Real-Time Prediction of Runway Occupancy Buffers

Abstract

To improve runway safety and efficiency, real-time prediction of the time separation between successive flights using the same runway would be valuable. In this paper, we develop methods for such predictions, focusing on the time difference between when the prior aircraft exits the runway and the next arriving aircraft crosses the runway threshold, a metric we term runway occupancy buffer. We use two modeling frameworks: a two-stage modeling framework that predicts runway occupancy buffer through prediction of leading aircraft's runway occupancy time and trailing aircraft's required time till arrival; and an integrated modeling framework which directly predicts runway occupancy buffer. Machine learning techniques, linear regression and random forest regression, are applied to train the model. Seven models are investigated and compared at different distances from the runway threshold. Random forest regression outperforms other models, and it suggests that separation is the most important factor in predicting the runway occupancy buffer.