Recurrent neural network (RNN) for delay-tolerant repetition-coded (RC) indoor optical wireless communication systems

Area of Science:

• Optical Wireless Communications
• Machine Learning in Communications

Background:

• Indoor optical wireless communications are crucial for high-speed user connectivity.
• Repetition-coded (RC) multiple transmitters enhance robustness and capacity but require precise signal synchronization.
• Synchronization is a significant challenge in high-speed wireless systems.

Purpose of the Study:

• To propose and demonstrate a delay-tolerant RC indoor optical wireless communication system.
• To overcome synchronization challenges in high-speed optical wireless communications.
• To enhance system robustness and capacity through a novel signal processing technique.

Main Methods:

• Development of a recurrent neural network (RNN)-based symbol decision scheme.
• Implementation of the RNN for delay-tolerant RC indoor optical wireless communication.
• Experimental validation of the proposed RNN scheme.

Main Results:

• The proposed RNN scheme significantly improves the bit-error-rate (BER) by approximately one order of magnitude.
• Performance improvement increases with longer signal delays.
• The RNN outperforms previously studied fully connected neural network schemes.

Conclusions:

• RNN-based symbol decision is an effective method for delay-tolerant RC indoor optical wireless communication.
• This approach addresses critical synchronization challenges in high-speed systems.
• The RNN offers a superior solution compared to traditional neural network methods for this application.