Bidirectional Branch and Bound Based Antenna Selection in Massive MIMO Systems

Antenna selection is a low cost solution to massive MIMO systems. There are two well-known design criteria of antenna selection in MIMO systems, i.e., maximizing MIMO capacity and maximizing post-processed Signal to Noise Ratio (SNR). This paper focuses on the latter one that can be achieved by selecting the largest Minimum Singular Value (MSV) of channel submatrices. A novel antenna selection is proposed by using the bidirectional Branch And Bound (BAB) searching algorithm to find the globally optimal channel submatrix with largest NISV. Simulation results demonstrate that, with both independent and identically distributed (i.i.d.) and sparse channels, the proposed method not only can achieve the same Bit Error Rate (BER) as the exhaustive search, but also has much lower complexity than the exhaustive search. Although the bidirectional BAB based antenna selection still has a high complexity, it can serve as a benchmark purpose for the future low complexity antenna selection design, especially when the exhaustive search is infeasible in massive NIIMO systems, which is the main motivation of this paper.