service process based on the key behavior features; S4, generating a dynamic optimization strategy to optimize the service process. 2.The AI-based airport service flow optimization method of claim 1, wherein, The positioning data includes data obtained through Bluetooth beacons, mobile application indoor positioning and luggage radio frequency identification tags; The visual data includes camera monitoring data based on a target detection algorithm; The business system data includes departure system data and equipment management system data connected through an application programming interface. 3.The AI-based airport service flow optimization method of claim 1, wherein, The step S2 comprises: Performing time-space calibration on the multi-modal data using a filtering algorithm; Extracting key behavior features through a feature screening algorithm; The feature screening algorithm integrates an automatic feature extraction network to generate and update the behavior feature library. 4.The AI-based airport service flow optimization method of claim 1, wherein The step S3 comprises: Building a service process model based on a network model, defining service nodes and transfer rules; Combining a time series prediction model to predict passenger flow fluctuations; Combining a graph neural network model to analyze service node dependency relationships; and detecting queuing time abnormalities and equipment failures in real time. 5.The AI-based airport service flow optimization method of claim 4, wherein The service process model integrates a rule engine, which dynamically updates the service node transfer rules based on real-time data. 6.The AI-based airport service flow optimization method of claim 1, wherein The step S4 comprises: Performing digital twin simulation based on a simulation model framework; Generating a strategy through a reinforcement learning algorithm, which synchronously performs the following operations: Dynamically adjusting the number of security check channels and check-in counters; optimizing passenger flow planning; and achieving information guidance through mobile terminal path planning. 7.The AI-based airport service flow optimization method of claim 6, wherein, The simulation model framework deploys edge computing nodes for high-concurrency data processing in steps S2 and S3; Dynamically adjusting the number of resources specifically refers to adjusting the number of security check channels and check-in counters; and optimizing passenger flow planning specifically refers to reducing the detour distance. 8.The AI-based airport service flow optimization method of claim 1, wherein, After step S4, visualizing and interacting, comprising: Generating a heat map based on a three-dimensional visualization engine; Outputting multi-modal instructions, including application programming interface control instructions, mobile terminal optimization suggestions and multi-language reports; Forming a closed-loop workflow from data collection to strategy execution.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, the method for optimizing an airport service process based on artificial intelligence is realized.

10. An artificial intelligence based airport service process optimization system, characterized in that, The method comprises the following steps: A multi-modal data collection module collects positioning data, visual data and business system data; An intelligent data processing module performs time-space calibration using a filtering algorithm and generates a behavior feature library through a feature screening algorithm that integrates an automatic feature extraction network; A behavior analysis module builds a service process model containing service nodes and transfer rules based on a network model, and detects process abnormalities in combination with a time series prediction model and a graph neural network model; A dynamic decision-making module performs digital twin simulation through a simulation model framework, and generates resource adjustment, flow optimization and mobile terminal information guidance strategies using a reinforcement learning algorithm; A visualization module displays a heat map based on a three-dimensional visualization engine, outputs multi-modal instructions and forms a closed-loop process.

Images