Performance Analysis and Optimization of 3D Massive MIMO Multi-Pair Relaying with SWIPT

In this paper, we study a massive multiple-input multiple-output (mMIMO) relay network where multiple source-destination pairs exchange information through a common relay equipped with a massive antenna array. The source users perform simultaneous wireless information and power transfer (SWIPT) and the power-splitting (PS) scheme is used at the relay to first harvest energy from the received signals, and then, to transmit the decoded signals using the harvested energy. Relay performs maximum ratio combining/maximum ratio transmission (MRC/MRT) beam-forming on the received signal. Under the three-dimensional (3D) directional channel model, we derive a closed-form lower bound expression for the average signal-to-interference-plus-noise ratio using results from random matrix theory, which leads to an asymptotic approximation of the achievable sum-rate. Based on that, we study a joint optimization problem over the tilt and PS ratio to maximize the achievable sum-rate. Grid search algorithm is used to solve the non-convex problem. Simulation results verify our theoretical analysis and the efficiency of our optimized design. In particular, our optimized system outperforms a conventional system with $\pi/4$ tilt and PS of $0.5$, by at least $62\%$.