Learning Based Proactive Handovers in Heterogeneous Networks

Today, the number of versatile real-time mobile applications is vast, each requiring different data rate, Quality of Service (QoS) and connection availability requirements. There have been strong demands for pervasive communication with advances in wireless technologies. Real-time applications experience significant performance bottlenecks in heterogeneous networks. A critical time for a real-time application is when a vertical handover is done between different radio access technologies. It requires a lot of signalling causing unwanted interruptions to real-time applications. This work presents a utilization of learning algorithms to give time for applications to prepare itself for vertical handovers in the heterogeneous network environment. A testbed has been implemented, which collects PHY (Physical layer), application level QoS and users context information from a terminal and combines these Key Performance Indicators (KPI) with network planning information in order to anticipate vertical handovers by taking into account the preparation time required by a specific real-time application.