Cognitive antenna selection for optimal source localization

Performance of array processing algorithms is directly determined by the number of array elements. Processing of data from a large number of receive array elements involves a large number of receivers, resulting in high cost and complex implementations. In many practical applications, a large number of receiving antenna elements is available, however, the number of receivers is limited due to their high cost. In this work, we address the problem of direction-of-arrival (DOA) estimation with a large number of antenna array elements and a small number of receivers, where the receivers are connected to the array elements via a reconfigurable switching matrix. A cognitive approach, named cognitive antenna selection (CASE), for sequentially switching the elements of the array based on previous observations is proposed. The criterion for antenna selection is based on minimization of the conditional Bobrovski-Zakai bound (BZB) given history observations on the mean-squared-error (MSE) of the DOA estimate. The performance of the CASE algorithm is evaluated via simulations and compared to two other non-adaptive approaches. It is shown that the CASE algorithm outperforms the other compared algorithms in terms of MSE both asymptotically and in the threshold region.