Ricerc@Sapienza

We are experiencing what is called the fourth industrial revolution, a transformation of the world which will drastically change our way of living. In this paradigm, technology plays a fundamental role, and connection between humans and devices is at the base of this transformation. Fundamental parts of this change are Internet of Things (IoT), Industry 4.0, automated vehicles, etc. The connection between billions of devices and their management also require a new rethinking of the network, which finds its solution in 5G networks. 5G system plans to go beyond the mere enhancement of the system capacity. Indeed, this system will enable high reliability and low latency services with such different requirements for each sector (IoT, Industry 4.0) that a flexible design and a proactive allocation of resources is necessary. An important role in this framework will be played by Multi-Access Edge Computing (MEC), a new technology that will bring computation and storage capabilities at the edge of the network, reducing the latency in accessing them. Radio and computation resource will be then managed together. In this complex system, the role of machine learning becomes essential in communication networks, paving the way to the estimation/prediction of user habits and mobility patterns, to place content and services before they are actual requested, in order to reduce latency in providing them. Once the learnt parameters are available, a dynamic optimization is required to make the Quality of experience (QoE) of the user above the threshold that 5G promises. In this proposal, machine learning algorithms and resource allocation strategies will be coupled to achieve the results necessary to meet requirements that are at the base of the 5G challenge.