Improving Routing Performance in Mobile Ad Hoc Networks Using Artificial Neural Networks for Mobility Prediction using deep learning

Abstract

In this study, a research was done to examine the effectiveness of deep learning models, such as the Artificial Neural Networks, Convolutional Neural Networks, Gated Recurrent Units and Generative Adversarial Networks, to boost routing’s efficiency and performance in the Mobile Ad Hoc Networks through mobility prediction. In total, 10 experiments were conducted, and ANNs demonstrated an average prediction accuracy of 0.86, MAE of 0.10, RMSE of 0.16, and correlation coefficient of 0.93. CNNs showed the most impressive performance, featuring the following indicators: the average prediction accuracy of 0.87, MAE of 0.09, RMSE of 0.15, and the correlation coefficient of 0.94. GRUs, in their turn, displayed quite decent performance, with an average prediction accuracy of 0.87, MAE of 0.10, RMSE of 0.16, and the correlation coefficient of 0.94; in the meanwhile, GANs can be also seen as offering a decent performance, with the average prediction accuracy of 0.87, MAE of 0.10, RMSE of 0.16, and the correlation coefficient of 0.94. In the end, the identified findings imply that deep learning models can be used to enhance the routing efficiency in MANETs, since they can predict the node movement with the high level of accuracy. The identified changes are possible to lead to the creation of effective wireless communication networks.