Design and Implementation of Real-Time Optimal Power Allocation System with Neural Network in OFDM-Based Channel of Optical Wireless Communications

In recent years, many studies have been conducted on OFDM-based optical wireless communications to develop a 6G communication infrastructure to improve data transmission and reduce the BER. Real-time optimal power management can enhance the data transmission speed and received power in an optical wireless channel under various conditions. This paper discusses implementing a real-time optimal power allocation system using a neural network for OFDM-based optical wireless communications. The system is designed to manage transmitter power, enhancing data transmission rates in optical wireless channels. In system design, data concerning power allocation for various types of OFDM-based optical wireless channels are calculated analytically, including the BER, SNR, fog effects, and fading types in the channel model. Next, a DNN neural model is trained using data generated from the analytical method. The trained model is finally integrated into wireless optical communication transmitter hardware. The experimental results indicate that the embedded power allocation system processes power allocation quickly. The proposed system achieves an average accuracy of 98% in power allocation, surpassing the analytical method. When used in wireless optical communication transmitters, this embedded system enhances speed and accuracy in power management, optimizing the data transmission rate up to 16 Gbps for a 500 m channel.