Hypergraph-Based SCMA Codebook Allocation in User-Centric Ultra-Dense Networks with Machine Learning

Abstract

In this paper, the optimum design of the downlink sparse code multiple access (SCMA) based user-centric ultra-dense networks (UUDNs) is studied, in which lots of access points (APs) are deployed to provide service to many user equipments (UEs). In UUDNs, the network architecture is shifted from traditional cell-centric to user-centric, where many APs can serve for one single UE, with the density of APs higher than that of UEs. One main challenge faced by UUDN design is the large interference due to the dense AP/UE deployments. The objective of this paper is to find the optimum SCMA codebooks allocation scheme to minimize the network interference that can maximize the system throughput, subject to the quality of service (QoS) constraints for each UE. The design is formulated as a mixed-integer nonlinear program (MINLP) problem by using different codebooks allocation among different APs serving the same UE, and it is a well known NP-hard problem. To tackle this problem, we use a weighted hypergraph model to transform it into a clustering problem, where machine learning (ML) algorithms are proposed to solve this MINLP problem efficiently. Simulation results show that the proposed hypergraph based ML algorithm outperforms the existing algorithm.