A Novel Routing Control Method Using Federated Learning in Large-Scale Wireless Mesh Networks

Currently, the volume of communication by mobile terminals are increasing owing to 5G and other technologies. A robust network and appropriate routing control methods are requied to transmit information in unstable wireless communication environments and avoid congestion. Therefore, in recent years, numerous studies have been conducted on wireless mesh networks (WMNs), which provide a fault-tolerant communication environment by securing multiple communication paths and whose topology can be freely configured and extended. Additionally, machine learning routing is attracting attention as a new routing method for wireless communication environments. However, when performing machine learning on a large WMN, the learning time increases and rapid routing control may be impossible. In this study, we apply federated learning to machine learning and propose a machine-learning-based routing method that can be applied to large-scale WMNs. Furthermore, experimental results demonstrate the effectiveness of the proposed method in various environments: congestion avoidance is achieved in a large-scale WMN by machine-learning routing using federated learning. This study is expected to serve as a basis for significant progress in the realization of large-scale WMNs as wireless communication infrastructure.