Reinforcement learning-based power amplifier aware Peak-to-Average power ratio reduction in optical OTSM systems

Visible Light Communication (VLC) has emerged as a promising technology for high-speed and energy-efficient wireless communication. However, Optical Orthogonal Time Sequence Multiplexing (OTSM) systems suffer from high Peak-to-Average Power Ratio (PAPR), which causes power amplifier (PA) nonlinearity, reduced energy efficiency, and degraded transmission performance. This paper proposes a Reinforcement Learning-Based Power Amplifier Aware (RL-PAA) PAPR reduction framework for Optical OTSM systems. The proposed method formulates the PAPR reduction problem as a Markov Decision Process, where a reinforcement learning agent dynamically selects peak reduction actions according to instantaneous waveform characteristics and PA operating conditions. By jointly considering PAPR, PA efficiency, and bit error rate (BER), the proposed framework learns an optimal transmission strategy that mitigates nonlinear distortion while enhancing communication reliability. Simulation results demonstrate that the proposed RL-PAA scheme achieves a PAPR of 5.9 dB at a CCDF of 10 −4 , outperforming conventional OTSM, clipping, tone reservation, and partial transmit sequence techniques. Furthermore, the proposed approach attains a BER of 10 −4 at only 13 dB SNR and improves PA efficiency to 65 % at 12 dB input back-off. The reinforcement learning agent exhibits rapid convergence and stable performance, confirming the effectiveness of RL-PAA for next-generation Optical OTSM-based VLC systems.