Computation-Aided Adaptive Codebook Design for Millimeter Wave Massive MIMO

Different from the existing predefined hierarchical codebook before the beam training, we design a computation-aided adaptive codebook and propose a beam training algorithm based on it. At each layer of the hierarchical codebook, we first estimate the channel angle of arrival (AOA) or angle of departure (AOD) according to the beam training results from the previous layers and then adaptively design a codeword in the current layer to align with the estimated AOA or AOD. Benefiting from the computation resources used for the adaptive codebook design and beam alignment, the proposed algorithm can improve the success rate of beam training as well as reducing the training overhead comparing with the existing algorithms. Simulation results verify the effectiveness of the proposed algorithm.