Machine learning based dimming control of orthogonal frequency division multiplexing based light fidelity

Abstract

This thesis proposes a new hybrid orthogonal frequency division multiplexing (OFDM) modulation technique termed as DC-biased Pulse amplitude modulated Optical OFDM (DPO-OFDM) for light fidelity (LiFi) systems. The proposed DPO-OFDM scheme is the combination of two existing OFDM formats: DC-biased Optical OFDM (DCO-OFDM) and Pulse Amplitude Modulated discrete multitone (PAM-DMT). In DPO-OFDM, the odd index subcarriers carry DCO-OFDM, and the imaginary part of even index subcarriers carry PAM-DMT component. Analysis indicates that the required DC-bias for DPO-OFDM is a function of the dimming and root mean square of its two components' values. Simulation results show that for an uncoded bit error rate (BER) of 10-3, the dimming range for DPO-OFDM is 3% to 97% with acceptable spectral efficiency. Furthermore, a switching algorithm for HDAP-OFDM is proposed where the individual components of HDAP-OFDM are switched according to a target dimming level. Next, machine learning algorithms are used to find the appropriate proportions of DPO-OFDM components. For this, a dataset is created for DPO-OFDM system simulated using the MATLAB tool. Using Python programming language, it is shown here that polynomial regression of degree 4 can reliable predict the constellation size of DCO-OFDM component of DPO-OFDM for a given constellation size of PAM-DMT and a given target dimming level. Hence, the findings of this thesis can contribute in developing a practical LiFi system capable of reliable Internet connection and room illumination.