Staged sliding time prediction and sliding control method and system

Abstract

The invention provides a staged taxiing time prediction and taxiing control method and system, and relates to the field of air transportation. A staged taxiing time prediction and taxiing control method comprises the steps that flight data of a plurality of departure taxiing processes are collected, and each departure taxiing process is divided into a plurality of taxiing stages; extracting flight arrival and departure characteristics of each taxiing stage and departure taxiing time by using the preprocessed flight data; selecting arrival and departure features of a part of flights in each taxiing stage to obtain a plurality of interaction features; training to obtain a departure taxiing time prediction model by using the arrival and departure features, interaction features and departure taxiing time of all flights in each taxiing stage; calculating the sliding time slot of each sliding stage to obtain the actual sliding time to optimize the model, and obtaining the predicted departure sliding time of each sliding stage. According to the method, the prediction precision of the sliding time is improved in the peak period and different sliding stages, so that the overall operation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of air transportation, and more particularly, to a method and system for predicting and controlling taxiing time in stages. Background Art

[0002] With the rapid development of the air transportation industry, global airport operations are facing increasing pressure. Against this background, airlines and airport management parties need to focus on improving flight punctuality and overall operational efficiency while adhering to the safety bottom line. Therefore, the prediction and control of aircraft taxiing time have received extensive attention from the academic and industrial circles in recent years and have become an important research direction for improving airport operational efficiency. Traditional taxiing time prediction usually regards the entire taxiing process as a whole and is difficult to effectively cope with the complexity of different stages (such as pushback and taxiing to the runway, etc.). This has led to insufficient prediction accuracy of taxiing time during peak periods and different taxiing stages, thus affecting the overall operation efficiency. At present, there is a need for a method and system for predicting and controlling taxiing time in stages, which can improve the prediction accuracy of taxiing time during peak periods and different taxiing stages, thereby accelerating the overall operation efficiency. Summary of the Invention

[0003] The object of the present invention is to provide a method and system for predicting and controlling taxiing time in stages, which can improve the prediction accuracy of taxiing time during peak periods and different taxiing stages, thereby accelerating the overall operation efficiency.

[0004] To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0005] A method for predicting and controlling taxiing time in stages, which includes: collecting flight flight data of multiple departure taxiing processes, each of the above departure taxiing processes being divided into multiple taxiing stages; extracting multiple flight arrival and departure characteristics and departure taxiing time of each of the above taxiing stages by using the preprocessed above flight flight data; selecting partial flight arrival and departure characteristics of each of the above taxiing stages multiple times to obtain interaction characteristics; preprocessing the multiple above interaction characteristics of each of the above taxiing stages; training a departure taxiing time prediction model by using all the above flight arrival and departure characteristics of each of the above taxiing stages, the preprocessed multiple above interaction characteristics, and the above departure taxiing time; optimizing the above departure taxiing time prediction model by calculating the taxiing time slots of each of the above taxiing stages to obtain the actual taxiing time; and obtaining the predicted departure taxiing time of each of the above taxiing stages during the departure taxiing process by using the optimized above departure taxiing time prediction model.

[0006] Each of the above-mentioned departure taxiing processes is divided into multiple taxiing stages, including: dividing each of the above-mentioned departure taxiing processes into three taxiing stages according to the period from after tow taxiing to before autonomous taxiing, from after autonomous taxiing to before entering the runway, and from after entering the runway to before takeoff.

[0007] The predicted departure taxiing times of each of the above-mentioned taxiing stages during the departure taxiing process are obtained by using the optimized above-mentioned departure taxiing time prediction model, including: for the taxiing stages from after tow taxiing to before autonomous taxiing and from after autonomous taxiing to before entering the runway, respectively select the corresponding above-mentioned departure taxiing time prediction model to obtain the taxiing time; for the taxiing stage from after entering the runway to before takeoff, use the statistical time median to obtain the taxiing time; sum the taxiing times of the three taxiing stages to obtain the predicted departure taxiing time of the entire departure taxiing process.

[0008] The interaction features are obtained by repeatedly selecting the in-port and out-port characteristics of some flights in each of the above-mentioned taxiing stages, specifically including: obtaining the usage frequencies of the runway and taxiway with respect to time and the space requirements of the runway and taxiway according to the number of in-ports; recording the idle times of the runway and taxiway according to the number of in-ports and out-ports or the time interval between the nearest in-port and out-port flights; calculating the in-out throughput according to the number of in-ports and out-ports to obtain the comparison result between the above-mentioned in-out throughput and the peak-hour throughput; performing dummy variable encoding on the combination of the parking position and the runway, and mapping each combination to a binary feature.

[0009] Preprocess the above-mentioned flight data, including: screening the flights that perform de-icing during taxiing, screening and deleting outliers, and using the median to fill in the missing values, either one or more of these methods;

[0010] Preprocess the above-mentioned interaction features, including: using label encoding for categorical variable features, using target encoding methods for the parking position and runway features, and performing standard scaling for standardizing numerical features, either one or more of these.

