AIR TRAFFIC COMPLEXITY THROUGH LOCAL COVARIANCE IN THE CONTEXT OF LARGE AREAS OF OPERATIONS
Résumé
The notion of air traffic complexity has many facets and can be related to workload, which is a perception of a given situation by a human controller or to disorder, which is intrinsic. The present work falls within the second category and aims at computing the level of organization in a neighborhood of a point on the earth. It is based on a two steps approach: in the first one, a smooth time-dependent vector field is inferred from the sampled traffic using local linear models. Since positions are measured on a sphere, some special care must be taken as it is not a vector space. Using the Levi-Civita connection and its associated parallel transport, a local linear model can be defined in the tangent space at any point. The time evolution is captured through the kernel function that take the form of a product with one term being time dependent. In a second phase, the underlying dynamical is characterized at each point using a symmetric positive definite matrix. Thanks to the Riemannian manifold structure of the set of such matrices, a complexity indicator is then defined.
Origine : Fichiers produits par l'(les) auteur(s)