Ultra-reliable MU-MIMO detector based on deep learning for 5G/B5G-enabled IoT

Abstract In this paper, we propose an ultra-reliable multiuser multiple-input multiple-output (MU-MIMO) detector based on deep learning for the fifth-generation and beyond the fifth-generation (5G/B5G) enabled Internet of Things (IoT), where the system is operating in interfering environments correlated over the time or frequency domain. For this system, we employ an iterative detection framework of a conventional symbol-by-symbol detector and a deep convolutional neural network (DCNN), where the DCNN is used to suppress the interfering signals by capturing the characteristics through deep learning. The conventional detector in the framework can be either ZF-MLD or MMSE-MLD, where the conventional zero-forcing (ZF) or minimum mean square error (MMSE) is initially used and then the near-by signal candidates are searched through the maximum likelihood detection (MLD). Thus, the proposed MU-MIMO detector can suppress the influence of the correlated interferences with low computational complexity and finally improve the reliability of the practical MU-MIMO systems in the presence of correlated interferences. To further enhance the system detection performance, user scheduling is employed, where several user selection criteria are proposed to choose one best user among multiple ones. Simulation results are finally presented to show that an ultra-reliable detection performance can be achieved for the 5G/B5G-enabled IoT.