[0011] The actual taxiing time is obtained by calculating the taxiing time slots of each of the above-mentioned taxiing stages, including: using the predicted departure taxiing time and the actual departure taxiing time of each of the above-mentioned taxiing stages during the departure taxiing process to calculate the predicted taxiing time error of the above-mentioned departure taxiing time prediction model; obtaining the probability density function and its cumulative distribution function of the above-mentioned predicted taxiing time error through the KDE method; using the upper quantile of the above-mentioned cumulative distribution function as the upper bound and the symmetric quantile of the above-mentioned cumulative distribution function as the lower bound to obtain the predicted taxiing time error range; calculating the actual taxiing time through the above-mentioned predicted taxiing time error of each of the above-mentioned taxiing stages, or, after summing the predicted taxiing time error ranges of all the above-mentioned taxiing stages during the departure taxiing process, distributing them to each of the above-mentioned taxiing stages to obtain the actual taxiing time.

[0012] The above-mentioned departure taxi time prediction model adopts the Extreme Gradient Boosting model; when training the above-mentioned departure taxi time prediction model for the corresponding taxiing phase, any one or more of the model parameters such as the learning rate, maximum depth, sample sampling ratio, feature sampling ratio, and number of trees are optimized.

[0013] A phased taxi time prediction and taxi control system includes: a data collection module: collecting flight data of multiple departure taxiing processes, and each of the above-mentioned departure taxiing processes is divided into multiple taxiing phases; a feature extraction module: extracting multiple flight arrival and departure features and the departure taxi time of each of the above-mentioned taxiing phases by using the preprocessed above-mentioned flight data; a feature interaction module: obtaining interaction features by repeatedly selecting partial flight arrival and departure features of each of the above-mentioned taxiing phases; a model training module: preprocessing multiple above-mentioned interaction features of each of the above-mentioned taxiing phases; training a departure taxi time prediction model by using all the above-mentioned flight arrival and departure features of each of the above-mentioned taxiing phases, the preprocessed multiple above-mentioned interaction features, and the above-mentioned departure taxi time; a model optimization module: optimizing the above-mentioned departure taxi time prediction model by calculating the actual taxi time of each of the above-mentioned taxiing phases; a taxi splitting module: obtaining the predicted departure taxi time of each of the above-mentioned taxiing phases in the departure taxiing process by using the optimized above-mentioned departure taxi time prediction model.

[0014] An electronic device includes a memory, a processor, and a computer program running on the above-mentioned processor. When the above-mentioned processor executes the above-mentioned computer program, the steps of the above-mentioned phased taxi time prediction and taxi control method are implemented.

[0015] A computer-readable storage medium stores a computer program, and when the above-mentioned computer program is executed by a processor, the steps of the above-mentioned phased taxi time prediction and taxi control method are implemented.

[0016] Compared with the prior art, the present invention has at least the following advantages or beneficial effects:

[0017] The present application provides a method for predicting phased taxiing time and taxiing control. By dividing the departure taxiing phase into multiple taxiing phases, extracting multiple flight arrival and departure characteristics of flight data, and the taxiing time of each phase, a prediction model for the departure taxiing time for different phases is constructed; and by further mining the interaction characteristics of the departure taxiing through the extracted characteristics, a prediction model for the departure taxiing time for the corresponding taxiing phase is trained, and the taxiing time slot is calculated according to multiple taxiing phases, so as to optimize the model, and the total taxiing time can be obtained by summing up the predicted values of each taxiing phase using the model. The present invention combines spatial requirements, idle time, and throughput, trains the model to predict the taxiing time of runways and taxiways, and can improve the prediction accuracy of taxiing time during peak periods and different taxiing phases, thereby accelerating the overall operation efficiency. The present invention comprehensively considers the performance of each model in different taxiing phases, shows a lower error level and stronger generalization ability, and can provide more reliable and stable prediction results. It has stronger applicability in actual operation and performs well in airport scenarios with greater uncertainty in taxiing time. The reasonable setting of the taxiing time slot for each phase helps to cope with complex operating conditions and uncontrollable factors, ensuring the smoothness of the taxiing process. This flexible taxiing time slot management method not only provides an optimization direction for the future operation of the airport, but also provides the possibility of dynamic adjustment for taxiing in different seasons, time periods, and operating environments, ensuring the robustness and flexibility of the operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a flowchart of the method for predicting phased taxiing time and taxiing control according to the embodiment of the present invention;

[0020] Figure 2 It is a comparison chart of the predicted value and the true value of the departure taxiing time prediction model in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0022] Embodiment

[0023] As shown Figures 1 - 2 in the figure, the phased taxiing time prediction and taxiing control method provided by the embodiment of the present application includes: collecting flight flight data of multiple departure taxiing processes, each of the above departure taxiing processes being divided into multiple taxiing stages; using the preprocessed flight flight data to extract multiple flight arrival and departure characteristics and the departure taxiing time of each of the above taxiing stages; selecting partial flight arrival and departure characteristics of each of the above taxiing stages multiple times to obtain interaction characteristics; preprocessing the multiple interaction characteristics of each of the above taxiing stages; using all the above flight arrival and departure characteristics of each of the above taxiing stages, the preprocessed multiple interaction characteristics, and the above departure taxiing time to train a departure taxiing time prediction model; calculating the actual taxiing time by calculating the taxiing time slots of each of the above taxiing stages to optimize the above departure taxiing time prediction model; using the optimized above departure taxiing time prediction model to obtain the predicted departure taxiing time of each of the above taxiing stages during the departure taxiing process.

