Joint Beamforming and Segmentation Optimization for Segmented RIS-Assisted Cell-Free mMIMO Systems Under Perfect and Imperfect CSI

Abstract

The integration of reconfigurable intelligent surface (RIS) technology with cell-free massive MIMO (CF mMIMO) enhances wireless network sum-rate performance. This paper proposes a novel segmented RIS-assisted CF mMIMO architecture for downlink transmission, where RIS elements within each segment share a common reflection coefficient to reduce optimization complexity. A weighted sum-rate maximization problem is formulated by jointly optimizing AP beamforming, segment-level RIS coefficients, and RIS element-to-segment association under perfect and imperfect CSI. For perfect CSI, the problem is reformulated via fractional programming and decomposed into subproblems solved by FP-CVX, FP-SCA, and swap matching. For imperfect CSI, an Minimum Mean Square Error–Discrete Fourier Transform (MMSE–DFT)-based channel estimation method is adopted with the same alternating optimization framework. A Joint Beamforming and Segmentation Parameter Optimization (JBSPO) algorithm is developed for the resulting mixed-integer non-convex problem. The framework is extended to finite backhaul capacity, phase-dependent amplitude response, discrete phase shifts, and spatially correlated channels. To capture the impact of spatial correlation on RIS segmentation, a new element-level grouping suitability indicator is introduced. Simulation results show that the proposed PCS-RIS achieves a performance–complexity trade-off, outperforming fixed segmentation, random phase shift, and no-RIS schemes. It improves WSR by about 1.21%, 1.41%, and 1.58% over PCFS-RIS, random, and no-RIS, respectively, with only 0.97% gap to ideal RIS, while reducing runtime by up to 48.98%.