Linear Transmission of Composite Gaussian Measurements over a Fading Channel under Delay Constraints

TL;DR

This paper investigates optimal linear transmission strategies for composite Gaussian measurements over fading channels with delay constraints, revealing how delay and channel knowledge influence performance and strategy optimality.

Contribution

It characterizes optimal linear transmission schemes under various delay constraints and channel state information scenarios, providing new insights into power allocation and scheme structure.

Findings

Optimal schemes are derived for known CSI with strict delay constraints.

Relaxing delay constraints improves distortion performance.

One-to-one linear mappings are suboptimal in multi-measurement scenarios.

Abstract

Delay constrained linear transmission (LT) strategies are considered for the transmission of composite Gaussian measurements over an additive white Gaussian noise fading channel under an average power constraint. If the channel state information (CSI) is known by both the encoder and decoder, the optimal LT scheme in terms of the average mean-square error distortion is characterized under a strict delay constraint, and a graphical interpretation of the optimal power allocation strategy is presented. Then, for general delay constraints, two LT strategies are proposed based on the solution to a particular multiple measurements-parallel channels scenario. It is shown that the distortion decreases as the delay constraint is relaxed, and when the delay constraint is completely removed, both strategies achieve the optimal performance under certain matching conditions. If the CSI is known only by the decoder, the optimal LT strategy is derived under a strict delay constraint. The extension for general delay constraints is shown to be hard. As a first step towards understanding the structure of the optimal scheme in this case, it is shown that for the multiple measurements-parallel channels scenario, any LT scheme that uses only a one-to-one linear mapping between measurements and channels is suboptimal in general.