Localization of multiple sources with a moving array using subspace data fusion

We study a direct location estimator for the problem of calculating the positions of multiple sources from measurements made with a moving antenna array. In the first pre-processing step, subspaces are formed from the raw antenna outputs at all positions of the moving array. Then the parameters of interest are directly estimated from a cost function that results from fusing all subspaces. This Subspace Data Fusion (SDF) approach requires only a low-dimensional optimization and avoids the data association problem inherent in Bearings-only Localization (BOL) methods. In Monte Carlo simulations, we compare SDF with BOL, where the data association is solved with a Kalman filter-based tracking algorithm. We find that the SDF estimator approaches the Cramer-Rao Bound (CRB) and always performs better than the BOL method. In the case of small signal-to-noise ratio, closely spaced targets, and crossing bearings the SDF estimator considerably outperforms the BOL estimator.