Deep learning-based transceiver design for multi-user MIMO systems

Abstract

Listen

Translate article

Multi-user multiple-input multiple-output (MIMO) is a key technique to increase both the channel capacity and the number of users that can be served simultaneously. One of the main challenges related to the deployment of such systems is the complexity of the transceiver processing. Although the conventional optimization algorithms are able to provide excellent performance, they generally require considerable computational complexity, which gets in the way of their practical application in real-time systems. In contrast to existing work, we study a DL-based transceiver design scheme for a downlink MIMO broadcasting channel (MIMO BC) system, which consists of a base station (BS) serving multi-users. The objective of this work is to maximize the sum-rate of all users by jointly optimizing the transmitter and receivers under the total power constraint, while suppressing interference as much as possible. Due to the inter-user interference in such system, the considered problem is nonconvex and NP-hard. Different from traditional optimization algorithms, we rely on the convolutional neural networks (CNNs) to optimize the transceivers in an adaptive way. In the proposed scheme, we develop an unsupervised learning strategy, where a loss function is constructed innovatively for reducing the inter-user interference. Simulation results show that the inter-user interference is reduced effectively by our proposed CNN-based transceiver optimization method.