Prediction Of Taxi Time Of Departure Aircraft Based On Uncertain Parameters

The surface layout of large busy airport is complex,and the number of aircrafts landing and departing the airport is superimposed,which causes the airport to be in high load operation for a long time,and makes the aircraft delay due to long taxiing time.At the same time,the extensive use of Airport-Collaborative Decision Management system(A-CDM),which integrates the data of air traffic management,airport and airlines,provides a data basis for the study of airport surface operation efficiency and the prediction of aircraft critical time.Taxi time,as one of the key indicators to evaluate the ground operation support efficiency of the airport,the accuracy of its prediction will directly affect the takeoff and landing sequence of the aircraft,and then affect the operational efficiency of the airport.Based on the above reasons,this paper establishes a deep learning based model to predict the departure aircraft taxiing time under uncertainty circumstances.First of all,the data needed for the study are preprocessed and the factors affecting the taxi time are analyzed.According to the multi-source heterogeneous data used in this study,the data standardization algorithm is used to eliminate the difference of different data dimensions to improve the data quality.In order to solve the problem of model robustness reduction caused by data imbalance,data resampling technology is used to balance the amount of each type of data.At the same time,the paper analyzes the factors that affect the aircraft taxiing time and the influence degree.According to the data characteristics of the factors,it further divides them into static deterministic factors(aircraft type,runway operation mode,taxiing distance,etc.)and dynamic uncertain factors(airport surface flow,weather,etc.).Secondly,the dynamic uncertainty factors(airport surface flow)are predicted.According to the temporal-spatial data,the sliding time window method is leveraged to smooth the data to ensure the continuity and stability of the subsequent prediction.The combined model of Long Short Term Memory Network and Deep Neural Network(D-LSTM)is established to predict the actual flow of airport surface.The D-LSTM model is leveraged to forecast and verify the surface flow of Hong Kong airport.The results show that the accuracy of the D-LSTM model is 88.0 %,which effectively captures the pattern and periodic characteristics of surface traffic as compared to other machine learning models.Finally,a deep learning model is established to predict the departure aircraft taxiing time.According to the definition and statistical method of aircraft taxiing time at the present research,two models are established to predict the departure aircraft taxiing time,which are: Wide-Deep model for aircraft taxiing time prediction based on historical statistical data(dynamic uncertainty is not predicted);and dynamic Wide-Deep model for changing airport surface flow into uncertainty.The results show that the prediction accuracy of the above two deep learning models is significantly better than other machine learning algorithms,which can be used to predict the taxiing time of departure aircraft in large airports under various meteorological conditions.