Beam Alignment for Millimeter Wave High Speed Train Communication Systems: A Bayesian Bandit Learning Approach

In this paper, we present a new approach for beam alignment in millimeter wave high speed train systems. The core idea of our approach is using Bayesian multi-armed bandit in the beam search process. In the proposed method, by approximating the signal on each beam direction as a Gaussian random variable, we update the posterior mean and variance once the beam direction is searched. With the obtained posterior mean and variance, we compute the upper and the lower confidence bounds, by which we identify the beam candidate to be searched in the next time. By selecting one whose confidence gap is large, we explore the beam direction that includes large amount of uncertainty, i.e., not much explored yet. Using this method, we also analyze the regret bound. Via simulation, we demonstrate the proposed Bayesian bandit learning approach provides better performance compared to other methods in terms of the beam alignment probability.