FedStar Caching: Decision Center Assisted Federated Cooperative Edge Caching

With the explosion of data, Wireless Edge Caching (WEC) has become a promising approach for locally accessing cached contents. Due to the limited storage capacity of local caching devices and the varying preferences of humans for content, it is necessary to predict popular contents. In this article, we will address two distinct objectives: 1) Popularity prediction of the content that should be cached at the edge to effectively utilize edge device memory. 2) Edge devices collaboration for efficient content delivery. Accordingly, we propose an FL-based Star Cooperative Caching (FedStar Caching), utilizing a Star network topology for realizing an efficient cooperation among Femto-cell Access Points (FAPs) using a Decision Center (DC) to enhance power efficiency and delay performance. Considering human sensitivity to privacy, collecting users’ data on a central server is not desirable, therefore, we leverage Federated Learning (FL) to alleviate this challenge. Simulation results demonstrate that the proposed method outperforms alternatives in terms of cache efficiency, as well as delay and power consumption.