Scholarlink

<p>This paper investigates the Peak-to-Average Power Ratio (PAPR) performance of Orthogonal Time Sequence Multiplexing (OTSM) waveforms under various reduction schemes, highlighting their suitability for beyond-Fifth Generation (5G) and future Sixth Generation (6G) wireless systems. A novel Deep Reinforcement Learning-based Tone Reservation (DRL-TR) approach is proposed, which leverages a continuous actor–critic-based DRL framework that optimizes reserved-tone coefficients to effectively control high signal peaks. Simulations are conducted under Rayleigh fading channels using 256-QAM modulation across different sub-carrier configurations (64, 256, 512, and 1024), incorporating channel estimation errors up to 20% to reflect realistic conditions. At a complementary cumulative distribution function (CCDF) of 10⁻³, DRL-TR achieves a substantial PAPR reduction of 8.5 dB to 9.2 dB over the original OTSM waveform and outperforms contemporary methods—including Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Autoencoder (AE), Partial Transmit Sequence (PTS), Selective Mapping (SLM), Tone Reservation (TR), and Clipping—by margins of 2.1 dB to 8.7 dB. Notably, this translates into power savings of up to 81%. Bit Error Rate (BER) analysis further shows that the proposed DRL-TR framework reduces the required SNR at a target BER of 10⁻³ by up to 12.6 dB under severe channel estimation errors and practical nonlinear transmission conditions, demonstrating its robustness and reliability for beyond-5G and future 6G wireless communication systems. Complexity analysis shows that while DRL-TR demands higher training resources due to neural optimization, it offers practical inference complexity. These findings establish DRL-TR as a first-of-its-kind integration of deep learning with tone reservation for OTSM, demonstrating the potential for highly efficient, reliable, and spectrally robust transmission in future wireless networks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

PAPR reduction using Deep Reinforcement Learning based Tone Reservation for 5G OTSM waveform

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • Arun Kumar,
  • Venkatachalam Revathi,
  • Nishant Gaur,
  • Mehedi Masud,
  • Aziz Nanthaamornphong

摘要

This paper investigates the Peak-to-Average Power Ratio (PAPR) performance of Orthogonal Time Sequence Multiplexing (OTSM) waveforms under various reduction schemes, highlighting their suitability for beyond-Fifth Generation (5G) and future Sixth Generation (6G) wireless systems. A novel Deep Reinforcement Learning-based Tone Reservation (DRL-TR) approach is proposed, which leverages a continuous actor–critic-based DRL framework that optimizes reserved-tone coefficients to effectively control high signal peaks. Simulations are conducted under Rayleigh fading channels using 256-QAM modulation across different sub-carrier configurations (64, 256, 512, and 1024), incorporating channel estimation errors up to 20% to reflect realistic conditions. At a complementary cumulative distribution function (CCDF) of 10⁻³, DRL-TR achieves a substantial PAPR reduction of 8.5 dB to 9.2 dB over the original OTSM waveform and outperforms contemporary methods—including Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Autoencoder (AE), Partial Transmit Sequence (PTS), Selective Mapping (SLM), Tone Reservation (TR), and Clipping—by margins of 2.1 dB to 8.7 dB. Notably, this translates into power savings of up to 81%. Bit Error Rate (BER) analysis further shows that the proposed DRL-TR framework reduces the required SNR at a target BER of 10⁻³ by up to 12.6 dB under severe channel estimation errors and practical nonlinear transmission conditions, demonstrating its robustness and reliability for beyond-5G and future 6G wireless communication systems. Complexity analysis shows that while DRL-TR demands higher training resources due to neural optimization, it offers practical inference complexity. These findings establish DRL-TR as a first-of-its-kind integration of deep learning with tone reservation for OTSM, demonstrating the potential for highly efficient, reliable, and spectrally robust transmission in future wireless networks.