Towards Cognitive Autonomous Networks

4Modelling Cognitive Decision Making

Stephen S. Mwanjeand Henning Sanneck

Nokia Bell Labs, Munich, Germany

The need for automating network management has been articulated for at least a decade [1,2]. Successive work has studied the different aspects of the Network Management Automation (NMA) challenge and different solutions for the specific challenges. This has resulted in what is known in the mobile network industry as Self‐Organizing Networks (SONs) [1,2]. The core idea in SON (as discussed in Chapter 3) was the implementation of closed‐loop control mechanisms that evaluate radio network state and propose the appropriate configurations or reconfigurations of the network parameters. For some key use‐cases (e.g. auto‐connectivity, commissioning, and automatic neighbour relationship setup) the automation gains have been proven for many operators in the field. Still, those use cases concern relatively simple workflows and parameter types. Their automation gain comes more from the fact that a very frequent, but simple use case is automated. For use cases which are less frequent but more complex to solve, the degree of automation in deployed systems is still low.

At the same time, the challenges underlying the need for network automation have continued to expand. The complexity of network operations exponentially increases with new technologies and the related density of cells. This has motivated for better mechanisms for decision making when deriving actions taken by the autonomous ...