A method and system for predicting the performance of a high-altitude controller based on a smart bracelet

Abstract

This invention discloses a method and system for predicting the performance of air traffic controllers in high-altitude areas based on smart bracelets. It relates to the fields of air traffic management, high-altitude medicine, and wearable computing. The method includes collecting physiological data such as heart rate, blood oxygen, and blood pressure via a smart bracelet; simultaneously acquiring environmental data such as altitude and visibility; extracting reaction time and error rate from voice data; performing data cleaning and alignment; inputting multimodal data into a deep fusion neural network model to calculate comprehensive risk indicators; triggering alerts and visually outputting the status when the indicators exceed dynamic thresholds, while simultaneously generating automated intervention suggestions. This invention, specifically designed for the low-oxygen environment of high altitudes, achieves deep attentional fusion of physiological, environmental, and behavioral data, constructing a dedicated prediction model suitable for high-altitude airports and forming a complete closed loop of "perception-prediction-intervention," providing an intelligent and non-intrusive proactive management tool for air traffic control safety.

Description

Technical Field

[0001] This invention relates to the fields of air traffic management, high-altitude medicine, and wearable computing technology, and more specifically to a method and system for predicting the performance of high-altitude air traffic controllers based on a smart bracelet. Background Technology

[0002] Currently, high-altitude airports have thin air and low oxygen levels, and air traffic controllers are in a low-oxygen environment for a long time, which significantly affects their cognitive function, reaction ability and decision-making performance, making them a major potential source of risk to aviation safety.

[0003] Existing controller condition monitoring technologies are mostly designed for lowland environments, failing to consider the unique impact of high-altitude hypoxia on baseline and fluctuation patterns of physiological indicators. Some physiological monitoring requires wearing specialized medical equipment, interfering with control operations; while relying solely on post-event voice analysis cannot provide real-time early warning of coupled physiological and behavioral changes. High-altitude airports have complex and variable weather conditions (such as sudden changes in visibility), and existing technologies sever the dynamic connection between controllers' physiological state and the external operating environment, making it difficult to accurately predict performance inflection points in real-world work scenarios. Existing technologies largely remain at the "monitoring-alarm" level, failing to provide actionable, automated intervention suggestions that conform to control procedures, thus limiting their decision support value.

[0004] Therefore, how to achieve accurate perception, integrated prediction, and closed-loop intervention of the physiological, cognitive, and behavioral states of controllers in the special environment of plateaus is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed to provide a method and system for predicting the performance of high-altitude air traffic controllers based on a smart bracelet, which overcomes or at least partially solves the above problems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, embodiments of the present invention provide a method for predicting the performance of high-altitude air traffic controllers based on a smart bracelet, characterized by comprising: S1. Real-time collection of controllers' physiological data from previous periods through integrated smart wristbands, simultaneous acquisition of environmental data of plateau airports from airport meteorological systems, and extraction of behavioral performance data from control voice collection and analysis systems; S2. Preprocess and spatiotemporally align physiological, environmental, and performance data to obtain a standardized multimodal dataset; S3. Input the standardized multimodal dataset into the pre-built multimodal deep fusion neural network model to calculate the controller's real-time comprehensive risk index; S4. Set a dynamic risk threshold T. Compare the comprehensive risk index with the preset dynamic risk threshold. When the comprehensive risk index exceeds the preset dynamic risk threshold T, trigger a command reminder and output the controller's current status through the visual interface. At the same time, generate and push automated intervention suggestions.

[0008] Furthermore, the physiological data in S1 includes at least heart rate, blood oxygen saturation, and blood pressure data; Environmental data should include at least the altitude and visibility data of the plateau airport; Behavioral performance data should include at least the controller's verbal response time and instruction error rate.

[0009] Furthermore, the preprocessing and spatiotemporal alignment of physiological, environmental, and performance data in S2 specifically involves: Physiological data is filtered to remove noise and outliers, environmental and performance data are standardized with timestamps, all data are sliced ​​into sliding slices according to a preset time window to achieve spatiotemporal alignment of multi-source heterogeneous data, and missing data is interpolated or filled forward.

