Rician κ-factor-aware min-max latency optimization for uplink NOMA-assisted cell-free massive MIMO-MEC under imperfect CSI

This paper investigates min-max latency minimization for uplink non-orthogonal multiple access (NOMA)-assisted cell-free massive multiple-input multiple-output (CF-mMIMO) systems integrated with mobile edge computing (MEC), where all channels follow Rician fading and channel state information (CSI) is imperfect. Three analytical contributions characterize how the Rician κ-factor shapes system behavior: (i) closed-form expressions for the fourth-order channel moment and power variance factor V(κ), (ii) a rigorous uplink SINR expression under the additive CSI error model with maximum ratio combining (MRC) that subsumes perfect CSI and Rayleigh fading as special cases, and (iii) a SINR variability characterization that rigorously links the Rician κ-factor to the fading-induced channel power fluctuation. A min-max latency optimization over transmit power and partial offloading ratios is formulated. A channel-correlation-based greedy NOMA pairing strategy exploits the LoS-induced spatial structure, and a successive convex approximation (SCA) algorithm with first-order (Karush–Kuhn–Tucker, KKT) stationarity at a converged fixed point under standard SCA regularity conditions efficiently solves the resulting non-convex problem. Monte Carlo simulations over κ=0 (Rayleigh) and κ ∈ {3, 6, 10, 15} dB confirm that the LoS component reduces latency variability by up to 13% relative to Rayleigh fading while providing consistent median latency reductions, with gains saturating near κ ≈ 10 dB in agreement with the analytical saturation characterization. The proposed framework achieves more than 28% latency reduction over co-located massive MIMO, converges within 4–5 SCA iterations, and exhibits frequency-dependent graceful degradation under imperfect CSI; comparison against a fully sequential SIC baseline further confirms that correlation-aware pairwise pairing, not raw cancellation count, governs min-max latency in URLLC CF-mMIMO-MEC deployments.