[0024] Each of the above departure taxiing processes is divided into multiple taxiing stages, including: dividing each of the above departure taxiing processes into three taxiing stages according to after tow taxiing until before autonomous taxiing, after autonomous taxiing until before entering the runway, and after entering the runway until before takeoff.

[0025] The above using the optimized above departure taxiing time prediction model to obtain the predicted departure taxiing time of each of the above taxiing stages during the departure taxiing process includes: for the taxiing stages after tow taxiing until before autonomous taxiing and after autonomous taxiing until before entering the runway, respectively selecting the corresponding above departure taxiing time prediction model to obtain the taxiing time; for the taxiing stage after entering the runway until before takeoff, using the statistical time median to obtain the taxiing time; summing the taxiing times of the three taxiing stages to obtain the predicted departure taxiing time of the entire departure taxiing process.

[0026] The above selecting partial flight arrival and departure characteristics of each of the above taxiing stages multiple times to obtain interaction characteristics specifically includes: obtaining the usage frequency of the runway and taxiway according to the arrival number and time, and the space requirements of the runway and taxiway; recording the idle time of the runway and taxiway according to the arrival and departure numbers or the time interval between the nearest arrival and departure flights; calculating the throughput of arrivals and departures according to the arrival and departure numbers to obtain the comparison result of the above throughput of arrivals and departures and the peak hour throughput; performing dummy variable encoding on the combination of the parking position and the runway, and mapping each combination to a binary feature.

[0027] Preprocess the above flight data, including any one or more of the following methods: screening flights that perform de-icing during taxiing, screening and deleting outliers, and filling missing values with the median; preprocess the above interaction features, including any one or more of the following: using label encoding for categorical variable features, using target encoding for stand and runway features, and standardizing numerical features using standard scaling.

[0028] The actual taxi time is obtained by calculating the taxi time slots of the above taxiing stages, including: using the predicted taxi-out time and the actual taxi-out time of each of the above taxiing stages during the taxi-out process to calculate the prediction taxi time error of the above taxi-out time prediction model; obtaining the probability density function and its cumulative distribution function of the above prediction taxi time error through the KDE method; using the upper quantile of the above cumulative distribution function as the upper bound and the symmetric quantile of the above cumulative distribution function as the lower bound to obtain the prediction taxi time error range; calculating the actual taxi time through the above prediction taxi time error of each of the above taxiing stages, or, after summing the prediction taxi time error ranges of all the above taxiing stages during the taxi-out process, allocating them to each of the above taxiing stages to obtain the actual taxi time.

[0029] The above taxi-out time prediction model uses the Extreme Gradient Boosting model; when training the above taxi-out time prediction model for the corresponding above taxiing stage, optimize any one or more of the model parameters including the learning rate, maximum depth, sample sampling ratio, feature sampling ratio, and the number of trees.

[0030] The embodiments of the present invention divide the taxi-out process of civil aircraft into three stages, and the specific definitions are as follows: Stage 1: After towing taxiing and before autonomous taxiing, that is, from the moment of removing the wheel chocks to the moment of starting the aircraft's own power taxiing. After the tractor tows the aircraft to the designated position, the aircraft brakes and starts the engine, and then the tractor withdraws. After completing this series of operations, the pilot conducts an inspection and sends a taxiing request to the ground control. After receiving the taxiing permission, the pilot enters the next stage. Stage 2: After autonomous taxiing and before entering the runway, that is, from the start of own power taxiing to the moment the aircraft enters the runway. This stage includes autonomous power taxiing on the taxiway and queuing to wait to enter the runway. Stage 3: After entering the runway and before takeoff, that is, from the moment the aircraft enters the runway to the actual takeoff moment. In this stage, the aircraft stays on the runway, waits for the takeoff instruction, and finally completes the takeoff.

[0031] From the perspective of air traffic control work, each stage of the departure taxiing process has clear air traffic control responsibilities and requirements. In stage one, apron controllers need to ensure that there are no interferences from other aircraft or ground vehicles after the aircraft reaches the designated position. Apron controllers need to communicate with the pilots in a timely manner to ensure safety. In stage two, apron controllers need to continuously monitor the taxiing position and status of the aircraft to ensure that it moves along the designated taxiing route. When the aircraft reaches the junction of the main taxiway and the apron, it will be handed over to the ground seat of the tower for continued air traffic control monitoring. In stage three, the tower control needs to monitor the runway environment in real time to ensure the safety of the aircraft during takeoff. After confirming safety, an instruction to allow takeoff will be issued.

