Green MEC-Enabled Cell-Free Massive MIMO: Co-Design of Power Allocation, Computation Offloading, and NOMA User Clustering

Abstract

Future wireless networks demand extremely low latency and high energy efficiency to support emerging applications in industrial automation and real-time communications. This paper proposes a green mobile edge computing (MEC) enabled cell-free massive MIMO (CF-mMIMO) framework that co-designs power allocation, computation offloading, and non-orthogonal multiple access (NOMA) user clustering. Instead of conventional heuristic pairing, NOMA user groups are formed via a normalized K-means algorithm that exploits both large-scale fading and users’ geographical distribution, enabling scalable and topology-aware clustering. On top of this clustering stage, we formulate a joint optimization problem that minimizes the worst-case end-to-end latency while incorporating transmission and local-computing energy consumption through a tunable penalty term, thereby enabling flexible latency-energy trade-offs. We develop a computationally efficient two-phase algorithm that combines K-means clustering with successive convex approximation to tackle the ensuing non-convex problem. Comprehensive simulations show that the suggested system reduces the median latency by 14.8% compared with the traditional mMIMO architecture, achieves convergence within 4–5 iterations, and exhibits near-linear scalability with approximately $1.13~\mu $ s additional latency per extra user. Under severe channel estimation errors, the performance degrades by nearly 6.7% while preserving a clear advantage over baseline schemes. These results demonstrate that the proposed co-design offers a viable and robust green solution for next-generation CF-mMIMO networks with MEC and NOMA.