[0010] Furthermore, the multimodal deep fusion neural network model in S3 includes: Physiological Feature Encoding Subnetwork: Extracts deep temporal features from physiological data to obtain a deep physiological feature vector H. phy Deep temporal features include at least heart rate variability, blood oxygen fluctuation trends, and blood pressure change rate; Environmental feature embedding subnetwork: maps environmental data into an environmental risk modulation factor vector E env ; Behavioral performance feature encoding subnetwork: Extracting deep behavioral feature vectors B from behavioral performance data beh ; Attention Fusion Layer: Calculate the attention weight α of the environmental risk modulation factor on the deep physiological features, and perform weighted modulation on the deep physiological features according to the attention weight α to obtain the environmental weighted physiological features H'. phy ; Calculate the environmental weighted physiological characteristics H' phy The coupling attention weight β between the deep behavioral features and the deep behavioral features is used to weight and modulate the deep behavioral features to obtain the physiological-behavioral coupled weighted behavioral features B'. beh ; Environmental weighted physiological characteristics H' phy Environmental risk modulation factor vector E env Physiological-behavioral coupling weighted behavioral features B' beh Perform concatenation and linear transformation to generate a deep fusion vector F. fusion ; Risk mapping layer, used to fuse deep vector F fusionIt is mapped to a quantitative comprehensive risk indicator.

[0011] Furthermore, the attention fusion layer specifically includes: Environmental-physiological attention unit: Calculate the attention weight α: α=softmax(W env ·[E env ;H phy ]+b env ) Computational environment-weighted physiological characteristics H' phy : H' phy =α⊙H phy Physiological-behavioral coupled attention unit: Calculate the coupling attention weight β: β=softmax(W beh ·[H' phy ;B beh ]+b beh ) Calculate the physiological-behavioral coupled weighted behavioral feature B' beh : B' beh =β⊙B beh Feature splicing and linear transformation unit: Calculate the deep fusion vector F fusion : F fusion =W concat ·[H' phy ;E env ;B' beh ]+b concat Among them W env b env W beh b beh W conca b concat represents learnable network parameters; [;] represents the concatenation operation of two; [;;] represents the concatenation operation of three; softmax represents the normalized exponential function; ⊙ represents element-wise multiplication.

[0012] Furthermore, the preset dynamic risk threshold T in S4 is adaptively adjusted based on the controller's personal historical baseline and current plateau environmental parameters. The calculation formula is as follows: T = Individual historical baseline mean + σ Personal historical baseline standard deviation Where σ represents the environmental adaptation coefficient, which is dynamically calculated based on the current altitude and visibility.

[0013] Furthermore, the automated intervention suggestions in S4 include at least rest reminders, seat rotation suggestions, auxiliary monitoring reminders, and workload adjustment reminders.

[0014] Secondly, embodiments of the present invention provide a performance prediction system for high-altitude air traffic controllers based on a smart bracelet, characterized in that it includes: The smart wristband terminal, worn on the controller's wrist, integrates a heart rate sensor, a blood oxygen saturation sensor, and a blood pressure monitoring module for continuous collection of physiological data; The environmental data interface module is securely connected to the airport's automatic weather observation system to acquire real-time altitude and visibility data. The voice analysis and behavior extraction module is connected to the air traffic control ground and air communication recording system to extract reaction time and command error rate in real time. The edge computing server is deployed at the control site and has a built-in deep fusion neural network model, data preprocessing and alignment module, model inference module and risk calculation and threshold judgment module to perform data alignment, model inference and risk calculation. A visual early warning and intervention terminal is set up at the duty officer's seat to output controller status, risk level, and automated intervention suggestions.

