Content-aware and QoE Optimization of Video Stream Scheduling over LTE Networks using Genetic Algorithm and Random Neural Networks

Abstract

Long Term Evolution (LTE) networks support Quality of Service (QoS) of multimedia services with fast communication connectivity, high data transfer speed and high level of security. Video streaming over LTE networks is one of the highest proportions of global mobile data traffic and is growing; this has led to the development of several scheduling algorithms aimed at improving the performance of these networks. The performance analysis and evaluation of existing scheduling algorithms are generally limited to QoS parameters. It is not clear how these scheduling algorithms perform in terms of Quality of Experience (QoE) which is the overall acceptability of a service or application, as perceived subjectively by end users. Video content has a major impact on QoE; thus its analysis in scheduling algorithms performance is critical. The aim of this study is to classify video content based on the impact of video content on quality over LTE networks. This classification is then used to develop novel QoE-aware optimization scheduling of video traffic in order to achieve maximum QoE. Our approach focuses on the development of optimization downlink scheduling based on a novel integration between random neural networks (RNN) and genetic algorithms (GA) to learn complex non-linear mapping of QoE and to search for the optimal parameters, respectively. An open source simulation tool for LTE networks (LTE-Sim) has been used to collect unique RNN training database based on existing scheduling algorithms. A comparison between the proposed scheduler and state-of-the-art LTE downlink scheduling algorithms (FLS, EXP-rule, and LOG-rule) has been made under different network conditions. Simulation results showed an increase in performance of about 15% in terms of QoE and throughput while maintaining fairness.