Deep Learning-Based Joint Channel Estimation and Implicit CSI Feedback

Previous studies often address channel estimation and channel state information (CSI) implicit feedback issues in 5G new radio (NR) system independently, posing a challenge to achieve effective coordination between the two processes in a shared environment. Consequently, we propose a deep learning-based scheme that jointly addresses channel estimation and implicit CSI feedback. In this scheme, the received pilot information from user equipment (UE) is regarded as a two-dimensional image. Channel estimation is performed employing the ResMLP architecture, whereas the Transformer architecture is utilized for implicit CSI feedback. Furthermore, a novel training approach is introduced to further enhance the performance of the proposed scheme. Simulation results illustrate a substantial improvement compared to the reference scheme employed in the 3rd Wireless Communication Artificial Intelligence (AI) Competition (WAIC).