The inherently large amplitude variations of OFDM signals in multi-user MIMO systems impose strict linearity requirements on power amplifiers, reducing energy efficiency and degrading effective throughput due to nonlinear distortion. Although null-space-based peak cancellation techniques such as PCCNC enable distortion-free PAPR reduction by confining correction signals to the channel null space, their performance remains restricted by static projection control and limited adaptability to dynamic channel conditions. In particular, insufficient regulation of projection strength results in residual distortion that directly impacts effective SINR and system throughput. To address these challenges, this paper proposes a learning assisted null-space peak cancellation framework that integrates adaptive projection scaling with throughput-aware optimization. The proposed method dynamically adjusts the null-space correction strength using real-time performance metrics, including PAPR evolution, peak reduction gain, antenna power variance, and effective SINR. By incorporating a lightweight feedback-driven learning mechanism, the algorithm improves projection efficiency, reduces residual nonlinear distortion, and enhances convergence behavior. Extensive MATLAB-based simulations demonstrate that the proposed approach achieves improved effective SINR and noticeable throughput enhancement compared to conventional PCCNC methods, while maintaining distortion-free transmission and low computational complexity. The results confirm that learning-assisted null-space peak cancellation provides a practical and scalable solution for throughput optimization in next-generation high-capacity MIMO-OFDM wireless systems
Kumari et al. (Fri,) studied this question.