Service-Differentiated Joint Distributed Communication and Computing Resource Allocation for Wi-Fi Networks Based on Federated Learning and MADRL

To effectively support diverse services and applications, operation of future Wi-Fi networks has to be highly intelligent. AI/ML has been considered an important component of the next-generation Wi-Fi standard (i.e., Wi-Fi 8). In addition, the distributed and relatively stable Wi-Fi operational environment bring more realistic AI/ML applications than other wireless networks. By opportunistically leveraging distributed characteristics of Wi-Fi, this article focuses on the service-differentiated joint optimization of communication and computing resource allocation with varying privacy sensitivity. By extending multiagent deep reinforcement learning (MADRL), a new semi-centralized and fully distributed joint resource optimization structure is created. Our purpose is to maximize the service quality of the deep reinforcement learning (DRL) model for privacy-insensitive users while minimizing private information exchange during the training period of privacy-sensitive users during the DRL interaction process. The proposed approach leverages the varying service requirements of different stations (STAs), facilitates real-time optimization of local communication resource allocation, and enables concurrent decision-making for computing resources. In addition, we explored the heterogeneous differences in channel states between communication nodes and utilized the federated weighting (FedWgt) method to address this issue, further improving the stability of the distributed model in solving the service-differentiated joint optimization problem of resource allocation. Extensive simulation experiments demonstrate that the proposed scheme outperforms baseline methods significantly in terms of throughput, calculation latency, and energy consumption improvement.