Fuzzy Q-Learning for Mobility Robustness Optimization in wireless networks

The high popularity of smartphones and mobile PCs is expected to increase wireless data traffic in the order of 1000 times by 2020 [1]. However, the current situation of Mobile Network Operator (MNO)s is characterized by increasing margin pressure due to declining revenues and an increasing cost base. Self-optimization functionalities, e.g. for Mobility Robustness Optimization (MRO), are essential means for reducing Operational Expenditure (OPEX). In particular, mobile user groups or moving networks at high speeds impose challenges and may severely degrade network performance as well as user experience. The Fuzzy Q-Learning-based approach presented in this paper aims at providing a generic basis for enabling self-optimizing and self-healing network operations. The designed concept consists of the following key components: Fuzzy Inference System (FIS), heuristic Exploration/Exploitation Policy (EEP), and Q-Learning (QL). Its performance in a reference scenario is compared with a trend-based handover (HO) optimization scheme presented in [2] and a scheme that assigns time-to-trigger (TTT) values based on velocity estimates.