Smart Concurrent Learning Scheme for 5G Network: QoS-Aware Radio Resource Allocation

Abstract

The continuous performance race led wireless industry to a ubiquitous adoption of heterogeneous architecture with small cells. Although, the extreme densification offers the largest increase in the network capacity, it also challenges more valuable system metrics like quality of service (QoS) for users with various traffic types. While previously used to boost capacity in the cellular system, radio resource management schemes now need to be refocused to address the requirements of the next generation services. In this paper, we propose a novel power profile construction framework designed specifically for scenarios with multiple traffic types and a smart way to adopt it to a distributed learning algorithm. The main goal is to provide each cell with the ability to make its decision autonomously while taking into account the QoS metrics of the surrounding cells. We illustrate the application of this distributed learning strategy for the subband allocation and propose several mechanisms to improve the convergence speed in the absence of communication. To address the arising convergence challenge we propose to additionally enhance the proposed algorithm with a smart model fitting stage. Taking advantage of this ideas, we were able to properly utilize flexibility and meet the strict requirements of machine learning algorithm for QoS scenarios. The performances of the proposed method are evaluated in the case of Long Term Evolution (LTE-A) setup and compared to a number of traditional resource allocation schemes. System level simulations show that it achieves a considerable improvement in system performance for heterogeneous deployment, without compromising the quality of service of the overall system.