Coordination of Self-Organizing Network (SON) functions in next generation radio access networks

Overview & Synopsis

To remain competitive operators have to continuously improve the capabilities of their networks. This increases the management complexity translating into increased CAPital EXpenditures (CAPEX) and OPerational EXpenditures (OPEX). Consequently Release 8 of 3GPP introduced the Self Organizing Network (SON) functions. For all these SON functions to work properly together (especially in a multi-vendor environment) we have to detect if they conflict and enforce a resolution mechanism if this is the case. For this purpose Release 10 of 3GPP introduced the SON COordinator (SONCO) function. In this thesis we provide two frameworks. The first one is for SON conflict diagnosis (SONCO-D). The SONCO-D has to be able to identify potential and active conflicts, i.e. which ones effectively degrade the network performance. For this purpose, we use Naive Bayesian Classifiers. The second framework is for SON conflict resolution (SONCO-R). Once a conflict is identified, a SONCO-R mechanism can be applied. It decides when to favor one SON function or another based on a predefined criterion. To this end, we use a Reinforcement Learning framework as it allows us to improve our decisions based on past experience. Throughout our work we consider the handover parameters to be established per cell (the same for all neighboring cells). We motivate this choice in a dedicated chapter.