Abstract
The 5G heterogeneous network (Het-Net) comprises macro cells and small cells. The small cells with the ultra-dense deployment can offload mobile data traffic from macro cells and extend service area while consuming less energy. However, frequent handoffs between the two types of cells result in high signaling costs and interference. Thus, determining when to switch small cells between active and inactive modes is crucial to reducing operation cost. This paper proposes a Reinforcement Learning-based network management mechanism for 5G HetNet, and simulation experiments were conducted to evaluate its performance, in contrast to previous works that utilized 3GPP standardized Self-Organizing Network (SON) for network management mechanisms.