Importance sampling for probabilistic prognosis of sector-wide flight separation safety

This study develops a probabilistic simulation methodology for the en-route safety assessment of multiple aircraft flying within an airspace sector. This assessment is affected by multiple sources of uncertainty, such as external inputs such as wind gust, internal factors such as aircraft component failure, and model errors (e.g., in the aircraft dynamics model) and variability in observed trajectories. The concern is to evaluate whether any pair of aircraft fail to maintain a minimum required lateral separation, and the safety assessment problem involves the prognosis of separation violations between all possible aircraft pairs within the sector, over a duration of interest. Some of the uncertainty sources are modeled as random variables and others as random processes. We combine two deterministic trajectory prediction models – the kinematics-based Generalized National Airspace Trajectory-Prediction System (GNATS), and the dynamics-based Base of Aircraft Data (BADA) – to formulate a probabilistic aircraft dynamic model which also includes model errors as additional external inputs. A computationally efficient importance sampling-based scheme is developed for this safety assessment problem involving a large number of aircraft. This approach is demonstrated using historical flight data, for aircraft operating in the Houston air traffic sector.