Extreme Learning Machine Based Channel Estimator and Equalizer for Underground Mining VLC Systems

Visible light communication (VLC) systems have numerous applications in harsh environments, such as underground mining. However, the irregular tunnel features, the random orientation of the optical components, physical phenomena such as shadowing and scattering, as well as the inherent generation of non-linearities of the optical signal greatly affect the VLC link. Therefore, it is necessary to investigate novel adaptive mitigation techniques to improve the performance of underground mining VLC systems. The literature has initially presented channel estimation schemes knowing a prior the channel state information (CSI). Currently, these schemes have evolved favorably given the research done regarding the use of machine learning techniques to estimate and equalize the communication channel. In this paper, channel estimation and equalization based on the standard extreme learning machine (ELM) scheme applied to underground mining VLC systems is presented. To verify the performance of the proposed technique, typical channel estimation schemes such as least square (LS), minimum mean square error (MMSE), and spectral temporal averaging (STA) were implemented using zero forcing (ZF) as equalizer. Furthermore, the ideal VLC channel and perfect channel estimation with matched filter (MF) and ZF are considered as reference cases. The performance analysis is validated through Monte Carlo simulations in terms of the bit error rate (BER). Results demonstrate that the novel ELM scheme performs better than traditional techniques and performs similarly compared to the perfect channel estimation case.