Detection des evolutions d'un mobile et estimation de sa cinematique par une methode de tests d'hypotheses

Abstract

The proposed method deals with the problem of detecting the evolutions of a mobile in motion (aircraft) on the basis of the general theory of significance tests (tests of BAYES, WALD).

The principle is as follows : several possible aircraft flight paths are continuously considered which correspond to different evolutions of the aircraft. Each path is represented as a model in a simple manner ; a 2nd order stationary filter allows the implementation to be performed in real time : considering that the evolutions are performed only at a maximum acceleration the filter is therefore submitted to an acceleration type control, the acceleration being then regarded as a white Gaussian noise with an average value, varying according to the evolution amplitude.

The different filters operate "in parallel" thanks to the measurements provided (Cartesian or polar coordinates). Each filter gives its own estimate of the target kinematics.

The proposed algorithm selects at every moment the estimate given by the filter whose model best corresponds to reality. Taking into account some probabilities of error, such a selection is carried out following a significance test designed on the basis of pseudo-innovations peculiar to each filter (such a test makes allowances for non-stationarities of the measurement noise).