Optimal QoS Provisioning For Narrow Band IoT Devices Using Machine Learning

Abstract

With the prediction of thousands of IoT devices within a single cell in the future, it is the need of the hour to come up with solutions to mitigate the expected problems among these devices. The competition to acquire up and downlink channels for operations would become intense for these large numbers of devices as 3GPP approved NB-IoT bandwidth for long-range communication is only 200khz with a subcarrier spacing of 15khz. In addition, the competition would also contribute to the problems of lower data rate, decrease devices battery lives and disrupt continuity for devices that require continuous communication. The thesis study focuses on introducing device to device connection among IoT devices which would help save the battery life of the devices. The new connections made among the devices will utilize lower power as the algorithm designed focuses on ideal potential pairing devices with least bit error rate, least distance, and lowest transmission power. The results show the supremacy of the proposed solution which can be used to reduce the interference among device to device IoT network in addition, to the power saving capability. Deployment of the proposed cognitive power-based device to device pairing solution can help reduce the power consumed by the IoT network which can be put in the network security algorithms to make the IoT network communication more secure. Moreover, machine learning is deployed to further speed up the device to device link formation between IoT devices after the initial set of training data is produced and processed at the operator end.