A Multilayer Perceptron Model for Station Grouping in IEEE 802.11ah Networks.

With the rapid development of smart devices and wireless communication technologies, IEEE 802.11ah (WiFi HaLow) is designed to solve one of the major problems of Internet of Things (IoT): high collision probability in dense networks. It proposes the Restricted Access Window (RAW) mechanism, where stations (sensors) are partitioned into groups for time-division channel access. The grouping strategy, which highly influences network performance, needs to consider factors including the number of stations per group, and stations’ data rates, and locations. With the advance of artificial intelligence technologies, we ponder whether deep learning can help solving this station grouping problem. In this paper, we propose a multilayer perceptron (MLP) model to predict RAW performance. More precisely, the model predicts the corresponding throughputs and packet loss rates of a given set of RAW configurations. Thus, based on the predicted results, we can determine proper RAW parameters. We have validated the proposed method by ns-3 simulations.