A Non-Coherent Distributed Massive Multi Antenna System Under Rician Propagation

Abstract

Non-coherent massive multi-antenna (mMA) communications have demonstrated performance comparable to coherent communications and they are also more suitable for certain use cases. Despite mMA's advantages, practical deployment limitations arise as the number of antennas increases. These limitations can be overcome using distributed (D-)mMA. However, most previous work on non-coherent mMA has provided solutions assuming simplified Rayleigh channel models, while these scenarios are predominantly characterized by Rician propagation channels. This paper provides a comprehensive analysis of the impact of Rician channels in D-mMA and offers light machine learning (ML)-based solutions to mitigate this impact. Analytical and numerical results demonstrate that the proposed methods outperform previously studied solutions in both the Rician and Rayleigh cases. This allows for constellations twice the size of the previous state-of-the-art proposals. These results are an important milestone in generalizing non-coherent mMA to more demanding and realistic channel conditions.