Parallel Solution for Per-Antenna Power Constrained Symbol-Level MU-MISO Precoding

Abstract

This paper designs a parallel solution framework for constructive interference based symbol-level precoding (CI-SLP) in the downlink of a multi-user multiple-input single-output (MU-MISO) system. Most existing works on SLP have considered the sum-power constraint, while in practical systems each transmit antenna is equipped with its dedicated power amplifier. Therefore, it is more realistic to design SLP approaches that incorporate the per-antenna power constraint (PAPC). In this paper, we focus on two specific PAPC-based problems: the constructive interference per-antenna power constraint signal to interference plus noise ratio (SINR) balancing (CI-PASB) problem and the constructive interference per-antenna peak power minimization (CI-PAPM) problem. Similar to sum-power constraint, for the CI-PASB problem, we demonstrate that it is separable, allowing the existing parallel proximal Jacobian alternating direction method of multipliers (PJ-ADMM) algorithm to be directly used. As for the CI-PAPM problem, although it is unseparable, we can leverage the established duality to obtain its solution based on the solution of the corresponding CI-PASB problem. Numerical results verify our proposed parallel methods and show that they are more efficient than the existing centralized schemes, which showcases the advantages of parallel computing and promotes the implementation of CI precoding under practical PAPC scenarios.