Spectrum sensing optimization in uncalibrated massive antennas systems

Abstract

In this paper, we propose a spectrum sensing (SS) design for massive MIMO (M-MIMO) antennas system based on decision threshold optimization given a fixed subset of antennas and sample size able to minimize the total error probability subject to target detection and false alarm probabilities. This formulation results in a nonlinear optimization problem (NLP). Empirical expressions based on the central limit theorem (CLT) with estimated sample-mean parameter values are deployed to find the optimal decision threshold. Two operational modes are considered: (i) spectral-efficient (SE) mode, i.e., low number of samples combined with a high number of antennas, implying in a reduced sensing time and increased transmission rate; (ii) energy-efficient (EE) mode, meaning reduced number of sensing antenna and massive number of processing samples, while the remaining antennas are in sleep mode, saving energy. To solve the NLP, an algorithm based on the sequential quadratic programming (SQP) is proposed, being able to attain convergence in a few iterations. Moreover, to deal with uncalibrated antennas, more elaborated detectors are evoked, including the Hadamard detector (HD), volume detector (VD) and the covariance detector (CAV). Such detector parameters are optimized in the sense of maximizing the SS task subject to uncalibrated massive antennas effects.