Backhaul-aware small cell DTX based on fuzzy Q-Learning in heterogeneous cellular networks

In this paper, we investigate optimal control of cell discontinuous transmission (DTX) for small cells in heterogeneous cellular networks (HetNets). The small cell transmission activity is dynamically orchestrated in a centralized way to jointly minimize the power consumption of Radio Access Network (RAN) and backhaul (BH), while at the same time satisfying user Quality of Service (QoS) constraints. We propose a Fuzzy Q-Learning scheme that combines Reinforcement Learning (RL) and Fuzzy Inference System (FIS) theory to enable system optimization in realistic environments. Our analysis shows that 1) joint RAN and BH optimization is necessary to correctly design network energy saving functions and 2) the proposed controller results in notable energy saving with respect to the classic DTX approach.