Joint Deployment and Beamforming Design for RIS-Assisted Cell-Free Networks Under Practical Phase Shift Model
Abstract
Reconfigurable intelligent surface (RIS)-assisted cell-free massive multiple-input multiple-output (mMIMO) networks hold great potential for significantly improving wireless system capacity and energy efficiency. This paper investigates a weighted sum-rate maximization problem by jointly optimizing RIS deployment, access point active beamforming, and RIS passive reflection coefficients under the practical RIS amplitude-phase coupling model. To solve the resulting mixed discrete-continuous non-convex problem, a surrogate-assisted bi-level optimization framework is developed. Specifically, the inner layer jointly optimizes the active and passive beamforming under the practical amplitude-phase coupling constraint, while the outer layer searches for the RIS deployment based on the beamforming-optimized weighted sum-rate (WSR). To avoid repeatedly executing the computationally intensive inner-layer optimization during the deployment search, an online deep surrogate model is employed to predict the WSR of candidate deployment schemes and is progressively updated using selected exact inner-layer evaluations. The resulting deployment search is implemented through a discrete particle swarm optimization (DPSO) algorithm. Extensive simulations validate that the surrogate-assisted approach significantly accelerates the discrete deployment search without sacrificing convergence accuracy, yielding a highly efficient and topology-aware RIS placement strategy. Furthermore, compared to traditional algorithms relying on ideal assumptions, the proposed practical design explicitly accounts for RIS non-idealities, thereby avoiding the performance degradation caused by model mismatch.