SGRL-Selective Gateway and Reinforcement Learning-based routing for WMN

Abstract

Thanks to their flexibility and their simplicity of installation, Wireless Mesh Networks (WMNs) allow a low cost deployment of network infrastructure. They can be used to extend wired networks coverage allowing connectivity anytime and anywhere. However, WMNs may suffer from drastic performance degradation (e.g., increased packet loss ratio and delay) because of interferences and congestion. Generally, the network may be unexpectedly congested at one or more gateways (GWs) since their number is limited and most traffic is oriented to/from Internet and passes through them. In this paper, we propose a combination between a selective gateway scheme and an adaptive routing scheme in WMNs, called SGRL (Selective Gateway and Reinforcement Learning-based routing). SGRL (1) considers a probabilistic gateway selection strategy to avoid route flapping which improves network stability and traffic fairness between gateways and (2) adaptively learns an optimal routing policy taking into account multiple metrics, such as loss ratio, interference ratio and load at the gateways. Simulation results show that SGRL can significantly improve the overall network performance compared to interference and channel switching (MIC), Reinforcement Learning-based Distributed Routing (RLBDR), Expected Transmission count (ETX), load at gateways as routing metrics.