Beamspace Direct Localization for Large-Scale Antenna Array Systems

Abstract

Direct localization systems can outperform two-step localization ones by jointly processing all the measurements observed at base stations. However, such systems encounter large communication and computational load when the signal dimension is high, e.g., with large-scale antenna arrays. In this paper, we establish a general framework for beamspace direct localization, which consists of beamspace design and position determination. By deriving the performance bound of beamspace direct localization, we first prove that the high-dimensional array signal can be represented in a low-dimensional beamspace without information loss, and then cast the beamspace design as an optimization problem under dimension constraints. In the presence of parameter uncertainty, we propose a robust beamspace scheme to guarantee the performance in the worst case. Using the low-dimensional signals, we develop an efficient direct localization algorithm with the computational complexity and communication overhead orders of magnitude smaller than the conventional methods. Simulation results show that the proposed approach attains near-optimal localization performance.