Factor Graph-based Message Passing Technique for Distributed Resource Allocation in 5G Networks

Next generation 5G cellular networks are intended to provide higher data rates, excellent user coverage, low latency, and low power consumption. Since 5G is expected to have a multitier architecture consisting of macro cells, small cells and Device-to-Device communication altogether but the reported works so far have considered either one of these but not all in resource allocation (RA) in heterogeneous interference model. Moreover, heterogeneous systems using centralized schemes are not scalable with respect to different constraints including interference. We propose a distributed scheme for RA in Orthogonal Frequency Division Multiple Access (OFDMA) based multi-tier cellular network for maximizing the user's data rate by keeping all kinds of interference constraints into mind. The formulated problem is an NP-hard non-convex non-linear optimization problem. The proposed algorithm solves RA problem through a distributed message passing approach which uses factor graph to model the decision making nodes. We prove the convergence of the proposed algorithm and analyze its time complexity. Through simulation experiments, we show the effect of the different parameter on RA and compared it with the existing work.

I. Introduction

The next-generation 5G wireless network is expected to have a multi-tier heterogeneous deployment of a large number of small cells along with Device-to-Device (D2D) users [1]. The presence of small cell user equipment (SUE), D2D user equipment (DUE) along with macrocell user equipment (MUE) give a significant improvement over the spectrum utilization, higher data rate, better energy efficiency in the networks. However, this heterogeneity creates the challenge in term of interference management and resource allocation (RA) due to the discrepancy in capacity, coverage area and working power [2]. There could be different interference at the different tier of the network such as interference within small cells, macro cells, macro cell and small cell, DUE and macro cell, DUE and small cell DUE. Thus, due to such heterogeneity in the network, the centralized solution to RA problem along with interference management becomes computationally expensive and give rise to computational overhead because of which there is a need to come up with a distributed RA scheme.