Method for identifying situational awareness of controller during fusion operation of unmanned aerial vehicles in controlled airspace

Abstract

The invention discloses a method for identifying situational awareness of a controller during fusion operation of unmanned aerial vehicles in a controlled airspace. The method is suitable for high-cognitive-load task scenes such as unmanned aerial vehicle operation and air traffic control, eye movement data and electroencephalogram signals of a controller are synchronously collected through eye movement tracking equipment and electroencephalogram equipment, and precise assessment of situational awareness is achieved in combination with multi-modal physiological features. Firstly, an eye movement tracking instrument is used for collecting eye movement characteristics, and electroencephalogram equipment is used for collecting frequency domain characteristics; secondly, cleaning, band-pass filtering and independent component analysis processing are carried out on the collected data, and relevant physiological features are extracted; and finally, carrying out real-time classification identification on the situational awareness state based on the hybrid classification model. The method can realize objective and real-time monitoring of situational awareness, is suitable for task environments with high pressure and high cognitive load, and has wide application value.

Description

Technical Field

[0001] The present invention relates to the field of situational awareness assessment, and in particular to a method for identifying situational awareness of a controller during fusion operation of unmanned aerial vehicles in a controlled airspace. Background Art

[0002] With the rapid development of the drone industry, the complexity of air traffic controllers' work continues to increase. As a key indicator to ensure safe flight, situational awareness has attracted more and more attention. Existing situational awareness assessment methods are mostly subjective scale assessments, which are easy to operate but lack real-time and objectivity.

[0003] In recent years, eye movement and EEG signals, as physiological characteristics of situational awareness, have gradually been used to monitor the cognitive state of controllers, with the advantages of real-time and objectiveness. However, how to integrate these two physiological signals to build an efficient and accurate situational awareness recognition method is still a problem that needs to be solved urgently in current technology. Summary of the invention

[0004] Purpose of the invention: In order to solve the problem of lack of objective and real-time measurement means in the prior art in the process of evaluating the situational awareness of controllers, the present invention discloses a method for identifying the situational awareness of controllers during the fusion operation of UAVs in controlled airspace.

[0005] Technical solution: A method for identifying controller situational awareness during the fusion operation of unmanned aerial vehicles in controlled airspace, including:

[0006] Build an experimental environment and use computers to simulate air traffic control tasks involving the integration of manned and unmanned aircraft;

[0007] Install the eye tracker and EEG equipment, and calibrate the eye tracker according to each subject's situation;

[0008] The subjects began to perform the computer-simulated air traffic control task, and the eye tracker and EEG equipment were started to collect the subjects' eye movement data and EEG signals in real time during the task. At the same time, the computer was used to record the subjects' performance during the experiment.

[0009] After the simulated air traffic control mission is completed, the eye tracker and EEG equipment are stopped from collecting data, and the eye movement data and EEG signals collected by the eye tracker and EEG equipment are preliminarily processed to extract their key features;

[0010] Based on the extracted key features of eye movement data and EEG signals, in-depth analysis is performed through machine learning models to identify the subject's situational awareness state.

[0011] Furthermore, the preliminary processing of the eye movement data specifically includes the following steps:

[0012] S1.1 performs interpolation processing on the collected eye movement data, and fills in the eye movement data with missing time less than the preset time through linear interpolation. First, the scale factor S of data interpolation is determined according to the timestamps of valid data samples before and after the missing time. factor , the formula is:

[0013]

[0014] Among them, t timestamp of sample to be replaced Indicates the timestamp of the sample that needs to be interpolated or replaced, t timestamp of first sample after gap Refers to the timestamp of the first valid sample after the data missing interval, t timestamp of last sample after gap Refers to the timestamp of the last valid sample after the data missing interval;

[0015] The time interpolation of the missing samples and the position information of the valid samples are calculated to fill the missing sample points. The formula is:

[0016] x m =x1+S factor ×(x2-x1)

[0017] y m =y1+S factor ×(y2-y1)

[0018] Among them, (x1, y1) and (x2, y2) are the screen coordinates of the valid sample points. m ,y m ) is the interpolated position of the missing sample point;

[0019] S1.2 performs noise reduction on the eye movement data after interpolation processing in S1.1, using a non-weighted moving average filter, assuming the filter sample window size is 2N, and the formula is as follows:

[0020]

[0021] Where N represents the number of sample points taken before and after the current sample point, n represents the index position of the current sample point, k represents the interval from the current sample point forward, x represents the sample value in the current time series, and y(n) represents the result of weighted averaging of the 2N+1 sample values ​​before and after the current sample point.

