An Efficient and Low-complexity Beam Training Technique for mmWave Communication

An efficient and low-complexity codebook-based beam training technique is proposed for mmWave communication systems. To identify the pair of beam patterns maximizing the received signal power with a given beam resolution, the proposed technique formulates the beam training process as a combinational optimization problem. It considers all of the beam pairs as a set of feasible solutions and the received signal power as the objective function. The main idea is to implement the Nelder-Mead simplex algorithm iteratively and recursively, each time finding the best beam pair with beam patterns of increased resolution. This technique benefits from the efficiency of the Nelder-Mead algorithm whilst avoiding its drawbacks, such as sticking in a local optimum near the optimum solution. Simulation results show that, for short-range indoor communications with line-of-sight (LoS) multipath components, the proposed technique achieves the same beam selection compared to the exhaustive beam training algorithm while requiring around 28% of the search steps needed by the two-level training mechanism adopted in IEEE 802.15.3c when 32 antenna elements are adopted at both the transmitter and the receiver.