Data-Driven Modeling of Aircraft Midair Separation Violation

Significant research has developed approaches to establish separation criteria between aircraft. Current daily planning methods do not account for deviations between planned and actual flight trajectories. This paper proposes a method to predict the number of loss of separation events between aircraft for a proposed daily schedule while including the uncertainty regarding flight paths and altitudes. The methodology uses publicly available flight data to develop GPR models to stochastically predict the planned level, flight speed, and portion of flight in cruise for each flight. The flight data are also used to develop libraries of deviations from planned flight paths and flight levels, which can be used to stochastically simulate the deviation for any flight. The statistics of violations within an ARTCC for a given daily flight schedule are then predicted using Monte Carlo runs. The proposed model is validated using actual flight data. The approach is used to assess the impact of increasing traffic by adding synthetic traffic to a daily flight schedule for a given sector. The results quantify how the frequency of mid-air separation violations might increase with increasing flight traffic, thus supporting risk-informed decision-making on two fronts: current flight scheduling and future traffic growth.