Robust PAPR Reduction in Large-Scale MIMO-OFDM using Three-Operator ADMM-type Techniques

This paper deals with a distortion-based non-convex peak-to-average power ratio (PAPR) problem for large-scale multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems. Our work is motivated by the observation that the distortion stemming from the PAPR reduction schemes has a deleterious impact on the data rates of MIMO-OFDM systems. Recently, some approaches have been proposed to either null or mitigate such distortion seen at the receiver(s) side by exploiting the extra degrees of freedom when the downlink channel is perfectly known at the transmitter. Unfortunately, most of these proposed methods are not robust against channel uncertainty, since perfect channel knowledge is practically infeasible at the transmitter. Although some recent works utilize semidefinite programming to cope with channel uncertainty and non-convex PAPR problem, they have formidable computational complexity. Additionally, some prior-art techniques tackle the non-convex PAPR problem by minimizing the peak power, which renders a suboptimal solution. In this work, we showcase the application of powerful first-order optimization schemes, namely the three-operator alternating direction method of multipliers (ADMM)-type techniques, notably 1) three-operator ADMM, 2) Bregman ADMM, and 3) Davis-Yin splitting, to solve the non-convex and robust PAPR problem, yielding a near-optimal solution in a computationally efficient manner.