[0022] S1.3 calculates the eye movement velocity based on the eye movement data after noise reduction, sets the eye movement velocity threshold and classifies the corresponding sample points according to the eye movement velocity. The velocity calculation formula is:

[0023]

[0024] Wherein, (x1, y1), (x2, y2) are the screen coordinates of two consecutive sample points, t1 is the timestamp of the sample point with screen coordinates (x1, y1), and t2 is the timestamp of the sample point with screen coordinates (x2, y2); if the speed of the sample point is lower than the threshold, the sample point is classified as a fixation point, and if the speed of the sample point is equal to or higher than the threshold, it is classified as a scanning point.

[0025] Furthermore, the key features extracted from the preliminarily processed eye movement data include: the number of fixation points, fixation time, scanning speed, scanning angle speed, pupil diameter, fixation time ratio and gaze entropy.

[0026] Furthermore, the preliminary processing of the EEG signal comprises the following specific steps:

[0027] S2.1 Re-reference the collected EEG signals to the average value of the bilateral mastoid electrodes using the binaural reference method;

[0028] S2.2 performs bandpass filtering on the EEG signal to extract a specific frequency range. The formula is:

[0029]

[0030] Among them, X(f) is the frequency domain representation of the EEG signal, f represents the frequency, which is the independent variable in the integration, t represents the time variable, which represents the time point of the signal in the time domain, and F1 and F2 are the lower and upper limit frequencies of the filter respectively;

[0031] S2.3 uses independent component analysis to further remove artifacts from the signal. The independent component analysis formula is:

[0032] X=AS

[0033] Where X is the observed signal matrix, A is the mixing matrix, and S is the independent source signal matrix;

[0034] S2.4 performs fast Fourier transform on the EEG signal processed by S2.3 to extract the characteristic power of the four main frequency bands of delta wave, theta wave, alpha wave and beta wave in the EEG signal.

[0035] Furthermore, the key features are extracted based on the initially processed EEG signal, and the key features of the EEG signal include the θ / β ratio, the (θ+α) / β ratio, the α / β ratio and the (θ+α) / (α+β) ratio.

[0036] Furthermore, the in-depth analysis by the machine learning model includes the following specific steps:

[0037] S3.1 Perform sensitivity analysis on eye movement data and EEG signals to extract sensitive features that are significantly related to situational awareness;

[0038] S3.2 uses principal component analysis to reduce the dimension of these sensitive features, remove redundant and noisy features, and distinguish linear features from nonlinear features;

[0039] S3.3 inputs the sensitive feature data after dimensionality reduction into the machine learning model for classification and identification, and uses a hybrid model combining support vector machine and random forest model for analysis, where the support vector machine uses radial basis function kernel function to process the nonlinear part in the data.

[0040] Furthermore, for the sensitivity analysis described in S3.1, for the collected eye movement data, the Levene test was first performed to evaluate the homogeneity of variance. If the homogeneity was met, the t test was used to analyze the mean difference. If it was not met, the Mann-Whitney U test was used. For the EEG signals, repeated measures analysis of variance was used.

[0041] Furthermore, the specific process of calibrating the eye tracker is that the subject looks at a plurality of different calibration points in turn, and the eye tracker determines the movement trajectory of the subject's eyeballs during the process and adjusts the positioning; each subject needs to recalibrate the eye tracker before conducting the experiment.

[0042] Furthermore, the specific process of building the experimental environment is to use Python to design a two-dimensional plane flight environment, set the flight parameters of manned aircraft and unmanned aircraft in a simulated air traffic control mission, where manned aircraft fly along fixed routes; unmanned aircraft fly in fixed directions and appear randomly.

[0043] Furthermore, based on the extracted sensitive features and the subjects' performance during the experiment, the machine learning model is combined to identify the subjects' situational awareness state under different experimental conditions.

