An Efficient Handover Trigger Scheme for Vehicular Networks Using Recurrent Neural Networks

The future of intelligent transportation systems has become increasingly dependent on the integration of heterogeneous wireless technologies over connected vehicular networks. In order to provide efficient safety, traffic control management, and assistance to drivers. Managing the transition and migration of active communication session of vehicles between different point of access is essential for seamless mobility. However, the rapid mobility of vehicles creates a challenging problem toward the efficiency of wireless communication between vehicles and access routers. To address this issue, an accurate mobility management protocol is needed, which anticipate the vehicles movement and network quality in order to derive a handover decision. In this paper, we present an efficient neural network-based handover trigger scheme for vehicular networks to accurately predict the handover trigger time using time-series quality measurements of the network. We adopt a recurrent neural network model to predict the upcoming sequence of received signal quality to derive a handover trigger estimation. In the performance evaluation, the proposed time-series estimation method shows high accuracy rates compared to several machine learning methods over generated mobility traces.