5G is the latest wireless standard designed to comply with stringent performance requirements to support diverse use cases over different verticals, considered not feasible with previous cellular network technology. In addition, the rapid increase of mobile devices and the rising popularity of mobile applications introduce unprecedented demands on wireless networking infrastructure. Mobile Network Operators (MNOs) will need to collect an immense amount of data in order to monitor network performance and provide better services.
Machine Learning (ML) techniques are of main interest in predicting the UE’s mobility behavior in vehicular networks (V2X) and networks having Unmanned Aerial Vehicles (UAVs) as UEs, because they exhibit strong dynamics in terms of traffic patterns, network topologies and radio propagation. In these cases, the network can extract knowledge of the mobility patterns of its UEs, based on their contextual information and signaling messages, proceeding in optimized network operations, such as handovers, proactive caching and traffic offloading. This way the network can reduce the signaling overhead and network load, minimizing any recurrent delays.
Mobility management is a key component to guarantee successful service delivery, so the objectives of this Thesis are: i) the design of enhanced mobility management solutions in order to reduce control signaling overhead and ii) their performance evaluation in real-world scenarios. The solutions will be reviewed in terms of their capability to address the requirements of selected use cases in 5G and Beyond 5G networks.