Resolving SON Interactions via Self-Learning Prediction in Cellular Networks

A novel Self Organizing Networks (SON) approach is developed which is capable to handle and simultaneously optimize several highly coupled and strongly interacting configuration parameters and effects in modern cellular wireless networks. Fully distributed SON entities located in each cell predict the quality of potential candidate parameter configurations via fast offline calculations without the need of any direct system feedback. The thereby used prediction model is adapting itself via several self-learning techniques to the particular cell individual situation. This approach can optimize the system performance as well as consider energy efficiency aspects. System simulations in a heterogeneous LTE-A scenario validate this solution approach scenario and its capabilities, characteristical effects and limitations are discussed. It is a generic concept, which can be applied and transferred to several typical SON use cases with interacting parameters and coupled effects.