Energy Efficiency Optimization in Integrated Satellite-Terrestrial UAV-Enabled Cell-Free Massive MIMO

TL;DR

This work addresses downlink energy efficiency in an integrated satellite-UAV CF-mMIMO network, where UAVs act as terrestrial CF-mMIMO APs coordinated by a LEO satellite. A successive convex approximation ($SCA$) algorithm is developed to maximize the UAV-layer $EE$ under per-UAV power budgets and QoS constraints, leveraging maximum-ratio ($MR$) precoding with statistical CSI at users. The authors derive closed-form expressions for the downlink $EE$ and validate them with simulations, showing that deploying tens of UAVs with optimized power substantially improves spectral efficiency and area coverage, especially when satellite and UAVs cooperate in a CF-mMIMO fashion. The results highlight the practical impact of joint NTN and TN operation for energy-efficient, high-capacity non-terrestrial networks, and point to extensions involving mobile-edge computation and caching.

Abstract

Integrating cell-free massive MIMO (CF-mMIMO) into satellite-unmanned aerial vehicle (UAV) networks offers an effective solution for enhancing connectivity. In this setup, UAVs serve as access points (APs) of a terrestrial CF-mMIMO network extending the satellite network capabilities, thereby ensuring robust, high-quality communication links. In this work, we propose a successive convex approximation algorithm for maximizing the downlink energy efficiency (EE) at the UAVs under per-UAV power budget and user quality-of-service constraints. We derive a closed-form expression for the EE that accounts for maximum-ratio transmission and statistical channel knowledge at the users. Simulation results show the effectiveness of the proposed algorithm in maximizing the EE at the UAV layer. Moreover, we observe that a few tens of UAVs transmitting with a fine-tuned power are sufficient to empower the service of satellite networks and significantly increase the spectral efficiency.

Figures (3)

• Figure 1: CDF of the per-GU SE for $L\!=\!60$, $K\!=\!40$ and $\mathrm{SE}^{\min}_k \!=\! 0.2$ [b/s/Hz], $\forall k$, under different $P^{\textrm{SN}}_{\mathrm{dl}}$ values and network operations. Results are obtained via Monte Carlo simulations. Label "CF" refers to the results obtained by implementing the derived closed-form expressions for the SE.
• Figure 2: Average EE at the UAV layer achieved by the NTN & TN system against the number of UAVs. We consider both FPA and EEM PA resulting from Algorithm \ref{['alg:EEM']}. Here, $K=40$ GUs and $P^{\textrm{SN}}_{\mathrm{dl}} = 10$ W.
• Figure 3: Average EE at the UAV layer against the number of GUs, under different network operations. We consider both EPA and EEM PA resulting from Algorithm \ref{['alg:EEM']}. Here, $L=60$ UAVs and $P^{\textrm{SN}}_{\mathrm{dl}} = 10$ W.