[0032] Preprocessing flight flight data mainly includes screening out flights that perform de-icing during taxiing, screening and deleting outliers, and filling missing values with the median. Moreover, reprocessing the flight flight data to obtain features such as the number of arrivals and departures during the flight taxiing. When constructing a departure taxiing time prediction model, the random sampling method can be used to divide the training set and the test set, and 80% of the data is selected as the training set, and the remaining 20% is the test set.

[0033] This article will select multiple flight arrival and departure features, such as: any one or more of the information of date, flight number, VIP identification, aircraft type, parking position, runway, SID (departure procedure), ETD time, PUS time, TAX time, aircraft entering the runway time, ATD time, time entering the de-icing apron, and initial sector. Flight arrival and departure features, such as runway, taxiway, parking position, number of arrivals and departures, arrival and departure time intervals, peak hour throughput, airline, flight month, judgment result of smooth or congested, aircraft type, departure procedure, initial sector, and VIP flight.

[0034] Optionally, the specific selected multiple flight arrival and departure features are shown in Table 1 below:

[0035]

[0036]

[0037]

[0038] Table 1

[0039] To further improve the prediction ability of the model, interaction features are introduced. Interaction features refer to combining multiple features together to generate new features, thereby capturing more complex relationships in the data.

[0040] Optionally, preprocess the interaction features as follows:

[0041] (1) Categorical variable encoding: For multiple categorical features such as aircraft type, airline, SID (Standard Instrument Departure), initial sector, and smooth / congested judgment, Label Encoder was used. Label Encoder can convert categorical features into integers, enabling these features to be directly utilized by the model while preserving the order relationship between categories.

[0042] (2) Target encoding: To further enhance the expressive power of categorical features, the Target Encoding method was adopted for the parking position and runway features. This method converts categorical variables into numerical features by calculating the average value between each category and the target variable. This conversion not only improves the model's ability to capture subtle relationships between categories but also effectively enhances the discrimination effect of categorical variables.

[0043] (3) Numerical feature standardization: Standard Scaler was used to standardize numerical features, adjusting the mean of the feature values to 0 and the variance to 1. Standardization can eliminate the differences between different feature dimensions and avoid the unbalanced influence of certain features on model training. In the embodiments of this application, features such as the number of inbound flights, the number of outbound flights, and the number of inbound and outbound flights have all been standardized to ensure that all numerical features participate in model training on the same scale, improving the convergence and prediction performance of the model.

[0044] Optionally, the interaction features are shown in Table 2 below:

[0045]

[0046] Table 2

[0047] The specific analysis process of the principle of the above interaction features is as follows:

[0048] Number of inbound flights and number of outbound flights: Since the increase in inbound flights may lead to an increase in the utilization frequency of runways and taxiways, while the increase in outbound flights may lead to an increase in the space demand for takeoff preparation. The combined effect of the two can affect the utilization rate of taxiways and runways, especially during high-density traffic periods. Through the interaction feature "number of inbound flights * number of outbound flights", the model can identify the impact on taxiing time when both change simultaneously, rather than just the impact on a single variable separately.

[0049] Number of arrivals and departures and closest flight interval: During taxiing, the frequency of runway use and flight intervals are key factors affecting taxiing time. When the number of arrivals and departures increases, runway use becomes more intensive, especially in a short period of time. Changes in flight intervals directly affect the idle time of the runway. Shorter flight intervals mean higher runway utilization and possibly longer taxiing waiting time. By analyzing "Number of arrivals and departures / closest flight interval", you can understand how busy the runway is during taxiing, especially for resource usage in a short period of time.

[0050] Peak hour throughput and number of arrivals and departures: The relationship between peak hour throughput (i.e. the maximum number of flights arriving and departing at an airport per hour) and the number of arrivals and departures reflects the linkage between the demand for airport resources and the maximum throughput capacity during busy hours. The interaction term between the number of arrivals and departures and peak hour throughput can be used to evaluate how the airport's throughput capacity and resource allocation affect the overall flight taxi time within a certain time range. Differences in the number of arrivals and departures can affect the airport's resource allocation, thereby affecting throughput and taxi time.

[0051] Parking stands and runways: Different parking stand and runway combinations mean that the flight needs to taxi to different distances from the parking stand to the runway, which affects the taxi time. By capturing these features, the model can better handle the efficiency differences of different parking stand and runway combinations, thereby more accurately predicting taxi time. Through dummy variable (One-hot) encoding, each parking stand and runway combination is mapped to a binary feature, and the model can learn the impact of different combinations on taxi time. The generated features are dummy variable columns corresponding to each "parking stand-runway combination", which helps the model better identify the potential impact of different parking stand and runway combinations on taxi time.

