Flexible SON Function Coordination Framework based on Machine Learning

Abstract:
Self-Organizing Network (SON) functions were developed to simplify mobile network management by automatically optimizing individual performance objectives by tuning a set of configuration parameters. Due to the interdependencies among SON Functions (SFs), they may conflict when run in parallel, leading to suboptimal and unstable network behavior. This work presents an implicit coordination framework based on Machine Learning (ML) to minimize the conflicts among multiple SFs. The proposed framework utilizes a pre-trained ML model in combination with a Genetic Algorithm (GA) that allows the operation of the network with the best possible configuration from the beginning and provides near-real-time background optimization. Additionally, the proposed framework allows Mobile Network Operator (MNO) to set different degrees of priorities among SFs according to their business model. The framework is applied to two well-established conflicting SFs: Mobility Robustness Optimization (MRO) and Mobility Load Balancing (MLB). Simulation studies show that the conflicts between MRO and MLB are minimized, and the framework can be adapted to different trade-offs between functions.