Assignment Optimization of Reconfigurable Intelligent Surfaces

In a multi-cell architecture, each Reconfigurable Intelligent Surface (RIS) is assigned to at most one Base Station (BS) at a time. We aim to maximize the worst-user SINR by optimizing the BS beamforming vectors and/or the binary RIS–BS assignment variables. This leads to a highly non-convex, non-linear mixed-integer optimization problem. To obtain tractable solutions, we develop four schemes: a large-scale MISOCP formulation, an alternating-optimization method, a quota-based allocation scheme, and an RIS optimization with fixed beamforming. All proposed schemes achieve substantial performance gains compared to a conventional no-RIS baseline.