[0052] Extreme Gradient Boosting (XGBoost) is an efficient implementation based on gradient boosting trees. It enhances the prediction accuracy of the model by integrating multiple weak learners (decision trees). This integration method not only improves the performance of the model but also optimizes the combination of individual weak learners through the gradient boosting framework, thereby achieving higher accuracy and efficiency in prediction tasks. Traditional taxiing time prediction tends to handle the overall taxiing process, while this model can independently adjust and optimize each stage by splitting the taxiing process into multiple stages. The taxiing time is divided into three stages. Among them, Stage 1 and Stage 2 account for a relatively large proportion of the entire taxiing time and have relatively high requirements for prediction accuracy. Therefore, when judging whether the proportion of the taxiing time of each of the above taxiing stages in the departure taxiing process exceeds the preset threshold, it can be adjusted according to past experience to select the extreme gradient boosting model to predict Stage 1 and Stage 2 separately. Stage 3 accounts for a relatively small proportion of the taxiing time, so the median of the statistical time of Stage 3 is directly used as the prediction result of Stage 3 in the model. The sum of the prediction results of Stage 1, Stage 2, and Stage 3 is the predicted value of the entire process taxiing time.

[0053] The parameters of the model of the present invention are debugged. The main parameters include: learning_rate (learning rate), max_depth (maximum depth), subsample (sample sampling ratio), colsample_bytree (feature sampling ratio), n_estimators (number of trees). After multiple rounds of testing and parameter tuning, the key parameters of the models for Stage 1 and Stage 2 are shown in Table 3 below.

[0054]

[0055] Table 3

[0056] In each stage of aircraft taxiing, the application of taxiing time slots may vary. For example, during the aircraft pushback stage, the taxiing time slot may be used as redundant time to ensure the safe pushback of the aircraft from the parking position and reaching the taxiway. During the self-powered taxiing stage, the taxiing time slot may be used to cope with traffic congestion or situations that require switching between multiple taxiways. During the waiting stage before the runway, the dynamic taxiing time slot may be used to cope with changes in the takeoff sequence or other temporary airport operation restrictions. When calculating the taxiing time slot for Stage 3, the time slot range is small and the significance is small, so it is no longer considered. Therefore, the taxiing time slots are calculated only for Stage 1 and Stage 2, that is, the pushback taxiing time slot and the self-powered taxiing time slot on the taxiway.

[0057] Kernel density estimation is a non-parametric method for estimating the probability density function. To ensure the accuracy of kernel density estimation, the Gaussian kernel function is selected in the embodiments of the present application to fit the error distribution.

[0058] First, the prediction error of the flight taxiing time is:

[0059] e i = t pre-taxi，i - t taxi，i (1)

[0060] where t pre-taxi，i is the departure taxi-out time of the i-th flight predicted by the model; t taxi，i is the actual departure taxi-out time of the i-th flight.

[0061] The probability density function f(e) of the error and its cumulative distribution function F(e) can be obtained by the KDE method:

[0062]

[0063] where n is the number of data points, x represents a certain value of the error, x i is the error of the actual observed value, h is the bandwidth parameter, and K is the Gaussian kernel function.

[0064] The method for calculating the taxiing time slot is to determine an error range to ensure that the predicted taxiing time plus the value within this range makes the on-time rate of the actual taxiing time p. In other words, it is desired to find two boundaries e low and e high that satisfy the following conditions:

[0065] P(e low ≤ e ≤ e high ) = p (4)

[0066] The upper bound e high is the quantile of the cumulative distribution function:

[0067] F(e high ) = p (5)

[0068] The lower bound e low is the symmetric quantile of the cumulative distribution function:

[0069] F(e low ) = 1 - p (6)

[0070] Through the inverse function F -1 (p) of the cumulative distribution function, we can obtain:

[0071] e high = F -1 (p) (7)

[0072] e low = F -1 (1 - p) (8)

[0073] The error distribution obtained through KDE (K Desktop Environment), so it can be deduced that for arriving at the runway with a punctuality rate of p, the allowable deviation range of the departure time is: [-e high , -e low .

[0074] Flight flight data is collected according to different taxiing stages of multiple departure taxiing processes. The number of flight flights includes arrival and departure data, as shown in Table 4 and Table 5 below. Among them, Table 4 is departure flight data, and Table 5 is arrival flight data.

[0075] Date Flight Number VIP Indicator Aircraft Type Parking Bay Runway SID (Departure Procedure) 2021 / 02 / 28 CBJ5315 / A320 155 11L DOT01D 2021 / 02 / 28 CES9724 / B737 122 35R PEG01D 2021 / 02 / 28 CBJ5351 / A333 153 11L ELK01D

[0076] ETD Time RDY Time PUS Time TAX Time Time when the aircraft enters the runway CTOT 22:30:00 22:17:34 22:19:18 22:26:18 22:53:20 22:35:00 22:40:00 22:26:31 22:29:03 22:41:32 23:24:14 22:46:00 22:45:00 22:44:10 22:45:44 22:55:31 23:45:37 22:45:00

[0077] ATD Time Time to enter the de - icing apron Initial Sector 22:54:37 22:26:57 EDLV 23:26:40 22:47:06 WDLV 23:46:47 22:56:58 EDLV

[0078] Table 4

[0079]

[0080] Table 5

[0081] Flight flight data includes: any one or more of the information such as date, flight number, VIP identification, aircraft type, parking position, runway, SID (departure procedure), ETD time, PUS time, TAX time, aircraft entering the runway time, ATD time, time entering the de-icing apron, initial sector, number of arrivals and departures during flight taxiing, nearest flight interval, and airport peak hour throughput. TAX time, for example: taxi out time (min), taxi_phase1 time, taxi_phase2 time, taxi_phase3 time.

