Designing a real-time ramp arrival prediction tool

Airline efficiency may be increased by the timely dissemination of accurate ramp time-of-arrival predictions for inbound aircraft. Potential benefits include improvements to gate management, ramp management, and personnel/equipment resource allocation. These benefits can lead to shorter aircraft turn-around times and reduced airline operating costs. Collaboration with Federal Express has identified the following criteria for a real-time prediction system: a gate time-of-arrival accuracy of plus-or-minus two minutes for 95 percent of all arrivals, predicted as an aircraft crosses the TRACON boundary. The system should extend to deliver more accurate time-of-arrival predictions as a particular flight approaches the assigned gate. Sensis Corporation is developing the algorithms and models required for time-of-arrival prediction as part of a NASA and Volpe National Transportation Systems Center funded project called Dynamic Runway Occupancy Measurement System (DROMS). This investigation has identified several parameters that significantly contribute to arrival time variation and uses this information in the design of neural network models which predict arrival time. In addition to predicting gate arrival times, this research considers the related challenges of predicting ramp and threshold arrival times. A ramp arrival time prediction model is of particular interest because this is where aircraft control transitions to the airlines. A simple model, predicting a constant time-to-ramp of 23.2 minutes for each aircraft crossing the TRACON boundary, achieves 95 percent compliance with bounds of plus-or-minus 11.5 minutes. The real-time models developed in this research reduce these bounds to plus-or-minus 8.3 minutes.