[0044] Beneficial effects: The present invention combines eye movement data with EEG signals to classify and identify situational awareness. Compared with a single data source model, it can provide more comprehensive feature information and significantly improve classification accuracy. It can provide an important reference for the selection, training and assessment of controllers and has broad application prospects and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention is a flow chart of a method for identifying situational awareness of a controller during fusion operation of unmanned aerial vehicles in a controlled airspace. DETAILED DESCRIPTION

[0046] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0047] A method for identifying controller situational awareness during fusion operation of unmanned aerial vehicles in controlled airspace includes the following steps:

[0048] Build an experimental environment, set experimental condition parameters, and simulate the air traffic control task of the integrated operation of manned and unmanned aircraft. In this embodiment, the experimental environment is based on a two-dimensional plane, and the specific construction method is as follows: Use python to design a two-dimensional plane flight environment to simulate the flight of manned and unmanned aircraft on the route. Set the starting point and destination, and limit the movement trajectory of the manned aircraft to two mutually perpendicular routes to ensure that the manned aircraft runs along the predetermined route. UAVs appear randomly at different altitude levels, and their vertical positions are indicated by marking their altitude values ​​with text in the two-dimensional interface, and they fly in a fixed direction to form a UAV queue.

[0049] Different traffic conditions are set in the flight environment, including normal traffic conditions and special traffic conditions. Under normal traffic conditions, a certain number of manned and unmanned aircraft are set, and the experimental environment remains in a normal state; under special traffic conditions, the number of manned and unmanned aircraft is increased relative to normal traffic conditions. Special situations are introduced to increase the cognitive load and operational pressure of controllers, simulating more complex air traffic management scenarios.

[0050] Set the flight parameters of manned and unmanned aircraft to ensure that the experimental environment is consistent with the actual air traffic management scenario. Set a specific flight altitude and speed for manned aircraft to fly along the predetermined route; set a random flight altitude and speed for unmanned aircraft.

[0051] Two warning prompts are designed: text warning and highlight warning. Text warning displays the area and type of potential conflict in text on the operation interface; highlight warning highlights the area where conflict is about to occur in red on the operation interface, helping controllers quickly identify potential risks and make decisions.

[0052] After the experimental environment is set up, preparatory work before the experiment is carried out, including the installation and calibration of equipment to ensure that data collection proceeds normally.

[0053] Install the eye tracker for the subjects and calibrate the equipment to ensure that the eye tracker can accurately record the subjects' eye movement data.

[0054] Before the experiment began, the eye tracking device was started, and the Tobii Pro X3-120 eye tracker was used to collect the controller's eye movement data. All subjects were required to perform eye movement calibration using the device's calibration software, Tobii Pro Eye TrackerManager 2.5.0, to ensure the accuracy of the collected data. The calibration process displays multiple calibration points and asks the subjects to look at different positions so that the system can determine the movement trajectory of their eyeballs and perform precise positioning.

[0055] At the same time, wear an EEG recording device to record the EEG data generated during the experiment to ensure the integrity and accuracy of the experimental data.

[0056] Furthermore, the eye tracker and EEG equipment were activated to record the subjects’ eye movement data and EEG activity in real time during the task. The subjects’ performance in the task was monitored and recorded to ensure that all relevant data was fully collected for subsequent analysis.

[0057] The EEG amplifier used in this embodiment records 32 channels of EEG data to record the brain wave activity of the subject during the task. While recording the eye movement data and EEG signals, the computer tracks the subject's performance in the task, especially the accuracy of flight conflict handling and decision-making. At the same time, the computer automatically records the frequency and severity of conflicts, and records these task data synchronously with the eye movement and EEG data.

[0058] At the end of the experiment, stop the recording function of the device and save the collected eye movement and EEG data. Process these data separately to extract feature information. The subjects need to complete the situational awareness assessment scale to provide more reference information. In this embodiment, the specific data processing method is as follows:

[0059] S1.1 interpolates the collected eye movement data and fills in the missing eye movement data with a time less than the preset time of 75 milliseconds through linear interpolation to ensure the continuity and integrity of the data. The linear interpolation method is based on the timestamps and screen coordinates of the previous and next valid sample points to ensure that the interpolated data conforms to normal gaze and saccade behavior.

