A New Search Process for High-Accuracy Beam Training in MmWave Communication

Abstract

In millimeter wave communication with large antenna arrays, beam training is an effective method to find the optimal directional beamforming vector in pre-designed training codebook. Based on the truth that with more training pilots, the optimal beam can be identified more precisely, we proposed a new search process that adaptively allocates pilots to the beams and finds the best one. Specifically, the proposed process evaluates the remaining beams in search space and calculates the posterior probability, those beams that are obviously not the optimal one will be rejected, then more pilots will be allocated to the rest beams and further evaluations are to be made until the algorithm finds the optimal result. When using the parameters similar to a high-accuracy beam training algorithm, our proposed search process achieves lower misalignment probability compared with exhaustive linear search and original high-accuracy algorithm. Moreover, if some knowledge of channel state information can be obtained, it may achieve better performance in practical system.