28th International Conference on Information Fusion, FUSION 2025, Rio de Janeiro, Brezilya, 7 - 11 Temmuz 2025, (Tam Metin Bildiri)
This paper presents a new state-space model and tracking filter for Riemannian angle tracking on the unit sphere. We consider a scenario where noisy angle-only measurements (azimuth and elevation) of a target, moving with nearly constant velocity, are obtained by a stationary sensor. The angular state, composed of the angular position vector constrained on the unit sphere and the angular velocity vector defined over the tangent plane at the angular position, is tracked. Since the target's nearly constant velocity motion in 3D space induces correlated acceleratory motion on the unit sphere, we develop a state-space model that accounts for this acceleration. A tracking filter based on the Riemannian generalization of the unscented Kalman filter (UKF) given in the literature is proposed and implemented using this model. Simulation results demonstrate that the developed angle tracker outperforms two benchmark Riemannian filters, one based on a (Riemannian generalization of) nearly constant velocity model and the other on a (Riemannian generalization of) nearly constant acceleration model, particularly for close-range and/or fast-moving targets.