[0082] Preprocessing the original flight flight data mainly includes screening out flights that perform de-icing and anti-icing during taxiing, screening and deleting outliers, and filling in missing values with the median. Extract the arrival and departure characteristics and departure taxiing time of each taxiing stage from the flight flight data; thus select some arrival and departure characteristics to further obtain interaction characteristics. Use the random sampling method to divide the training set and the test set, and select 80% of the data as the training set, and the remaining 20% as the test set. Among them, the arrival and departure characteristics and interaction characteristics are as described above in the example, the specific arrival and departure characteristics are shown in Table 1, and the interaction characteristics are shown in Table 2. The arrival and departure characteristics and interaction characteristics of different taxiing stages of the same departure taxiing process can be partially the same, so as to predict the taxiing time of multiple taxiing stages of the departure taxiing process at the same time. The arrival and departure characteristics and interaction characteristics of each taxiing stage are introduced into the departure taxiing time prediction model for training, and the taxiing time of each stage can be predicted according to the characteristics of each flight data.

[0083] Table 6 below shows the comparison between the predicted values and the actual values of the pushback times for several randomly selected flights.

[0084] True Value (minutes) Predicted Value (minutes) 18.52 17.59 19.53 21.75 15.53 16.15 15.57 14.49 16.6 17.34 33.93 31.38

[0085] Table 6

[0086] To visually evaluate the performance of the departure pushback time prediction model, 200 predicted values and actual values were randomly selected for comparison. The results are as Figure 2 shown, where the blue broken line represents the actual pushback time and the red broken line represents the predicted pushback time by the model. The overall trend of the model's prediction results is consistent with the actual pushback time. Especially within the medium pushback time period (10 - 45 minutes), the predicted values are relatively accurate.

[0087] Furthermore, a comprehensive evaluation was conducted on four models, including Random Forest, Support Vector Regression, Extreme Gradient Boosting (XGBoost), and this model. By using evaluation metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Coefficient of Determination (R 2 ^2), the performance of the models on the training set and the test set was compared. The evaluation results of the departure pushback time prediction model are shown in Table 7.

[0088]

[0089] Table 7

[0090] The prediction accuracy of the models is shown in Table 8 below:

[0091]

[0092]

[0093] Table 8

[0094] The errors between the predicted values and the actual values of the departure pushback time are within the ranges of ±1, ±2, ±3, ±4, and ±5 minutes. The Extreme Gradient Boosting model and this model outperform by several percentage points in all error ranges. Especially at ±5 minutes, the accuracy of this model is 96%.

[0095] There are two strategies for dynamic taxi time slot control to allocate time slots. The first strategy is to calculate taxi time slots independently for each stage. In this method, an independent error range is set for each stage, and the allowable taxi time slots are derived based on the predicted time and error of the stage. Each stage can independently meet the set on-time rate during the entire taxiing process without affecting each other. The second strategy is to calculate the overall taxi time slot range and allocate it to each taxiing stage according to the proportion of the duration. This method first determines an overall allowable taxi error range, and then allocates it according to the proportion of the taxi time of each stage to ensure the overall on-time rate. The calculation results of the taxi time slots under Strategy 1 are shown in Table 9 below:

[0096]

[0097]

[0098] Table 9

[0099] The calculation results of the taxi time slots under Strategy 2 are shown in Table 10 below:

[0100]

[0101] Table 10

[0102] In actual operation, CTOT (Calculated Takeoff Time) is an important reference standard for analyzing taxi time slots. According to the "Rules for the Operation of Air Traffic Flow Management in Civil Aviation of China (Trial)" of the Civil Aviation Administration of China, the CTOT time slot tolerances are divided into two categories: Category I tolerance (-5, +10 minutes) and Category II tolerance (-3, +3 minutes). Therefore, when designing taxi time slots, it is necessary to ensure that the taxiing can be finally completed within this time window.

