Quality driven resource allocation for adaptive video streaming in OFDMA uplink

In this paper, we consider the problem of improving Quality of Experience (QoE) of multiple video streams in OFDMA uplink. The proposed system leverages both video adaptation and adaptation of resource allocation (RA). Three different time scales arises in this scenario, namely (1) long time scale of QoE, evaluated over the sequence of Groups of Pictures (GOPs), (2) medium time scale of video adaptation for each GOP, and (3) short time scale of RA inside each GOP. We deal with the time scale differences through a two step process. We first formulate, for each GOP, an optimization problem of video quality which takes into account the long time scale QoE. Then we convert this problem to a sequence of quality-driven RA optimization. At each RA step inside a GOP, resources are allocated to users according to their utility determined by QoE constraints and quality fairness. The performance of proposed quality-driven RA is evaluated under realistic uplink scenario. It is shown that the proposed system improves significantly users' QoE compared to other solutions.

I. Introduction

Nowadays, mobile video traffic is one of the main contributors to the exponential traffic growth in cellular networks [1]. Hence, a big challenge for future networks is, first, to respectively increase the throughput, and, second, to provide efficient quality managements for this specific type of traffic. Although many solutions such as small cells, HetN ets and massive MIMO have been proposed to increase user throughput, only a few have been specifically tailored for the video delivery with high Quality of Experience (OoE).