Distributed Massive MIMO System with Dynamic Clustering in LEO Satellite Networks

TL;DR

This work tackles scalable downlink performance in LEO satellite networks by introducing a distributed massive MIMO architecture with dynamic, user-centric clustering (UC-DMIMO). It combines RSAP-based coordination and phase shift–aware precoding to compensate propagation delays, enabling coherent joint transmission within serving clusters. The paper develops algorithms for initial access, pilot assignment, and cluster handover, and compares UC-DMIMO against non-cooperative and full-cooperative baselines, showing near FC-DMIMO spectral efficiency with significantly smaller cluster sizes and reduced backhaul complexity. Simulation results demonstrate that phase compensation markedly enhances performance and that the proposed approach offers a scalable path to high-throughput, low-latency broadband in dynamic LEO constellations.

Abstract

Distributed massive multiple-input multiple output (mMIMO) system for low earth orbit (LEO) satellite networks is introduced as a promising technique to provide broadband connectivity. Nevertheless, several challenges persist in implementing distributed mMIMO systems for LEO satellite networks. These challenges include providing scalable massive access implementation as the system complexity increases with network size. Another challenging issue is the asynchronous arrival of signals at the user terminals due to the different propagation delays among distributed antennas in space, which destroys the coherent transmission, and consequently degrades the system performance. In this paper, we propose a scalable distributed mMIMO system for LEO satellite networks based on dynamic user-centric clustering. Aiming to obtain scalable implementation, new algorithms for initial cooperative access, cluster selection, and cluster handover are provided. In addition, phase shift-aware precoding is implemented to compensate for the propagation delay phase shifts. The performance of the proposed user-centric distributed mMIMO is compared with two baseline configurations: the non-cooperative transmission systems, where each user connects to only a single satellite, and the full-cooperative distributed mMIMO systems, where all satellites contribute serving each user. The numerical results show the potential of the proposed distributed mMIMO system to enhance system spectral efficiency when compared to noncooperative transmission systems. Additionally, it demonstrates the ability to minimize the serving cluster size for each user, thereby reducing the overall system complexity in comparison to the full-cooperative distributed mMIMO systems.

Figures (5)

• Figure 1: UC-DMIMO system in space. a) general system with a GEO satellite working as a CPU and a set of orbiting LEO satellites working as access points that jointly serve ground users; and b) the $n$-th SC of LEO satellites selected based on a user-centric clustering approach and cooperatively work as a distributed antenna system to serve the $n$-th user.
• Figure 2: Downlink per-user spectral efficiency for a) $M=100$ and b) $M=400$, when the RSAP is selected based on the maximum channel gain. Note that in this paper, we mainly focus on investigating phase synchronization.
• Figure 3: Comparison of the downlink per-user spectral efficiency for the proposed UC-DMIMO system with the FC-DMIMO and the NCT when the RSAP is selected based on the best channel and based on the maximum service time.
• Figure 4: User's SC size for the proposed US-DMIMO compared with that of FC-DMIMO when the RSAP is selected based on the best channel.
• Figure 5: Per-user coverage time without RSAP handover when the RSAP is selected based on the best channel and based on the maximum service time.