Abstract
In fifth generation (5G) wireless systems, millimetre wave (mmWave) frequency bands will be exploited to support high data demands. The exploitation of mmWave carrier frequencies allows the use of large-scale antenna arrays for three-directional beamforming. To utilise the beamforming gain, the transmitter and the receiver are required to perform beam training periodically to maintain reliable connectivity. This brings an inevitable training overhead for beam measurements. In this paper, an efficient beam training strategy is proposed and the trade-off between the data throughput and the training overhead is investigated for dynamic mmWave channels including blockage effects. By utilising the sparse nature of mmWave channels and the spatial consistency feature of realistic channels, only a limited number of beams are stored and searched based on the recent history of beam selection results. Simulation results show that the proposed beam training strategy can significantly reduce the training overhead while achieving comparable data throughput to the exhaustive beam search.