Big Data and Self-organization for Mobile Networks

We deal in this research line with the open challenges related with the autonomic management of future 5G networks, characterized by the coexistence of extremely dense and heterogeneous deployments, new architectural paradigms, such as CloudRAN or SoftRAN, new deployment approaches based on Network Functions Virtualization (NFV) and Software-Defined Networking (SDN), which are being introduced to make mobile network deployments more cost-effective. We address the management through the well-established paradigm of Self-Organizing Networks. We propose novel solutions for self-configuration, self-optimization, self-healing and self-coordination, often taking advantage of the theoretical framework of Markov Decision Processes and Reinforcement Learning. We believe that Machine Learning in general offers a powerful tool to provide the network with the capability of learning from experience and self-adapting to the dynamics of the always varying wireless environment. In addition to this, we believe that the huge amount of measurement, control and management information, generated during normal operation of 4G/5G networks can be gathered for network management purposes in data centres. In this context, big data approaches represent a huge opportunity for mobile network operators from both business and network wide optimization perspectives.