Energy Efficient Estimation of Gaussian Sources Over Inhomogeneous Gaussian MAC Channels

TL;DR

This paper investigates energy-efficient methods for estimating Gaussian sources over Gaussian MAC channels, establishing conditions for optimality, analyzing power minimization, and comparing joint, separate, and uncoded schemes in various network scenarios.

Contribution

It introduces a joint source-channel coding approach with necessary and sufficient conditions, and analyzes power minimization and optimal decoding order in Gaussian MAC networks.

Findings

Uncoded transmission remains optimal in homogeneous networks at high SNR.

Optimal decoding order depends only on channel quality rankings.

Joint and uncoded schemes share the same asymptotic SNR exponents.

Abstract

It has been shown lately the optimality of uncoded transmission in estimating Gaussian sources over homogeneous/symmetric Gaussian multiple access channels (MAC) using multiple sensors. It remains, however, unclear whether it still holds for any arbitrary networks and/or with high channel signal-to-noise ratio (SNR) and high signal-to-measurement-noise ratio (SMNR). In this paper, we first provide a joint source and channel coding approach in estimating Gaussian sources over Gaussian MAC channels, as well as its sufficient and necessary condition in restoring Gaussian sources with a prescribed distortion value. An interesting relationship between our proposed joint approach with a more straightforward separate source and channel coding scheme is then established. We then formulate constrained power minimization problems and transform them to relaxed convex geometric programming problems, whose numerical results exhibit that either separate or uncoded scheme becomes dominant over a linear topology network. In addition, we prove that the optimal decoding order to minimize the total transmission powers for both source and channel coding parts is solely subject to the ranking of MAC channel qualities, and has nothing to do with the ranking of measurement qualities. Finally, asymptotic results for homogeneous networks are obtained which not only confirm the existing optimality of the uncoded approach, but also show that the asymptotic SNR exponents of these three approaches are all the same. Moreover, the proposed joint approach share the same asymptotic ratio with respect to high SNR and high SMNR as the uncoded scheme.