Random Fourier Features based Post-Distortion for Massive-MIMO Visible Light Communication

Abstract

Massive multiple input multiple output (m-MIMO) visible light communication (VLC) has emerged as a viable technology to enhance the throughput of an existing VLC based systems in order to serve the high speed data requirement for beyond 5G and 6G communication systems. However, performance of m-MIMO VLC is severely impaired by (1) columns of m-MIMO VLC channel are highly correlated which makes the channel matrix ill-conditioned, and (2) nonlinear transfer-characteristics of light emitting diode (LED). The aforementioned channel impairments collectively degrade the overall bit error rate (BER) at the receiver. In this paper, finite memory budget adaptive precoder is proposed to decrease the channel correlation/condition number of MIMO channel matrix. Reproducing kernel Hilbert space (RKHS) based post-distorters have been proposed in the literature for VLC which rely on growing dictionary of kernel evaluations, and hence it is difficult to practically implement them under finite memory budget. In this paper, random Fourier features (RFF) based kernel least mean square (RFF-KLMS) algorithm is proposed for post-distortion over m-MIMO VLC channels which alleviates the requirement of dictionary and facilitates post-distortion under finite memory budget. Computer simulations performed over m-MIMO VLC channels show that the proposed RFF based algorithm exhibit similar BER performance with reduced computational complexity as compared to the dictionary based post-distortion algorithm. © 2020 IEEE