Machine learning based analog beam selection for 5G mmWave small cell networks

In 5G millimeter-wave (mmWave) small cell networks, a lot of mmWave small cell base stations (SBSs) are densely deployed, where SBSs uses large number of antennas to form directional analog beams for improving the spatial spectrum reuse. Every mobile terminals (MTs) could be served by concurrent transmissions from multiple SBSs simultaneously. However, as the large number increase of SBSs and MTs, it becomes very difficult for each SBS to precisely and quickly select the analog beams. Hence, in this paper, we utilize machine learning (ML) to enhance the network performance. First, we model the random distribution of SBSs by Poisson point process, where the probabilities that multiple SBSs serve the MT are derived to get the average sum rate (ASR) for mmWave small cell networks. Second, based on ML, we iteratively use the support vector machine (SVM) classifier to select the analog beam of SBS. Third, we proposed an iteration sequential minimal optimization (SMO) training algorithm to train data samples of all the links, where the computational complexity and algorithm convergence are also discussed. Last, the sample training and simulation are evaluated by Google TensorFlow. The results verified that our proposed algorithm not only gets a higher ASR than the traditional channel estimation (CE) algorithm, but also achieves a very substantial reduction of calculation complexity.