Autonomic management of future wireless networks

The traditional human controlled network management approaches may not cope with the envisioned virtualized and more dynamic mobile communication paradigm. We propose an autonomic network management and policy execution framework that re-factors the network functionalities by decomposing the network architecture into hierarchical layers. We propose an hybrid self-learning scheme and present an aggregate approach to efficient learning without reconstructing from scratch for each layer, cluster, and player, for a variety of learning algorithms widely used in practical network management. To evaluate performance of proposed framework, we develop a full-scale demonstrator. Results confirm that system learns autonomously.