MACHINE LEARNING BASED CAPACITY ENHANCEMENT OF FEMTOCELLS FOR 5G HETEROGENEOUS NETWORKS

Abstract

Small cells based ultradense heterogeneous networks (HetNets) are being considered as the one the promising solution for increased coverage and capacity in the 5G cellular networks. However, in the multi-tiered architecture, co-tier and cross tier interference are a performance-limiting factor. The interference can be effectively handled through efficient resource allocation techniques in either a cooperative or distributive manner. However, the complexity of such resource allocation schemes linearly increases with the density of the HetNets due to unplanned deployment and dynamic behavior of small cells. The HetNets can be implemented only through an adaptive and self-organizing algorithm that can adapt to the dynamic conditions. In this research paper, a machine learning (ML) based adaptive resource allocation scheme is proposed for the femtocell based dense HetNets. The Q-Learning based scheme consider each femtocell base station (FBS) as the agent of the network and model the HetNets as multi-agent network to allocate optimal power to the FBS to maximize the capacity of the femtocell user equipment (FUEs) an macrocell user equipment (MUEs) while considering the quality of service (QoS) requirements. The proposed cooperative Q-Learning scheme increases the sum capacity of the FUEs by seven-folds and always ensures the minimum QoS requirements as compared to the prior work. Furthermore, the proposed solution also increased the number of supported femtocells by two-fold in comparison to the state of the art solution.