https://enac.hal.science/hal-01004152Delahaye, DanielDanielDelahayeENAC - Ecole Nationale de l'Aviation CivilePuechmorel, StéphaneStéphanePuechmorelENAC - Ecole Nationale de l'Aviation CivileBoussouf, LoïcLoïcBoussoufTrajectory prediction : a functional regression approachHAL CCSD2008[MATH.MATH-OC] Mathematics [math]/Optimization and Control [math.OC]Porte, Laurence2014-07-01 17:13:022021-10-19 11:02:492014-07-01 17:13:02enConference papers1To solve the problem of trajectory prediction, traditional approaches can be classified into three categories : learning algorithms, nonparametric algorithms, simulation algorithms. The approach of this paper is to use the functional regression for the trajectory prediction, ie a method between learning algorithms and nonparametric methods. The proposed approach consists in the optimal decomposition of trajectories on a functional Karhunen Loève base and to learn the weights with a large set of registered trajectories. Those weights are also functions. Based on this optimal decomposition, generalization process enables to predict the shape of the trajectory in the near future. This method has been applied on real aircraft trajectories with successful results.