Abstract:
Keywords- LiFi, WiFi, O-OFDM, PAPR, DP-OFDM, GMSK, DFT, O-GFDM, HLWN, load balancing, RL, RWP, ORWP. A significant increase in wireless communication has been observed over the past decade. The existing radio frequency (RF) based communication network is not able to cope up with this influx in connections and data requirements. In order to meet future needs, researchers have started investigating Light Fidelity (LiFi) for the indoor environment. LiFi offers various advantages over RF, such as a vast spectrum, spatial reuse, and inherent security. Furthermore, LiFi does not interfere with the devices operating in the RF spectrum. However, LiFi technology has its limitations; the major challenges of the LiFi system include the non-linearity due to LiFi front end, limited front-end bandwidth, and susceptibility to blockages. In this dissertation, we have tried to address these aforementioned challenges. Firstly, we propose an adaptive learning architecture (ALA)-based predistoter to mitigate the effect of front-end non-linearity. The proposed ALA predistoter achieved near-linear performance in terms of amplitude-amplitude (AM/AM) distortions and constellation plots for different LiFi front-ends non-linearity. Secondly, in order to support high data rates with limited front-end bandwidth, highly spectral efficient modulation schemes such as optical orthogonal frequency division multiplexing (O-OFDM) are required. Nonetheless, the major drawback of O-OFDM is that it suffers from a high peak-to-average power ratio (PAPR), which causes clipping distortion, reduces the illumination-to-communication conversion efficiency, and affects the lifetime of the LED. Therefore, in this thesis, we propose advanced spectrally efficient low PAPR modulation schemes such as double precoded optical orthogonal frequency division multiplexing (DPOOFDM) and optical-generalized frequency division multiplexing (O-GFDM). The simulation results validate that the proposed DP-OOFDM with interleaved subcarrier mapping provides PAPR as low as 2.1 dB compared to 12.7 dB for the corresponding O-OFDM counterpart. Lastly, in order to deal with the problem of blockages in LiFi, the coexistence of LiFi and WiFi has been proposed in the literature. However, an appropriate load balancing strategy plays a vital role in the overall performance of such heterogeneous LiFi WiFi networks (HLWN). Nonetheless, the problem of load balancing of HLWN is a non-convex mixed-integer nonlinear programming (MINLP) optimization problem, i.e., it is mathematically intractable. Therefore, in this thesis, we propose a reinforcement learning (RL) based load balancing technique for HLWN. Additionally, we also explore the effect of different mobility models and link aggregation in HLWN. Simulation results illustrate that the proposed RL-based method can ensure near-optimal performance at relatively low complexity. The proposed frameworks in this dissertation can be utilized in LiFi standards. It will be helpful for LiFi communication engineers to design an efficient physical layer and intelligent load balancing scheme for HLWN without performing extensive simulations.