Beamforming Arrays with Faulty Sensors in Dynamic Environments

摘要

This paper addresses the problem of beamforming a uniform linear array when some of the receive elements are not operational. Bad sensors are often handled by either zeroing or interpolating the faulty elements prior to conventional beamforming. While zeroing faulty elements prior to conventional beamforming is the simplest approach, it often results in undesirably high beamformer sidelobes. Alternatively, minimum mean-square error (MMSE) interpolation of the missing data is not explicitly aimed at minimizing post- interpolation leakage of strong interference components into otherwise quiet directions. While true minimum variance (MV) adaptive beamforming is the optimal solution given long observation times, for large arrays in highly dynamic environments, severely limited snapshot support poses a difficult trade- off between desired interference suppression versus unwanted signal cancellation. In this work, we propose an alternative approach for conventional beamforming with faulty sensors based on adaptively synthesizing complete array data snapshots which minimize post-interpolation leakage into quiet directions subject to a constraint that the solution is sufficiently close to the data measured at the working elements. After reconstruction of the complete array data snapshots, computationally efficient conventional beamforming can be performed to both estimate the noise field directionality as well as produce time-series output for further temporal analysis.