Leveraging Common User Clustering for Improved Performance in Cell-Free NOMA Networks

Abstract

Abstract In existing cell-free non-orthogonal multiple access (NOMA) networks, each user equipment (UE) receives desired signals from a group of access points (APs) simultaneously, however, the UE belongs to only a single NOMA cluster. This prevents exploiting the potential benefits of having clusters with common UEs. Previous studies have investigated clusters with common UE in cellular NOMA networks by considering that UEs at the cell borders can communicate with base stations through multiple clusters using coordinate multi-point techniques. However, this approach has not been investigated for cell-free NOMA networks yet.In this paper, we consider the UE clustering in a cell-free NOMA network considering three strategies, including single-UE OMA (CF-OMA), double-UE NOMA with no common UE (CF-NOMA), and double-UE NOMA with common UE (CF-NOMAC) clusters. We analytically prove that the proposed CF-NOMAC clustering improves the sum rate compared to the CF-NOMA, and CF-OMA methods, provided that certain conditions hold. Considering the three characterized UE clustering methods, we formally define UE clustering, AP grouping, and power allocation problem for maximizing the sum rate in cell-free NOMA networks.We decompose the formulated mixed-integer non-linear program (MINLP) problem into subproblems with integers and continuous decision variables. The integer decision variables correspond to UE clustering and AP grouping which are obtained based on the analytically elaborated results, and the continuous ones correspond to the power values that are obtained by converting a non-convex signomial programming problem into geometrical programming using the monomial approximation technique.Analytical results reveal the outperformance of the proposed algorithm due to employing a lower number of orthogonal clusters in the network.