MADDPG-Based Power Control for Asynchronous Cell-Free Massive MIMO-OFDM System

Orthogonal frequency-division multiplexing-based cell-free massive multiple-input multiple-output (CF mMIMOOFDM) system is considered a promising technology with the potential to offer ubiquitous connectivity and high spectral efficiency (SE). Nevertheless, in such networks, the distances between APs and users are different, making it challenging for signals to arrive at the receivers simultaneously, giving rise to asynchronous reception effects. To address the issue, this paper proposes a multi-agent deep deterministic policy gradient (MADDPG)-based power control scheme to enhance SE. To be specific, we first derive minimum-mean-square error channel estimation under asynchronous reception effects for cell-free mMIMO-OFDM systems. Subsequently, user-centric clustering is performed based on the imperfect channel state information (CSI). Finally, in the downlink transmission, we develop a MADDPG-based power control scheme to maximize the SE under the constraint of AP transmitting power. Simulation results demonstrate that the proposed scheme exhibits good convergence. Furthermore, compared to the conventional power allocation scheme based on water-filling algorithm, the proposed scheme significantly improves the SE of CF mMIMO-OFDM systems with asynchronous reception effect.