[Paper Review] Performance of Media-based Modulation in Multi-user Networks

Merve Yüzgeçcioğlu, Eduard A. Jorswieck|arXiv (Cornell University)|Jan 30, 2018

Advanced Wireless Communication Technologies9 references3 citations

TL;DR

This paper investigates media-based modulation (MBM) in multi-user networks, showing that spectral efficiency remains high despite correlation in constellation points. By modeling the sum rate region of multi-user MAC systems, it derives tight bounds on achievable rates and proves that correlation's impact diminishes as constellation size increases, approaching AWGN capacity asymptotically.

ABSTRACT

High spectral efficiency and low power consumption are the most challenging requirements of 5G networks since the number of devices are increased drastically. Media-based modulation (MBM) is a promising scheme in order to achieve these requirements. In this paper, the spectral efficiency of MBM scheme in single- and multi-user networks is studied. To find the spectral efficiency, information theoretical methods are used and tight lower- and upper-bounds are derived for the achievable rate. The performance of the system is studied for both correlated and uncorrelated constellation diagrams. It is shown that the entries of the constellation diagram for the sum rate region of multi-user case are always correlated. The characteristic of covariance matrix is derived and its impact on performance assessed.

Motivation & Objective

• To analyze the spectral efficiency of media-based modulation (MBM) in single- and multi-user wireless networks.
• To investigate the impact of correlation in constellation diagram entries on achievable rate performance.
• To derive tight lower- and upper-bounds on the achievable rate for both single-user and multi-user (MAC) scenarios.
• To characterize the covariance structure of the joint constellation diagram in multi-user systems and assess its effect on sum rate.
• To evaluate how the system performance approaches the ergodic capacity region as constellation size M increases.

Proposed method

• Uses information-theoretic methods to derive tight lower- and upper-bounds on the achievable rate for MBM in single- and multi-user fading channels.
• Models the MBM system as a Gaussian mixture channel where constellation points are generated via channel perturbations, with M = 2^NP possible points.
• Derives the covariance matrix Σ for the joint constellation diagram in a K-user MAC, showing it has KM^{K-1} largest eigenvalue and K(M−1) eigenvalues of M^{K−1}, with the rest zero.
• Applies the replica method to approximate the achievable rate region and validates bounds via Monte Carlo simulations.
• Compares the MBM performance to the ergodic capacity region of the AWGN MAC and evaluates the gap under varying SNR and constellation sizes.
• Uses a layered MIMO-MBM framework and recursive decoding to reduce complexity, enabling practical implementation.

Experimental results

Research questions

• RQ1 How does correlation in the constellation diagram entries affect the achievable spectral efficiency in single-user MBM systems?
• RQ2 What is the structure of the covariance matrix for the joint constellation diagram in a K-user multi-access channel (MAC)?
• RQ3 How does the achievable rate region of MBM in a MAC system compare to the ergodic capacity region of an AWGN MAC?
• RQ4 Does the performance gap between MBM and AWGN capacity vanish as the constellation size M increases?
• RQ5 What is the impact of increasing the number of users K and modulation points M on the null space and eigenvalue distribution of the covariance matrix?

Key findings

• For a single-user system with M = 256 and NP = 8, the rate degradation due to strong correlation in constellation points is only 0.1 bits/s/Hz.
• The achievable rate region of a 2-user MAC with NP = 3 perturbations is close to the AWGN capacity region, with a small gap even at 20 dB SNR.
• The covariance matrix of the joint constellation diagram in a K-user MAC has KM^{K−1} as the largest eigenvalue and M^{K−1} as K(M−1) eigenvalues, with the rest being zero.
• For K = 3 and M = 3, 20 out of 27 eigenvalues are zero, indicating a large null space that grows with K and M.
• As M → ∞, the non-zero eigenvalues dominate and the effect of correlation vanishes, leading to asymptotic capacity-approaching performance.
• The sum rate gap between MBM and AWGN capacity is wider than individual user rate gaps, due to inherent correlation in the joint constellation diagram.