Feedback Design for Multi-Antenna K-tier Heterogeneous Downlink Cellular Networks

TL;DR

This paper analyzes and optimizes limited feedback strategies for multi-antenna K-tier heterogeneous cellular networks, improving spectral efficiency through analytical modeling and feedback partitioning in both cooperative and non-cooperative scenarios.

Contribution

It provides analytical expressions for spectral efficiency and introduces feedback partition solutions tailored for cooperative and non-cooperative multi-antenna heterogeneous networks.

Findings

Feedback is useful only when interference is low in a tier.

Allocating more feedback to stronger intra-cluster BSs is effective.

Feedback partition solutions are independent of instantaneous SIR.

Abstract

We characterize the ergodic spectral efficiency of a non-cooperative and a cooperative type of K-tier heterogeneous networks with limited feedback. In the non-cooperative case, a multi-antenna base station (BS) serves a single-antenna user using maximum-ratio transmission based on limited feedback. In the cooperative case, a BS coordination set is formed by using dynamic clustering across the tiers, wherein the intra-cluster interference is mitigated by using multi-cell zero-forcing also based on limited feedback. Modeling the network based on stochastic geometry, we derive analytical expressions for the ergodic spectral efficiency as a function of the system parameters. Leveraging the obtained expressions, we formulate feedback partition problems and obtain solutions to improve the ergodic spectral efficiency. Simulations show the spectral efficiency improvement by using the obtained feedback partitions. Our major findings are as follows: 1) In the non-cooperative case, the feedback is only useful in a particular tier if the mean interference is small enough. 2) In the cooperative case, allocating more feedback to stronger intra-cluster BSs is efficient. 3) In both cases, the obtained solutions do not change depending on instantaneous signal-to-interference ratio.