Over-the-Air Transmission Aided Distributed Resource Block Allocation for Wideband Cell-Free Systems

Distributed resource allocation is a pivotal and challenging problem for cell-free systems, where a large number of access points (APs) cooperate to serve user equipments (UEs). Compared to centralized optimization that suffers from pro hibitive computational complexity and huge fronthaul overhead, distributed methods offer a desirable alternative for practical deployment. This paper studies distributed resource block (RB) allocation in cell-free orthogonal frequency-division multiplexing (OFDM) systems. We first propose a distributed sequential algorithm along with an over-the-air (OTA) transmission scheme for optimizing RB allocation, where each AP locally updates decisions using the information obtained through OTA transmissions between APs and UEs. To achieve fast convergence and reduce the OTA transmission overhead, a clustering-based variant of the sequential algorithm is developed. To alleviate the estimation error of the information acquired via OTA transmission caused by pilot contamination and noise, we further propose a distributed deep learning-based method (DDM) to jointly optimize pilot allocation and RB allocation based on the proposed distributed algorithms. Simulation results demonstrate that the proposed distributed al gorithms perform close to the centralized algorithm under perfect OTA transmission. The proposed DDM significantly outperforms existing distributed baseline methods, indicating that the DDM effectively mitigates the impact of estimation error and reduces the overhead associated with OTA transmission. Moreover, the DDMexhibits good generalizability to different numbers of APs, UEs, and RBs as well as different channel distributions.