Skip to main content

Data-Driven Airborne Collision Risk Modelling using a Probability Density Function

Paper ID

ATM-2023-030

Conference

USA/Europe ATM R&D Seminar

Year

2023

Theme

Separation assurance and safety nets

Project Name

Keywords:

collision risk modelling, data-driven, kernel density estimation, probability density function, safety

Authors

Benoit Figuet, Raphael Monstein and Steven Barry

DOI

Project Number

Abstract

This paper introduces a novel data-driven mid-air collision risk model for an aircraft flying through a flow of aircraft, modelled using a probability density function to describe position, and a speed vector. The proposed model is, compared to traditional Monte-Carlo simulations, computationally efficient and, thus, facilitates exploration of risks as a function of key parameters, such as aircraft performance, or with different scenarios. Compared with traditional collision risk models, the proposed solution can handle more complex trajectories and traffic flows. The usefulness of the novel model is illustrated on a real-world example by applying it to the terminal airspace of Zurich airport, Switzerland. Specifically, the probability of collisions between go-arounds on Runway 14 and departures on Runway 16 is quantified. The results of the model were validated through comparison with Monte-Carlo simulations, with comparable outcomes but significantly lower computational costs.