[0103] In Strategy 1, the advance and delay times in Glide Phase 1 and Glide Phase 2 both increase with the improvement of the on-time rate. Specifically, the allowed advance times at 80% and 85% on-time rates in Strategy 1 are 209 seconds and 267 seconds respectively, and the allowed delay times are 238 seconds and 290 seconds respectively, both meeting the requirements of CTOT Class 1 tolerance (-5, +10 minutes). However, at 90% on-time rate, the allowed advance time of 356 seconds does not meet the CTOT Class 1 tolerance, so the CTOT Class 1 tolerance should be used as the boundary standard, and the allowed advance time in each phase should be reduced to a sum of 300 seconds. The advantage of this strategy is that it can precisely control the on-time rate of each glide phase, ensuring that the start time is within the error range, especially suitable for situations with large differences in glide time and error characteristics. However, independent calculation may lead to cumulative errors, and its own time cost is relatively large. In contrast, the advance times in Strategy 2 are 105 seconds at 80% on-time rate, 133 seconds at 85% on-time rate, and 176 seconds at 90% on-time rate. At the same time, the delay times at 80%, 85%, and 90% on-time rates are 120 seconds, 142 seconds, and 178 seconds respectively, all meeting the higher requirements of CTOT Class 2 tolerance (-3, +3 minutes). The advantage of this strategy is that it considers the error range as a whole, helps to reduce cumulative errors, and is simple to implement without calculating stage by stage. However, if the glide time distributions in each phase vary greatly and the error is relatively high in a specific phase, it may lead to unreasonable allocation and cannot effectively meet the punctuality of each phase. Generally speaking, Strategy 1 provides stable time control at high on-time rates, but its time cost is relatively high. In contrast, Strategy 2 can more effectively control the glide time and improve the efficiency of flight scheduling under the condition of meeting high on-time rates.

[0104] The present invention divides the departure taxiing process into multiple phases according to the autonomous taxiing process and the taxiway, and uses the Extreme Gradient Boosting model (XGBoost model) to predict the taxiing time of each phase, and then sums them up for prediction. This method overcomes the limitation of traditional models that regard the entire taxiing process as a single process, making the prediction more targeted and accurate. The performances of four machine learning models (Random Forest, Support Vector Regression, Extreme Gradient Boosting, and this model) in predicting the departure taxiing time are evaluated. The results show that the average absolute error of this model on the test set is 1.796, which is better than Random Forest, Support Vector Regression, and Extreme Gradient Boosting. The mean squared error is significantly lower than other models, showing a smaller fluctuation of prediction errors. The coefficient of determination is 0.870, indicating its strong explanatory ability for the change trend of taxiing time. The CV MAE of the cross-validation result is 1.277±0.016, showing excellent stability. In terms of prediction accuracy, the accuracy rate of this model within ±5 minutes is 96%, which is higher than other models.

[0105] In addition, this paper presents a taxiing time slot control method based on chance constraints. By fitting the prediction error of the taxiing time, the taxiing time slots for each stage are calculated, enabling the airport to achieve precise control of the taxiing process. This time slot control method provides the airport with a more flexible time management strategy, ensuring punctuality during peak periods and enhancing operational efficiency during off-peak periods. With the application of big data and artificial intelligence technologies, future taxiing control will rely more on accurate taxiing time prediction and analysis.

[0106] In summary, a phased taxiing time prediction and taxiing control method and system provided by the embodiments of this application:

[0107] By dividing the departure taxiing stage into multiple taxiing stages, extracting multiple flight arrival and departure characteristics of flight data, as well as the taxiing time for each stage, a departure taxiing time prediction model for predicting taxiing in different stages is constructed; and through the extracted characteristics, the interaction characteristics of departure taxiing are further mined, training the departure taxiing time prediction model for the corresponding taxiing stage, calculating the taxiing time slots according to multiple taxiing stages, thereby optimizing the model, and using the model to predict that the sum of the taxiing times of each taxiing stage can obtain the total taxiing time. The present invention combines space requirements, idle time, and throughput, trains the model to predict the taxiing time on the runway and taxiway, and can improve the prediction accuracy of the taxiing time during peak periods and different taxiing stages, thereby accelerating the overall operational efficiency. The present invention comprehensively considers the performance of each model in different taxiing stages, shows a lower error level and stronger generalization ability, and can provide more reliable and stable prediction results. It has stronger applicability in actual operation and performs well in airport scenarios with greater uncertainty in taxiing time. The reasonable setting of the taxiing time slots for each stage helps to cope with complex operating conditions and uncontrollable factors, ensuring the smoothness of the taxiing process. This flexible taxiing time slot management method not only provides an optimization direction for the future operation of the airport, but also provides the possibility of dynamic adjustment for taxiing in different seasons, time periods, and operating environments, ensuring the robustness and flexibility of operation. The present invention demonstrates higher accuracy and stronger generalization ability, and while ensuring the prediction accuracy, can provide more relevant characteristics for each taxiing stage. It is adapted to the complex and changeable airport operating environment and provides reliable support for flight scheduling and operation management.

[0108] The above are only the preferred embodiments of this application and are not used to limit this application. For those skilled in the art, various changes and modifications can be made to this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A method for predicting staged taxiing time and taxiing control, characterized in that, Including: Collecting flight data of multiple departures during taxiing, and each of the said departure taxiing processes is divided into multiple taxiing stages; Extracting multiple flight arrival and departure characteristics of each of the said taxiing stages, as well as the departure taxiing time, by using the preprocessed flight data; Selecting partial flight arrival and departure characteristics of each of the said taxiing stages multiple times to obtain interaction characteristics; Preprocessing multiple said interaction characteristics of each of the said taxiing stages; using all the flight arrival and departure characteristics of each of the said taxiing stages, the preprocessed multiple said interaction characteristics, and the said departure taxiing time to train a departure taxiing time prediction model; Optimizing the departure taxiing time prediction model by calculating the taxiing time slots of each of the said taxiing stages to obtain the actual taxiing time; Using the optimized departure taxiing time prediction model to obtain the predicted departure taxiing time of each of the said taxiing stages during the departure taxiing process.