[0060] First, determine the scale factor S of data interpolation based on the timestamps of valid data samples before and after the missing time. factor , the formula is:

[0061]

[0062] Among them, t timestamp of sample to be replaced Indicates the timestamp of the sample that needs to be interpolated or replaced, t timestamp of first sample after gap Refers to the timestamp of the first valid sample after the data missing interval, t timestamp of last sample after gapRefers to the timestamp of the last valid sample after the data missing interval;

[0063] Furthermore, the time interpolation of the missing samples and the position information of the valid samples are calculated to fill the missing sample points. The formula is:

[0064] x m =x1+S factor ×(x2-x1)

[0065] y m =y1+S factor ×(y2-y1)

[0066] Among them, (x1, y1), (x2, y2) are the screen coordinates of the valid samples; (x m ,y m ) is the interpolated position of the missing sample point.

[0067] S1.2 performs noise reduction on the eye movement data after interpolation processing in S1.1, using a non-weighted moving average filter to reduce the impact of experimental equipment noise and slight eye movement jitter on the data. Assuming the filter sample window size is 2N, the formula is as follows:

[0068]

[0069] Where N represents the number of sample points taken before and after the current sample point, n represents the index position of the current sample point, k represents the interval from the current sample point forward, x(n) represents the sample value in the current time series, and y(n) represents the result of weighted averaging of the 2N+1 sample values ​​before and after the current sample point.

[0070] S1.3 calculates the eye movement velocity based on the eye movement data after noise reduction, sets the eye movement velocity threshold and classifies the corresponding sample points according to the eye movement velocity. The velocity of continuous sampling points is calculated based on the I-VT algorithm, and the velocity calculation formula is:

[0071]

[0072] Wherein, (x1, y1), (x2, y2) are the screen coordinates of two consecutive sample points, t1 is the timestamp of the sample point with screen coordinates (x1, y1), and t2 is the timestamp of the sample point with screen coordinates (x2, y2); if the speed of the sample point is lower than the threshold, the sample point is classified as a fixation point, and if the speed of the sample point is equal to or higher than the threshold, it is classified as a scanning point.

[0073] Key eye movement features are extracted from the data obtained in the above steps, including the number of fixation points, fixation time, scanning speed, scanning angle speed, pupil diameter, fixation time ratio and gaze entropy. The extracted key eye movement features are used to analyze the changes in the visual attention pattern of the controller during the task execution process, and provide basic data support for situational awareness assessment.

[0074] S2.1 Re-reference the collected EEG data. The binaural reference method is used to reference the EEG signal to the average value of the bilateral mastoid electrodes to eliminate the reference electrode noise, improve the spatial resolution of the signal, and ensure that the signal differences of different EEG channels reflect the actual activity of the brain area.

[0075] S2.2 performs a 1-30 Hz bandpass filter on the EEG signal to remove high-frequency EMG noise and low-frequency artifacts. The bandpass filter formula is:

[0076]

[0077] Among them, X(f) is the frequency domain representation of the EEG signal, f represents the frequency, which is the independent variable in the integration, and t represents the time variable, which represents the time point of the signal in the time domain;

[0078] S2.3 After bandpass filtering, the signal is further processed using independent component analysis to separate and remove common eye movement artifacts and noise signals such as poor electrode contact to improve the purity of the signal. Independent component analysis formula:

[0079] X=AS

[0080] Among them, X is the observation signal matrix, A is the mixing matrix, and S is the independent source signal matrix. Through the independent component analysis method, the independent source signal S is separated and the artifact components are removed.

[0081] S2.4 performs fast Fourier transform on the EEG signal processed by S2.3, converting the time domain signal into the frequency domain signal to extract the characteristic power of the four main frequency bands of δ wave, θ wave, α wave and β wave. The characteristic power of these frequency bands is used to evaluate the cognitive state and situational awareness level of the controller during the task execution. The formula of FFT is as follows:

[0082]

[0083] Among them, X(f) is the representation of the signal in the frequency domain, x(t) is the time domain signal, and f represents the frequency.

[0084] The key features of EEG signals were further analyzed, and the controllers’ situational awareness level was evaluated by calculating the θ / β ratio, (θ+α) / β ratio, α / β ratio, and (θ+α) / (α+β) ratio.

[0085] Furthermore, key features are extracted from the collected data, and in-depth analysis is performed through machine learning models to identify the subject's situational awareness state, and the situational awareness level is evaluated based on the data characteristics of different experimental conditions.

