Optimizing Access Demand for mMTC Traffic Using Neural Networks

Machine-type communications show unique spatial and temporal correlation properties that often lead to bursty access demand profiles. With the expected large-scale deployment of the Internet of Things (IoT), next-generation mobile networks should be redesigned to manage massive, highly synchronized arrivals of access requests by employing efficient access barring schemes. In this work, we first derived the analytical expres- sion of the optimal Access Class Barring (ACB) parameter as standardized by the Third Generation Partnership Project (3GPP). Secondly, we predict the type and number of accessing devices from measurements acquired by the Base Station (BS) by employing Neural Networks (NNs). These estimates are used to effectively implement the optimal barring scheme, achieving performance results close to the theoretical bound.