Learning-Assisted Probabilistic Precoding for PAPR Reduction in OTSM With Different Subcarrier Configurations

Orthogonal Time Sequence Multiplexing (OTSM) has emerged as a promising waveform candidate for sixth-generation (6G) wireless systems due to its robustness against high mobility, Doppler dispersion, and time-varying channel conditions, while maintaining low receiver complexity. However, similar to other multicarrier-derived waveforms, OTSM suffers from a high peak-to-average power ratio (PAPR), which degrades power amplifier efficiency and system reliability. This letter proposes a novel Learning-Assisted Probabilistic Precoding (LA-PP) framework to address the PAPR issue in OTSM systems. The proposed LA-PP exploits learned OTSM-specific peak-sensitive signal statistics to guide probabilistic precoding, distinguishing it from conventional PAPR reduction techniques such as selective mapping (SLM), partial transmit sequences (PTS), and optimization-based approaches. Consequently, LA-PP avoids exhaustive search procedures and the need for side information. An OTSM-aware peak cost function is formulated to capture the sequency-domain phase alignment effects responsible for peak formation. Simulation results demonstrate that LA-PP achieves significant PAPR reduction across various subcarrier configurations, providing up to an 8 dB reduction compared to baseline OTSM. In addition, LA-PP improves bit error rate (BER) performance over Rayleigh fading channels. The proposed scheme preserves the standard OTSM signal structure and introduces no additional online computational complexity, making it a practical and efficient solution for real-time, high-mobility 6G communication scenarios.