Estimating UAS-To-Crewed Aircraft Collision Risk Using Extreme Value Theory

Given the increasing number of uncrewed aircraft systems (UAS) operations and the need to ensure safe integration with crewed aviation, there is a critical need to have means for accurate airborne collision risk assessments. This work demonstrates such a method using Extreme Value Theory applied to simulation results. A set of simulated uncrewed aircraft (UA) trajectories are generated and compared with historical aircraft tracks to obtain a sample of closest-points-of-approach. The distribution of these samples nearer than a selected threshold are fit with a Generalized Pareto Distribution, which is then used to estimate the likelihood of approaches close enough to be tantamount to collision. Approach distances from Wing's real-life production flights are compared to simulation results to validate the model. The results suggest this technique is a valid means for tractably computing air risk for specific operations in a way that is empirically verifiable through comparison with the distribution of field data.