Group Owner Selection Based on Artificial Neural Networks in Mobile Ad hoc Wi-Fi Direct Networks

This paper investigates the use of Artificial Neural Networks (ANN) to enhance Group Owner (GO) selection in Mobile Ad hoc Networks based on Wi-Fi Direct technology. These networks are decentralised, with no fixed access point, and require optimal GO selection to ensure group persistence with maximum stability. Traditional GO selection methods, based on a single criterion such as Intent Value, can lead to inappropriate GO selection and fail to consider the heterogeneous and the dynamic nature of the network. To overcome this issue, we design a classification model and a regression model, both based on Artificial Neural Network techniques. The classification model identifies the most suitable node to act as the GO based on several parameters. It enables fast, efficient decision-making by directly selecting the GO from the available nodes, based on a binary output (0 for GO or 1 for noGO). The regression model provides a continuous estimate of Intent Value of each node based on several parameters collected on the node, offering a finer measure of a device’s willingness to become a GO. These models were trained with a weighted dataset and evaluated using the performance metrics recommended for classification and regression in Artificial Neural Networks. The results show that these ANN models offer a promising solution for improving the management of ad hoc networks by providing more adaptive and intelligent GO selection decisions. Our approach ensures that the most capable node is elected as group owner.