Analysis of Joint Transmit-Receive Diversity in Downlink MIMO Heterogeneous Cellular Networks

TL;DR

This paper analyzes the performance of joint transmit-receive diversity in downlink MIMO heterogeneous cellular networks using stochastic geometry, deriving coverage probabilities and insights on interference management and antenna strategies.

Contribution

It develops a tractable stochastic model for MIMO HetNets with joint transmit-receive diversity, providing new analytical insights into interference-aware and interference-blind MRC performance.

Findings

Interference-aware MRC is less effective than interference-blind MRC with transmit diversity.

Ignoring spatial interference correlation overestimates IA-MRC performance.

Selection combining can outperform MRC for a small number of receive antennas.

Abstract

We study multiple-input multiple-output (MIMO) based downlink heterogeneous cellular network (HetNets) with joint transmit-receive diversity using orthogonal space-time block codes at the base stations (BSs) and maximal-ratio combining (MRC) at the users. MIMO diversity with MRC is especially appealing in cellular networks due to the relatively low hardware complexity at both the BS and user device. Using stochastic geometry, we develop a tractable stochastic model for analyzing such HetNets taking into account the irregular and multi-tier BS deployment. We derive the coverage probability for both interference-blind (IB) and interference-aware (IA) MRC as a function of the relevant tier-specific system parameters such as BS density and power, path loss law, and number of transmit (Tx) antennas. Important insights arising from our analysis for typical HetNets are for instance: (i) IA-MRC becomes less favorable than IB-MRC with Tx diversity due to the smaller interference variance and increased interference correlation across Rx antennas; (ii) ignoring spatial interference correlation significantly overestimates the performance of IA-MRC; (iii) for small number of Rx antennas, selection combining may offer a better complexity-performance trade-off than MRC.