Machine-Learning-Based Indoor Localization under Shadowing Condition for P-NOMA VLC Systems.

The localization of agents for collaborative tasks is crucial to maintain the quality of the communication link for successful data transmission between the base station and agents. Power-domain Non-Orthogonal Multiple Access (P-NOMA) is an emerging multiplexing technique that enables the base station to accumulate signals for different agents using the same time-frequency channel. The environment information such as distance from the base station is required at the base station to calculate communication channel gains and allocate suitable signal power to each agent. The accurate estimate of the position for power allocation of P-NOMA in a dynamic environment is challenging due to the changing location of the end-agent and shadowing. In this paper, we take advantage of the two-way Visible Light Communication (VLC) link to (1) estimate the position of the end-agent in a real-time indoor environment based on the signal power received at the base station using machine learning algorithms and (2) allocate resources using the Simplified Gain Ratio Power Allocation (S-GRPA) scheme with the look-up table method. In addition, we use the Euclidean Distance Matrix (EDM) to estimate the location of the end-agent whose signal was lost due to shadowing. The simulation results show that the machine learning algorithm is able to provide an accuracy of 0.19 m and allocate power to the agent.

Deep Learning-Based Next-Generation Waveform for Multiuser VLC Systems.

Due to the growing number of users, power, and spectral effectiveness, most communication systems are complex and difficult to implement on a large scale. Artificial Intelligence (AI) has played an outstanding role in the implementation of theoretical systems in the real world, with less complexity achieving better results. In this direction, we compare the Non-Orthogonal Multiple Access (NOMA) technique for a multiuser Visible Light Communication (VLC) system with Successive Interference Cancellation (SIC) for two types of detectors: (1) the deep learning-based system and (2) the traditional maximum likelihood (ML) decoder-based system. For multiplexing, we compare the variations of novel Orbital Angular Momentum (OAM) multiplexing and Orthogonal Frequency Division Multiplexing (OFDM) with Index Modulation (IM). In this article, we implement OFDM-IM and OAM-IM for four users for the Gaussian fading MIMO Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS) VLC channels. The suggested systems' bit error rate (BER) performances are compared in simulations for a wide range of Signal-to-Noise Ratios (SNRs), which shows that deep learning-based systems outperform the ML-based system for both users to ensure better decoding at the receiver end, especially at higher SNR values. The detection error is lower in a deep learning-based system at around 20% and around 30% for low SNR and high SNR values, respectively.