Tracking quality monitoring based on information geometry and geodesic shooting

This paper explores a new way of monitoring tracking quality based on information geometry methods. Previous methods were not fully satisfying when facing certain maneuvers when both the bias and the variance vary at the same time. The method we propose takes into account all parameters of the movement of the target thanks to the use of an appropriate distance which reflects the mean and the covariance of the distribution we obtain with the Kalman filter. To achieve that, we develop the distance in the manifold of multivariate Gaussians, compute the Fisher-Rao distance, and compare it with bounds.