AI Enabled Resource Allocation in Future Mobile Networks

The recent past has advocated immense flexibility in the control and elasticity in the resources of the mobile networks. The emerging application domains including autonomous driving, eHealth, smart grid, etc. position the need for the right communication stretch at a pivotal level. It goes without saying that the network operators will experience the dynamic demands like never before owing to an extremely dynamic device layer i.e., IoT. An obvious consequence of this is the uncertainty in the demand estimation and capacity planning of the communication infrastructure. This paper studies the concept of dynamic demand estimation using AI approaches. We start with learning over the mobility and activity patterns of a single user and evaluate the performance of different machine learning (ML) approaches, for example, classification, regression, and clustering. We then move on to more realistic settings, where the learning is carried out for population and autonomous driving. To do so, we use the data-set from Ernst-Reuter-Platz collected as part a Berlin City project that made the data openly available.