Path Selection‐Based Beamforming Design for RIS‐Aided mmWave Massive MIMO Systems

ABSTRACT

This paper investigates beamforming design for reconfigurable intelligent surface (RIS)‐aided millimetre‐wave (mmWave) massive MIMO systems. Building on the sparse channel characteristics and array‐response‐based modelling commonly used in mmWave RIS systems, we adopt a path‐domain channel representation to enable low‐complexity beamforming design. Under large‐scale antenna and RIS configurations, the effective cascaded channel can be represented by a limited number of dominant propagation paths, which motivates a structured two‐stage beamforming framework. The key idea is to treat dominant path selection as a reduced‐subspace construction problem for spectral‐efficiency‐oriented transmission, rather than a per‐path ranking problem. Based on this idea, we propose a two‐stage framework termed OMP‐RCG. In the first stage, orthogonal matching pursuit (OMP) is used to iteratively construct RIS‐side reduced‐dimensional subspaces on both the BS–RIS and RIS–UE links by tracking the dominant residual channel components. In the second stage, RIS phase shifts are optimized within the selected reduced‐dimensional subspace via a Riemannian conjugate gradient (RCG) algorithm on the complex circle manifold. Simulation results show that the proposed OMP‐RCG framework outperforms existing path‐selection‐based schemes and achieves similar spectral efficiency to that of benchmark.