[0086] S3.1 For the collected eye movement data, the Levene test was first used to evaluate the homogeneity of variance. If the homogeneity was met, the t test was used to analyze the mean difference. If it was not met, the Mann-Whitney U test was used. For the EEG data, repeated measures analysis of variance was used.

[0087] S3.2 Based on the results of the sensitivity analysis, sensitive features significantly related to situational awareness are extracted. Subsequently, the sensitive features of the eye movement and EEG data are subjected to dimensionality reduction using principal component analysis, respectively, and the components that contribute the most to the variance are retained to remove redundant information and noise features. The principal component analysis dimensionality reduction results are also used to distinguish between linear features and nonlinear features: if the data concentrates variance information on a few principal components, it is judged to be a linear distribution; otherwise, it is a nonlinear distribution. Based on this analysis, linear features are input into the random forest model, and nonlinear features are input into the support vector machine model.

[0088] S3.3 inputs the reduced-dimensional sensitive feature data into a hybrid model combining support vector machine and random forest for classification. The support vector machine model uses radial basis function kernel function to perform high-dimensional mapping processing on nonlinear features to capture complex nonlinear patterns under situational awareness state and output corresponding classification results. Random forest inputs linear features, improves the stability of the model by integrating multiple decision trees, and outputs the classification results of situational awareness state.

[0089] To ensure the generalization ability of the model, K-fold cross validation is used to optimize the hyperparameters of the support vector machine and random forest. The specific operation is to divide the data set into K subsets, select K-1 subsets from the K subsets as the training set each time, and the remaining 1 subset as the validation set, repeating K times, so that each subset participates in the validation. For the support vector machine model, the classification effect of nonlinear features is optimized by adjusting the C and γ parameter combination of the radial basis function kernel function; in the random forest model, the classification performance of its linear features is optimized by adjusting the number of decision trees and the depth parameter. Cross-validation records the accuracy, precision and other indicators of each validation, and finally selects the best performing parameter combination for training to ensure the stability and prediction accuracy of the model.

[0090] Based on the extracted sensitive features and combined with the machine learning model, the situational awareness state of the subjects under different experimental conditions is identified.

[0091] The present invention discloses a method for identifying the situational awareness of air traffic controllers, which can accurately identify the situational awareness level of subjects, provide an important reference for the selection, training and assessment of air traffic controllers, and has broad application prospects and practical value.

Claims

1. A method for identifying situational awareness of controllers during fusion operation of unmanned aerial vehicles in controlled airspace, characterized in that: include: Build an experimental environment and use computers to simulate air traffic control tasks involving the integration of manned and unmanned aircraft; Install the eye tracker and EEG equipment, and calibrate the eye tracker according to each subject's situation; The subjects began to perform the computer-simulated air traffic control task, and the eye tracker and EEG equipment were started to collect the subjects' eye movement data and EEG signals in real time during the task. At the same time, the computer was used to record the subjects' performance during the experiment. After the simulated air traffic control mission is completed, the eye tracker and EEG equipment are stopped from collecting data, and the eye movement data and EEG signals collected by the eye tracker and EEG equipment are preliminarily processed to extract their key features; Based on the extracted key features of eye movement data and EEG signals, in-depth analysis is performed through machine learning models to identify the subject's situational awareness state.

2. The situational awareness recognition method according to claim 1, characterized in that: The preliminary processing of the eye movement data specifically includes the following steps: S1.1 performs interpolation processing on the collected eye movement data, and fills in the eye movement data with missing time less than the preset time through linear interpolation. First, the scale factor S of data interpolation is determined according to the timestamps of valid data samples before and after the missing time. factor , the formula is: Among them, t timestampofsampletobereplaced Indicates the timestamp of the sample that needs to be interpolated or replaced, t timestampoffirstsampleaftergap Refers to the timestamp of the first valid sample after the data missing interval, t timestampoflastsampleaftergap Refers to the timestamp of the last valid sample after the data missing interval; The time interpolation of the missing samples and the position information of the valid samples are calculated to fill the missing sample points. The formula is: x m =x1+S factor ×(x2-x1) yes m =y1+S factor ×(y2-y1) Among them, (x1, y1) and (x2, y2) are the screen coordinates of the valid sample points. m ,y m ) is the interpolated position of the missing sample point; S1.2 performs noise reduction on the eye movement data after interpolation processing in S1.1, using a non-weighted moving average filter, assuming the filter sample window size is 2N, and the formula is as follows: Where N represents the number of sample points taken before and after the current sample point, n represents the index position of the current sample point, k represents the interval from the current sample point forward, x represents the sample value in the current time series, and y(n) represents the result of weighted averaging of the 2N+1 sample values ​​before and after the current sample point. S1.3 calculates the eye movement velocity based on the eye movement data after noise reduction, sets the eye movement velocity threshold and classifies the corresponding sample points according to the eye movement velocity. The velocity calculation formula is: Wherein, (x1, y1), (x2, y2) are the screen coordinates of two consecutive sample points, t1 is the timestamp of the sample point with screen coordinates (x1, y1), and t2 is the timestamp of the sample point with screen coordinates (x2, y2); if the speed of the sample point is lower than the threshold, the sample point is classified as a fixation point, and if the speed of the sample point is equal to or higher than the threshold, it is classified as a scanning point.

