Application Driven Joint Uplink-Downlink Optimization in Wireless Communications

TL;DR

This paper develops a mathematical framework to optimize joint uplink and downlink rates in multi-user wireless systems, considering imperfect CSI, feedback schemes, and user velocity, to improve throughput.

Contribution

It introduces a novel joint uplink-downlink optimization framework incorporating realistic channel estimation and feedback models for wireless communications.

Findings

Optimal pilot densities depend on user velocity.

CSI feedback is more beneficial at low velocities with high downlink priority.

At high velocities, no CSI feedback and TDD with random beamforming are optimal.

Abstract

This paper introduces a new mathematical framework, which is used to derive joint uplink/downlink achievable rate regions for multi-user spatial multiplexing between one base station and multiple terminals. The framework consists of two models: the first one is a simple transmission model for uplink and downlink, which is capable to give a lower bound on the capacity for the case that the transmission is subject to imperfect CSI. A detailed model for concrete channel estimation and feedback schemes provides parameter input to the former model and covers the most important aspects such as pilot design optimization, linear channel estimation, feedback delay, and feedback quantization. We apply this framework to determine optimal pilot densities and CSI feedback quantity, given that a weighted sum of uplink and downlink throughput is to be maximized for a certain user velocity. We show that for low speed, and if downlink throughput is of particular importance, a significant portion of the uplink should be invested into CSI feedback. At higher velocity, however, downlink performance becomes mainly affected by CSI feedback delay, and hence CSI feedback brings little gain considering the inherent sacrifice of uplink capacity. We further show that for high velocities, it becomes beneficial to use no CSI feedback at all, but apply random beamforming in the downlink and operate in time-division duplex.