Learnable ADMM Based OFDM Phase Noise Compensation and Signal Detection Over High Mobility Channels

A novel joint phase noise compensation and signal detection algorithm based on deep learning (DL) is proposed for orthogonal frequency division multiplexing (OFDM) systems under rapidly time-varying channels. Specifically, we first develop an alternating direction method of multipliers (ADMM) based solution to the difficult mixed-integer maximum likelihood (ML) estimation, along with a low-complexity ADMM solution that exploits the inherent feature of the inter-carrier interference (ICI). Furthermore, by unfolding the ADMM iterations, we present a model-driven network with trainable variables, which avoids cumbersome parameter search and also results in faster convergence. Simulation results demonstrate that our proposed DL-aided ADMM algorithms significantly outperform existing baselines in terms of BER performance with acceptable complexity.