Research On Aircraft Taxi-Out Time Prediction For A-CDM

Taxi-out time prediction is an effective method for improving the predictability of the ground transportation. And it helps airport, airlines, air traffic control and other aircraft operations support department to achieve better collaborative decision making. At present,most airports in our country assume that the taxi-out time of departure aircraft is a default value of a uniform, and a series of decisions are made based on this assumption. For airports which are not busy, this approach is feasible. But for the busy airports, Usage of the unified default value will produce a big error because the actual taxi-out time distribution is big. This paper focuses on the research on aircraft taxi-out time prediction. And the specific work is as follows:As for the static prediction of aircraft taxi-out time, firstly, the taxi-out process is divided into two phases according to the aircraft departure process. One phase is the aircraft taxi to the runway queue without any obstacles and the other is the taxi-out delay while the aircraft in the queue. Secondly, the main influencing factors of the taxi-out time are determined using the correlation analysis. Based on the factors, an unimpeded taxi-out time calculation model is proposed. Meanwhile, the runway service process is modeled as M/G/1/∞ random service system, and a queue time prediction model is established based on Queuing Theory. Finally,the aircraft taxi-out time is got with the sum of unimpeded taxi-out time and the queue time.As for the real time prediction of aircraft taxi-out time, the paper proposed one model with two steps based on KNN and SVR. The first step predicts the number of departure and arrival aircrafts using the same runway during the aircraft taxiing out. Based on the prediction results in the first step and the main factors(i.e. taxiing out distance), the second step predicts the taxi-out time using SVR. In this paper, the historical data is grouped according to the traffic flow characteristics, and the prediction model is established for each group. The experiment result based on the actual operation data of Beijing Capital International Airport shows that compared with the prediction results of using SVR algorithm directly and other methods, the proposed method performs better in prediction accuracy of a single aircraft, and provide basis for the airport collaborative decision making.