抄録
This paper investigates how a priori localization information on a stationary ground target affects the optimal trajectories of cooperative unmanned aerial vehicles (UAVs) for estimating the target location. Each UAV is assumed to be equipped with a bearing-only sensor which always directs to the target. In order to formally incorporate the effect of a priori information into the estimation process, a Bayesian framework is employed: The optimal flying directions and paths of the multiple UAVs are determined by minimizing the Bayesian Cramér-Rao Lower Bound (BCRLB) with prior information on the ground target location represented via a bivariate Gaussian Probability Density Function (PDF). The effect of the prior information is examined with varying parameters that handle the size and shape of the prior PDF. It turns out that the more elliptical the prior PDF support, the more orthogonal the flying directions to the major principal axis of the ellipse in order to reduce the information imbalance in the direction of the major principal axis. In order to reduce the computational load for the calculation of the BCRLB, we employ the Bayesian Monte-Carlo integration method, which is shown to be fast enough to compute the BCRLB online in small UAVs.