2. The phased taxi time prediction and taxi control method according to claim 1, characterized in that, Each of the said departure taxiing processes is divided into multiple taxiing stages, including: Dividing each of the said departure taxiing processes into three taxiing stages according to the period from after tow taxiing to before autonomous taxiing, from after autonomous taxiing to before entering the runway, and from after entering the runway to before takeoff.

3. The phased taxi time prediction and taxi control method according to claim 2, characterized in that The step of using the optimized departure taxiing time prediction model to obtain the predicted departure taxiing time of each of the said taxiing stages during the departure taxiing process includes: For the taxiing stages from after tow taxiing to before autonomous taxiing and from after autonomous taxiing to before entering the runway, respectively selecting the corresponding departure taxiing time prediction model to obtain the taxiing time; for the taxiing stage from after entering the runway to before takeoff, using the statistical time median to obtain the taxiing time; summing up the taxiing times of the three taxiing stages to obtain the predicted departure taxiing time of the entire departure taxiing process.

4. The phased taxi time prediction and taxi control method according to claim 1, characterized in that, The step of selecting partial flight arrival and departure characteristics of each of the said taxiing stages multiple times to obtain interaction characteristics specifically includes: Obtaining the usage frequencies of runways and taxiways according to the number of arrivals and the time, as well as the spatial requirements of runways and taxiways; Recording the idle times of runways and taxiways according to the number of arrivals and departures or the time interval between the nearest arrival and departure flights; Calculating the throughput of arrivals and departures according to the number of arrivals and departures, and obtaining the comparison result between the said throughput of arrivals and departures and the peak-hour throughput; Performing dummy variable encoding on the combinations of parking positions and runways, and mapping each combination to a binary feature.

5. The phased taxi time prediction and taxi control method according to claim 2, wherein Preprocessing the flight data includes: Selecting any one or more of the methods of screening flights that perform de-icing during taxiing, screening and deleting outliers, and filling missing values with the median; Preprocessing the interaction characteristics includes: Using any one or more of label encoding for categorical variable features, target encoding for parking position and runway features, and standard scaling for normalizing numerical features.

6. The phased taxi time prediction and taxi control method according to claim 1, characterized in that The step of obtaining the actual taxiing time by calculating the taxiing time slots of each of the said taxiing stages includes: Using the predicted off - block taxi time and the actual off - block taxi time of each of the taxiing stages during the off - block taxiing process, calculate the prediction taxi time error of the off - block taxi time prediction model; obtain the probability density function and its cumulative distribution function of the prediction taxi time error through the KDE method; use the quantile of the cumulative distribution function as the upper bound and the symmetric quantile of the cumulative distribution function as the lower bound to obtain the prediction taxi time error range; calculate the actual taxi time through the prediction taxi time error of each of the taxiing stages, or, after summing up the prediction taxi time error ranges of all the taxiing stages during the off - block taxiing process and distributing them to each of the taxiing stages, obtain the actual taxi time.

7. The phased taxi time prediction and taxi control method according to claim 1, wherein The off - block taxi time prediction model adopts the extreme gradient boosting model. When training the off - block taxi time prediction model corresponding to each of the taxiing stages, optimize any one or more of the model parameters including the learning rate, the maximum depth, the sample sampling ratio, the feature sampling ratio, and the number of trees.

8. A staged taxiing time prediction and taxiing control system, characterized in that, It includes: Data acquisition module: Collect flight flight data of multiple off - block taxiing processes, and each of the off - block taxiing processes is divided into multiple taxiing stages. Feature extraction module: Use the pre - processed flight flight data to extract multiple flight arrival and departure features and the off - block taxi time of each of the taxiing stages. Feature interaction module: Select some of the flight arrival and departure features of each of the taxiing stages multiple times to obtain interaction features. Model training module: Pre - process the multiple interaction features of each of the taxiing stages; use all the flight arrival and departure features of each of the taxiing stages, the pre - processed multiple interaction features, and the off - block taxi time to train and obtain the off - block taxi time prediction model. Model optimization module: Optimize the off - block taxi time prediction model by calculating the actual taxi time through the taxi time slots of each of the taxiing stages. Taxi splitting module: Use the optimized off - block taxi time prediction model to obtain the predicted off - block taxi time of each of the taxiing stages during the off - block taxiing process.

9. An electronic device, comprising a memory, a processor, and a computer program running on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the method for predicting and controlling taxiing time in stages as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for predicting and controlling taxiing time in stages as described in any one of claims 1 to 7.

Images