[0015] Furthermore, the edge computing server also stores a database of controllers' personal historical baselines and dynamically updates the historical baseline data; the visual early warning and intervention terminal includes a real-time status visualization module, a multi-seat risk overview module, an early warning module, and an intervention suggestion generation and push module. The intervention suggestion generation and push module has a built-in rule engine that matches preset intervention rules according to risk level and dominant factors to generate personalized intervention suggestions.

[0016] Furthermore, it also includes a model management platform, which is deployed in the cloud or on a central server, for version management, online updates, performance monitoring, and user baseline management of deep fusion neural network models in edge computing servers.

[0017] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: This invention provides a method and system for predicting the performance of air traffic controllers in high-altitude areas based on smart bracelets. For the first time, a prediction model specifically for high-altitude airport controllers is constructed: for the low-oxygen environment of high altitude, blood oxygen saturation is innovatively introduced as a core physiological feature, and it is dynamically integrated with the environmental altitude and visibility through an attention mechanism, so that the model can adapt to the special characteristics of the high-altitude environment and greatly improve the prediction accuracy.

[0018] Achieving deep modeling of the "physiology-environment-behavior" triad: By designing a multi-level attention fusion network, not only is deep fusion of the three types of data achieved, but the dynamic coupling relationship between the three can also be identified, such as the synergistic effect of "blood oxygen decrease and reaction time prolongation in low visibility environment", thereby more accurately capturing performance inflection points.

[0019] Forming a complete technological closed loop of "perception-prediction-intervention": This invention not only realizes risk prediction and visualization, but also generates automated intervention suggestions that comply with regulatory operation procedures. It embeds the technology system into the actual management process, providing duty officers with direct and operable decision support, and realizing a leap from passive monitoring to proactive intelligent management.

[0020] Non-invasive and low-load engineering implementation: Using a smart wristband as the core data collection device, continuous and unobtrusive physiological data collection is achieved without increasing the operator's workload or interfering with existing workflows, making it highly valuable for engineering application and promotion. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 This is an overall flowchart of the solution provided in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the specific process from data collection to intervention suggestion generation provided in this embodiment of the invention. Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention discloses a method and system for predicting the performance of high-altitude air traffic controllers based on a smart bracelet, such as... Figure 1 As shown, it includes: S1. Real-time collection of controllers' physiological data from previous periods through integrated smart wristbands, simultaneous acquisition of environmental data of plateau airports from airport meteorological systems, and extraction of behavioral performance data from control voice collection and analysis systems; S2. Preprocess and spatiotemporally align physiological, environmental, and performance data to obtain a standardized multimodal dataset; S3. Input the standardized multimodal dataset into the pre-built multimodal deep fusion neural network model to calculate the controller's real-time comprehensive risk index; S4. Set a dynamic risk threshold T. Compare the comprehensive risk index with the preset dynamic risk threshold. When the comprehensive risk index exceeds the preset dynamic risk threshold T, trigger a command reminder and output the controller's current status through the visual interface. At the same time, generate and push automated intervention suggestions.

[0024] The specific implementation of this invention is as follows: The specific implementation of the method in Example 1 is as follows: Figure 2 As shown: Data Acquisition: Controller 1, on duty at a high-altitude airport, wears a smart bracelet that collects his heart rate, blood oxygen saturation (SpO2), and blood pressure (systolic / diastolic). Simultaneously, the system obtains local visibility data from the meteorological system via a secure interface. The voice system analyzes his air-to-ground communication in real time, calculating the reaction time (the time interval from receiving the pilot's reply to issuing the next instruction) and the instruction error rate (such as incorrect instructions, inconsistent repetition, etc.) for each instruction.

[0025] Data preprocessing and alignment: The raw data is processed using a 1-minute sliding window. For example, median filtering is applied to heart rate data to remove abnormal spikes caused by loose wristbands. All data is aligned to a uniform 1-minute time slice according to timestamps. If a certain type of data is missing within a window, forward padding or interpolation is used to complete it.

[0026] Model inference and risk calculation: The aligned 1-minute data feature vector is input into a deep neural network model deployed on a field edge server.

