Smart airports have been advancing, and with it comes more autonomous special
purpose vehicles used at the airport airfield. But because of the sensitive
nature of airport operation data, autonomous cars find it hard to get real-time
flight operation info. Ground vehicles don’t have exact timing info for making
choices. To address this issue, this paper proposes a method for predicting the
arrival time of the aircraft through sliding, in order to achieve more efficient
coordination between the aircraft and ground vehicles. Using actual operating
data from Shanghai Hongqiao International Airport, a multi-dimensional feature
system that includes aircraft type, parking bay, spatial and temporal features,
and derived features is created. Multi-level detection nodes are arranged along
taxi routes, and an independent prediction model is built through the XGBoost
algorithm to achieve multi-node incremental prediction of aircraft taxiing time
to parking position. This model constantly changes its predictions as the
aircraft moves, creating alerts about ground vehicle passage based on these
predictions. It allows autonomous vehicles to always have the most recent
operational information available, which gives them a time reference for their
activities on the airport’s airside. Operational results show that the shortest
distance has the biggest impact on taxiing time (0.113), and calibrating the
prior detection point predictions also greatly affects taxiing time predictions.
Model accuracy increases more obviously as prediction nodes get closer to the
parking spot. The last prediction got a MAE of 5.96 seconds and an R2
of 0.94. Operational results show that this approach can well balance the
prediction accuracy and computational efficiency, satisfying the real-time
operation needs and offering reliable data support for aircraft-vehicle
coordination and intelligent airport ground management.