DeepOFW: Deep Learning-Driven OFDM-Flexible Waveform Modulation for Peak-to-Average Power Ratio Reduction

Abstract

Peak-to-average power ratio (PAPR) remains a major limitation of multicarrier modulation schemes such as orthogonal frequency-division multiplexing (OFDM), reducing power amplifier efficiency and limiting practical transmit power. In this work, we propose DeepOFW, a deep learning-driven OFDM-flexible waveform modulation framework that enables data-driven waveform design while preserving the low-complexity hardware structure of conventional transceivers. The proposed architecture is fully differentiable, allowing end-to-end optimization of waveform generation and receiver processing under practical physical constraints. Unlike neural transceiver approaches that require deep learning inference at both ends of the link, DeepOFW confines the learning stage to an offline or centralized unit, enabling deployment on standard transmitter and receiver hardware without additional computational overhead. The framework jointly optimizes waveform representations and detection parameters while explicitly incorporating PAPR constraints during training. Extensive simulations over 3GPP multipath channels demonstrate that the learned waveforms significantly reduce PAPR compared with classical OFDM while simultaneously improving bit error rate (BER) performance relative to state-of-the-art transmission schemes. These results highlight the potential of data-driven waveform design to enhance multicarrier communication systems while maintaining hardware-efficient implementations. An open-source implementation of the proposed framework is released to facilitate reproducible research and practical adoption.

Figures (9)

• Figure 1: Illustration of the proposed OFDM-type transmission framework in continuous-time and discrete-time implementations.
• Figure 2: Possible deployment strategy of DeepOFW in a star topology, suitable for WLAN 802.11 standards. Each station interacts with the AP to obtain $\mathbf{q}$ and $\mathbf{Q}$. The channel state information is first delivered to the AP by the stations. The AP, equipment with DL accelerators, generate $\mathbf{q}_i,\mathbf{Q}_i$ for station $i$ accordingly, and and communicates them back to the station.
• Figure 3: An illustration of the element-wise multiplication detector $D_{\mathbf{q}}\{ \mathbf{r}\}$.
• Figure 4: E2E model architecture. The transmitter–receiver chain is trained end-to-end to learn a complex waveform matrix $\mathbf{Q}$ and detector parameters $\mathbf{q}$ from instantaneous time-domain channel impulse responses. A GRU-based network generates $\mathbf{Q}$ and $\mathbf{q}$, which are applied by the $\mathbf{Q}$ modulator and $\mathbf{Q}$ demodulator together with a one-tap detector to produce LLRs for decoding. An uncertainty network and a PAPR network predict per-sample weighting factors and a regularization coefficient, enabling joint optimization of BER and PAPR during training.
• Figure 5: CCDF comparison between conventional OFDM (green curve with star markers) and DeepOFW. Each DeepOFW curve corresponds to a different RMS delay spread realization, color-coded from blue (low delay spread) to red (high delay spread).