Interference minimization in visible light communication systems using LSTM-based predictive resource allocation

Abstract Visible light communication (VLC) has emerged as a promising technology for high-speed wireless access in indoor and vehicular environments. However, dense deployment of light-emitting diodes (LEDs) leads to severe co-channel interference, which degrades signal quality and system throughput. This paper proposes a Long Short-Term Memory (LSTM)-based interference prediction and mitigation framework for multiuser VLC networks. The temporal correlation of user mobility and channel variations is exploited using LSTM to forecast future interference levels and dynamically allocate transmit power and bandwidth. A mathematical model of the VLC channel, interference, and signal-to-interference-plus-noise ratio (SINR) is developed, and the optimization problem is formulated to minimize aggregate interference while satisfying quality-of-service constraints. Simulation results demonstrate that the proposed LSTM-based scheme significantly improves SINR, reduces bit error rate, and enhances throughput compared to conventional static and heuristic allocation methods. At a transmit power of 1 W, the throughput under interference is 114 Mbps, while the interference-free benchmark achieves 131 Mbps. The proposed LSTM-assisted framework restores throughput to 128 Mbps, closely approaching ideal conditions. These results confirm the effectiveness of the proposed predictive interference mitigation strategy for next-generation VLC systems.