Low Overhead Weighted-Graph-Coloring-Based Two-Layer Precoding for FDD Massive MIMO Systems

TL;DR

This paper introduces a low-overhead, graph-coloring-based two-layer precoding scheme for FDD massive MIMO systems that reduces channel state information acquisition overhead by intelligently orthogonalizing user clusters based on their angular-spreading-range overlaps.

Contribution

It proposes a novel heuristic edge-weighted vertex-coloring pattern division scheme to efficiently mitigate angular-spreading-range overlaps in practical massive MIMO scenarios.

Findings

Outperforms existing pattern division schemes in simulations

Effectively mitigates angular-spreading-range overlaps

Reduces overhead in downlink precoding for FDD massive MIMO

Abstract

A massive multiple-input multiple-output (MIMO) system, operating in Frequency Division Duplexing (FDD) mode of operation, suffers from prohibitively high overhead associated with downlink channel state information (CSI) acquisition and downlink precoding, due to the lack of uplink/downlink channel reciprocity. In this paper, a heuristic edge-weighted vertex-coloring based pattern division (EWVC-PD) scheme is proposed to alleviate the overhead of a two-layer precoding approach, in a practical scenario where the user clusters undergo serious angular-spreading-range (ASR) overlapping. Specifically, under a constraint of limited number of subchannels, an undirected edge-weighted graph (EWG) is firstly constructed, to depict the potential ASR overlapping relationship among clusters. Then, inspired by classical graph coloring algorithms, we develop the EWVC-PD scheme which mitigate the ASR by subchannel orthogonalization between clusters possessing serious ASR overlapping, and multiplexing the ones having slight ASR overlapping. Simulation results reveal that our scheme efficiently outperforms the existing pattern division schemes.