Intelligent based hybrid precoder for millimetre wave massive MIMO system.

The Millimeter-wave and Massive Multiple Input Multiple Output technologies are promising candidates to offer high data rates and system throughputs in the next-generation wireless communication systems. In massive MIMO systems, the precoder plays a leading role to cancel the interference between the data stream and thereby reduce the complexity of the receiver design. In the conventional fully digital precoding scheme, many Radio Frequency chains are essential for every antenna array. Therefore, a hybrid precoder is a feasible solution to reduce the RF chains and improve the antenna array gain by dividing the signal processing into analog and digital precoders. Designing a hybrid precoder is a non-convex optimization problem because the phase shifter in the analog precoder holds hardware constraints. To address this, an intelligent hybrid precoder is designed using deep learning algorithms. In this paper, the Deep Learning framework is incorporated for hybrid precoder as it renovates non-convex problems into a network training process. In this work, the hybrid precoder is based on decomposition techniques such as Uniform Channel Decomposition (UCD) and Generalized Triangular Decomposition method, which is implemented in the training process as it provides equal gain for all subchannels to diminish the inter subchannel interference. The simulation results validate the proposed work is superior in terms of the Bit Error Rate in comparison with other conventional decomposition techniques. The result shows that the deep learning-based hybrid precoder is better than the conventional with a 2 dB improvement between the UCD method and GMD method. [ABSTRACT FROM AUTHOR]