Robustness considerations for bearings-only tracking

This paper deals with the analysis of robustness for bearings-only tracking. We focus on the case where the trajectory of the target follows some assumed parametric model with an unknown additive perturbation. We investigate the parametric estimator obtained using least squares. We prove that the difference between this estimator and the true parameter may be expanded along directions with intensities that may be very large due to the presence of the unknown perturbation. For a given parametric trajectory, we are able to exhibit the perturbation that leads to the highest intensities. We show through simulations that other methods, like extended Kalman filter or particles filter, present also the same lack of robustness.