DeepGOMIMO: Deep Learning-Aided Generalized Optical MIMO with CSI-Free Detection

Generalized optical multiple-input multiple-output (GOMIMO) techniques have been recently shown to be promising for high-speed wireless communication (OWC) systems. In this paper, we propose a novel deep learning-aided GOMIMO (DeepGOMIMO) framework systems, where channel state information (CSI)-free blind detection can enabled by employing specially designed neural network (DNN)-based MIMO detector. The CSI-free DNN detector mainly consists of two modules: one is the pre-processing module which address both path loss and crosstalk issues caused transmission, other feed-forward used joint spatial constellation learning statistics input signal additive noise. Our simulation results clearly verify that, in typical indoor 4 $\times$ MIMO-OWC system using generalized modulation (GOSM) multiplexing (GOSMP) with unipolar non-zero 4-ary pulse amplitude (4-PAM) modulation, proposed achieves near same bit error rate (BER) performance as optimal maximum-likelihood (ML) detector, but much reduced computational complexity. Moreover, since does not require instantaneous estimation obtain accurate CSI, it enjoys unique advantages improved achievable data time delay comparison CSI-based zero-forcing (ZF-DNN)