[0027] The physiological encoder extracts heart rate variability (LF / HF ratio), rate of decrease in blood oxygen saturation (dSpO2 / dt), and blood pressure variability index.

[0028] The environment embedding layer encodes the current altitude and visibility into modulation vectors.

[0029] The behavior encoder extracts the mean and standard deviation of reaction time over the most recent 5 minutes, as well as the sliding cumulative value of the error rate.

[0030] The attention fusion layer calculates the weights of the modulation vector to physiological features. For example, when visibility is below 1000 meters, the attention mechanism automatically enhances the contribution of the rate of decline in blood oxygen saturation to the final risk.

[0031] The risk mapping layer outputs a comprehensive risk index R ranging from 0 to 100.

[0032] Threshold determination: The system stores the controller's personal historical baseline for the same time period over the past 7 days. The dynamic threshold T = baseline mean + 0.5 × baseline standard deviation. If R > T, the risk is considered excessive.

[0033] Visualization and intervention suggestion generation: When R>T and lasts for more than 3 minutes, the system will indicate the increased risk with an orange curve on the duty manager's terminal and mark the main contributing factors: "Main influencing factors: decreased blood oxygen, low visibility".

[0034] At the same time, an intervention suggestion was pushed to the duty manager's terminal: "The overall risk of controller 1 has increased. It is recommended to pay attention to its status and consider temporarily reducing the number of flights transferred in its sector. Review in 10 minutes."

[0035] Specific composition of the system in Example 2: Smart bracelet: integrates heart rate / blood oxygen monitoring module and blood pressure monitoring module with AFE chip, and communicates with edge server via Bluetooth BLE.

[0036] Data access layer: including meteorological data interface and voice analysis server interface.

[0037] Edge computing server: Embedded deep learning model, data alignment module and threshold management module.

[0038] Visual terminal: including the chief screen.

[0039] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0040] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the performance of high-altitude air traffic controllers based on smart wristbands, characterized in that, include: S1. Real-time collection of controllers' physiological data from previous periods through integrated smart wristbands, simultaneous acquisition of environmental data of plateau airports from airport meteorological systems, and extraction of behavioral performance data from control voice collection and analysis systems; S2. Preprocess and spatiotemporally align the physiological data, environmental data, and performance data to obtain a standardized multimodal dataset; S3. Input the standardized multimodal dataset into the pre-built multimodal deep fusion neural network model to calculate the controller's real-time comprehensive risk index; S4. Set a preset dynamic risk threshold T. Compare the comprehensive risk index with the preset dynamic risk threshold. When the comprehensive risk index exceeds the preset dynamic risk threshold T, trigger a command reminder and output the controller's current status through a visual interface. At the same time, generate and push automated intervention suggestions.

2. The method as described in claim 1, characterized in that, include: The physiological data described in S1 include at least heart rate, blood oxygen saturation, and blood pressure. The environmental data includes at least the altitude and visibility data of the plateau airport. The behavioral performance data includes at least the controller's voice response time and instruction error rate.

3. The method as described in claim 1, characterized in that, The preprocessing and spatiotemporal alignment of the physiological data, environmental data, and performance data in S2 specifically involves: The physiological data is filtered to remove noise and outliers, the environmental data and performance data are timestamped and standardized, all data are sliced ​​into sliding slices according to a preset time window to achieve spatiotemporal alignment of multi-source heterogeneous data, and missing data is interpolated or filled forward.

