A Survey of PAPR Techniques Based on Machine Learning

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) is the modulation technology used in Fourth Generation (4G) and Fifth Generation (5G) wireless communication systems, and it will likely be essential to Sixth Generation (6G) wireless communication systems. However, OFDM introduces a high Peak to Average Power Ratio (PAPR) in the time domain due to constructive interference among multiple subcarriers, increasing the complexity and cost of the amplifiers and, consequently, the cost and complexity of 6G networks. Therefore, the development of new solutions to reduce the PAPR in OFDM systems is crucial to 6G networks. The application of Machine Learning (ML) has emerged as a promising avenue for tackling PAPR issues. Along this line, this paper presents a comprehensive review of PAPR optimization techniques with a focus on ML approaches. From this survey, it becomes clear that ML solutions offer customized optimization, effective search space navigation, and real-time adaptability. In light of the demands of evolving 6G networks, integration of ML is a necessity to propel advancements and meet increasing prerequisites. This integration not only presents possibilities for PAPR reduction but also calls for continued exploration to harness its potential and ensure efficient and reliable communication within 6G networks.

Power Amplifier Predistortion Using Reduced Sampling Rates in the Forward and Feedback Paths

“…Hence, numerous studies have been conducted in order to alleviate this problem. Some of the proposed solutions include the peak-to-average power reduction techniques presented in [ 14 ], the reduced complexity DPD presented in [ 8 ], linearization under high data sparsity [ 15 ], and the feature selection methodologies presented in [ 16 ]. The concept of feature selection [ 16 ], in particular, is of great interest, as it attempts to limit the amount of basis functions in a DPD model required for adequate linearization.…”