Predictive Beamforming With Active Inference in Hierarchical Codebooks

Beamforming technology using massive MIMO in the millimeter wave (mmWave) band is attracting attention as a fundamental technology for next-generation wireless communication systems. Beamforming increases the signal-to-noise ratio (SNR) of signals received by terminals and enables high-speed communications. In beamforming, it is necessary to search for a beam with appropriate directivity from a predefined codebook and irradiate the beam toward the terminal. Although hierarchical codebooks can be used to reduce the search overhead, conventional beam training methods in hierarchical codebooks are not suitable under conditions where channel conditions change over time. This is because each time the beam is re-searched, a non-optimal beam is applied, and the SNR is repeatedly degraded temporarily and significantly. To solve this problem, this paper proposes a method to predict the optimal beam using active inference. This method avoids the problem of temporarily degrading SNR by predicting the optimal beam without searching for it. As a result, the method using active inference can increase the average SNR compared to the conventional beam training method in hierarchical codebooks.