AI-optimized power allocation for TDMA in visible light communications: a multi-stage grid search approach

Abstract This paper describes a new approach to light emitting diode (LED) deployment in a visible light communication (VLC) network for improving the capacity of the communication system through intelligent use of power allocation schemes within VLC networks. The utilization of this method has shown to yield much greater data throughput than current VLC methodologies and by utilizing the existing light emission capabilities from LED devices, it has reduced dependency on additional resources such as fiber, antennas, etc. In addition, the ability to automate the power allocation using an optimization-enhanced time-division multiple access (TDMA) approach allows the autonomous deployment of this invention, removing the need for manual input of the optimal LED locations and also reduces the chance of incorrect power allocation occurring, thus maximizing the use of the VLC system full potential. The core contribution of this work is a BER-driven power allocation strategy for TDMA-based VLC systems under discrete 4-PAM modulation, utilizing a practical multi-stage grid search rather than advancing artificial intelligence methodology. The method demonstrates the expected advantage over capacity-maximizing water-filling for discrete 4-PAM modulation, providing a practical BER-optimal allocation strategy.