Using artificial intelligence to support emerging networks management approaches

In emergent networks such as Internet of Things (IoT) and 5G applications, network traffic estimation is of great importance to forecast impacts on resource allocation that can influence the quality of service. Besides, controlling the network delay caused with route selection is still a notable challenge, owing to the high mobility of the devices. To analyse the trade-off between traffic forecasting accuracy and the complexity of artificial intelligence models used in this scenario, this work first evaluates the behavior of several traffic load forecasting models in a resource sharing environment. Moreover, in order to alleviate the routing problem in highly dynamic ad-hoc networks, this work also proposes a machine-learning-based routing scheme to reduce network delay in the high-mobility scenarios of flying ad-hoc networks, entitled Q-FANET. The performance of this new algorithm is compared with other methods using the WSNet simulator. With the obtained complexity analysis and the performed simulations, on one hand the best traffic load forecast model can be chosen, and on the other, the proposed routing solution presents lower delay, higher packet delivery ratio and lower jitter in highly dynamic networks than existing state-of-art methods