3. The situational awareness recognition method according to claim 2, characterized in that: The key features extracted from the preliminarily processed eye movement data include: the number of fixation points, fixation time, scanning speed, scanning angle speed, pupil diameter, fixation time ratio and gaze entropy.

4. The situational awareness recognition method according to claim 1, 2 or 3, characterized in that: The preliminary processing of the EEG signal comprises the following specific steps: S2.1 Re-reference the collected EEG signals to the average value of the bilateral mastoid electrodes using the binaural reference method; S2.2 performs bandpass filtering on the EEG signal to extract a specific frequency range. The formula is: Among them, X(f) is the frequency domain representation of the EEG signal, f represents the frequency, which is the independent variable in the integration, t represents the time variable, which represents the time point of the signal in the time domain, and F1 and F2 are the lower and upper limit frequencies of the filter respectively; S2.3 uses independent component analysis to further remove artifacts from the signal. The independent component analysis formula is: X=AS Where X is the observed signal matrix, A is the mixing matrix, and S is the independent source signal matrix; S2.4 performs fast Fourier transform on the EEG signal processed by S2.3 to extract the characteristic power of the four frequency bands of delta wave, theta wave, alpha wave and beta wave in the EEG signal.

5. The situational awareness recognition method according to claim 4, characterized in that: The key features extracted from the preliminarily processed EEG signal include the θ / β ratio, the (θ+α) / β ratio, the α / β ratio and the (θ+α) / (α+β) ratio.

6. The situational awareness recognition method according to claim 5, characterized in that: The in-depth analysis through the machine learning model includes the following specific steps: S3.1 Perform sensitivity analysis on eye movement data and EEG signals to extract sensitive features that are significantly related to situational awareness; S3.2 uses principal component analysis to reduce the dimension of these sensitive features, remove redundant and noisy features, and distinguish linear features from nonlinear features; S3.3 inputs the sensitive feature data after dimensionality reduction into the machine learning model for classification and identification, and uses a hybrid model combining support vector machine and random forest model for analysis, where the support vector machine uses radial basis function kernel function to process the nonlinear part in the data.

7. The situational awareness recognition method according to claim 6, characterized in that: For the sensitivity analysis described in S3.1, for the collected eye movement data, the Levene test was first performed to assess the homogeneity of variance. If the homogeneity was met, the t test was used to analyze the mean differences. If it was not met, the Mann-Whitney U test was used; For EEG signals, repeated measures ANOVA was used.

8. The situational awareness recognition method according to claim 7, characterized in that: The specific process of calibrating the eye tracker is that the subject looks at a plurality of different calibration points in turn, and the eye tracker determines the movement trajectory of the subject's eyeballs during the process and adjusts the positioning; each subject needs to recalibrate the eye tracker before conducting the experiment.

9. The situational awareness recognition method according to claim 8, characterized in that: The specific process of building the experimental environment is to use Python to design a two-dimensional plane flight environment, set the flight parameters of manned aircraft and unmanned aircraft in a simulated air traffic control mission, where manned aircraft fly along fixed routes; unmanned aircraft fly in fixed directions and appear randomly.

10. The situational awareness recognition method according to claim 9, characterized in that: Based on the extracted sensitive features and the subjects' performance during the experiment, the machine learning model is combined to identify the subjects' situational awareness state under different experimental conditions.

Images