4. The method as described in claim 1, characterized in that, include: The multimodal deep fusion neural network model described in S3 include: Physiological Feature Encoding Subnetwork: Extracts deep temporal features from the physiological data to obtain a deep physiological feature vector H. phy The deep temporal features include at least heart rate variability, blood oxygen fluctuation trend, and blood pressure change rate; Environmental feature embedding sub-network: maps the environmental data into an environmental risk modulation factor vector E. env ; Behavioral performance feature encoding sub-network: Extracts deep behavioral feature vector B from the behavioral performance data. beh ; Attention Fusion Layer: Calculates the attention weight α of the environmental risk modulation factor on the deep physiological features, and performs weighted modulation on the deep physiological features according to the attention weight α to obtain the environmental weighted physiological features H'. phy ; Calculate the environmental weighted physiological characteristic H' phy The deep behavioral features are coupled with attention weight β, and the deep behavioral features are weighted and modulated according to the coupled attention weight β to obtain physiological-behavioral coupled weighted behavioral features B'. beh The environmental weighted physiological characteristic H' phy Environmental risk modulation factor vector E env Physiological-behavioral coupling weighted behavioral features B' beh Perform concatenation and linear transformation to generate a deep fusion vector F. fusion ; Risk mapping layer, used to map the deep fusion vector F fusion It is mapped to a quantitative comprehensive risk indicator.

5. The method as described in claim 4, characterized in that, The attention fusion layer specifically includes: Environmental-physiological attention unit: Calculate the attention weight α: α=softmax(W env ·[E env ;H phy ]+b env ) Calculate the environmental weighted physiological characteristic H' phy : H' phy =α⊙H phy Physiological-behavioral coupled attention unit: Calculate the coupling attention weight β: β=softmax(W beh ·[H' phy ;B beh ]+b beh ) Calculate the physiological-behavioral coupled weighted behavioral feature B' beh : B' beh =β⊙B beh Feature splicing and linear transformation unit: Calculate the deep fusion vector F fusion : F fusion =W concat ·[H' phy ;E env ;B' beh ]+b concat Among them W env b env W beh b beh W conca b concat represents learnable network parameters; [;] represents the concatenation operation of two; [;;] represents the concatenation operation of three; softmax represents the normalized exponential function; ⊙ represents element-wise multiplication.

6. The method as described in claim 1, characterized in that, This includes: the preset dynamic risk threshold T mentioned in S4 is adaptively adjusted based on the controller's personal historical baseline and current plateau environmental parameters, and the calculation formula is: T = Individual historical baseline mean + σ Personal historical baseline standard deviation Where σ represents the environmental adaptation coefficient, which is dynamically calculated based on the current altitude and visibility.

7. The method as described in claim 1, characterized in that, Including; the automated intervention suggestions described in S4 include at least rest reminders, seat rotation suggestions, auxiliary monitoring reminders, and workload adjustment reminders.

8. A performance prediction system for high-altitude air traffic controllers based on smart wristbands, characterized in that, A method for implementing any one of claims 1-7, comprising: The smart wristband terminal, worn on the controller's wrist, integrates a heart rate sensor, a blood oxygen saturation sensor, and a blood pressure monitoring module for continuous collection of physiological data; The environmental data interface module is securely connected to the airport's automatic weather observation system to acquire real-time altitude and visibility data. The voice analysis and behavior extraction module is connected to the air traffic control ground and air communication recording system to extract reaction time and command error rate in real time. An edge computing server, deployed at the control site, is equipped with the deep fusion neural network model, data preprocessing and alignment module, model inference module, and risk calculation and threshold judgment module to perform data alignment, model inference, and risk calculation. A visual early warning and intervention terminal is set up at the duty officer's seat to output controller status, risk level, and automated intervention suggestions.

9. The system as described in claim 8, characterized in that, The edge computing server also stores a database of controllers' personal historical baselines and dynamically updates the historical baseline data. The visual early warning and intervention terminal includes a real-time status visualization module, a multi-seat risk overview module, an early warning module, and an intervention suggestion generation and push module. The intervention suggestion generation and push module has a built-in rule engine that matches preset intervention rules based on risk level and dominant factors to generate personalized intervention suggestions.

10. The system as described in claim 8, characterized in that, It also includes a model management platform, which is deployed in the cloud or a central server and is used to manage the deep fusion neural network models in the edge computing server, perform online updates, monitor performance, and manage user baselines.

Images