Abstract

This paper describes and evaluates an Innovative Algorithm for Improved Quality Multipath Delivery of Virtual Reality Content (QM4VR) that addresses the stringent communication requirements of Virtual Reality (VR) applications. Making use of the Multipath TCP (MPTCP) built-in multipath delivery features (subflows), QM4VR explores the subflows’ characteristics, evaluates their performance (e.g., delay, throughput or loss) and proposes a new management scheme to improve the Quality of Service (QOS) of the VR applications. glsqm4vr adopts a Machine Learning (ML)-based approach to evaluate the subflows’ performance which is implemented in two steps: 1) a linear regression scheme to forecast the subflow’s performance for a given feature; and 2) a linear classification scheme to arrange the results obtained in step 1. Based on these results QM4VR selects the most appropriate subflows for data delivery in order to achieve improvement of VR QOS levels.