Self-Organizing Network (SON) functionalities in forthcoming 5G systems stand as a promising field for the fertilization of Artificial Intelligence (AI) mechanisms. In this respect, this paper analyzes how SON functionalities for radio access network planning and optimization in future 5G systems can be built upon AI concepts. A framework is presented that processes input data from multiple sources and extracts, through learning-based classification, prediction and clustering models, relevant knowledge to drive the decisions made by 5G SON functionalities. Different candidate AI-based tools for knowledge discovery are identified together with the associated knowledge models that can be extracted. On this basis, the applicability of these models to a comprehensive range of 5G SON functions across the categories of self-planning, self-optimization and self-healing is analyzed. Finally, the paper identifies the research directions deriving from the proposed framework.

Peer Reviewed