Efficient Multi-Connectivity Handover Algorithm in Heterogeneous Cellular Networks by Graph-to-Sequence Reinforcement Learning

In the literature, there are many handover algorithms of selecting one target serving base station (BS) for a user equipment (UE). However, these algorithms are not suitable for a UE capable of the multi-connectivity communication since they do not address how to determine an adequate number of the target serving BSs for an individual UE and the best combination of the target serving BSs for a UE usually cannot be found by one-by-one greedy selection. On the other hand, the up-to-date handover algorithms of selecting multiple target serving BSs for a UE either do not determine an adequate number of the target serving BSs or demand a considerable time to choose a good set of target serving BSs. In this paper, we propose a Graph-to-Sequence reinforcement learning method to fill this gap. Simulations show that the proposed method outperforms the state-of-the-art algorithms in terms of the average quality of experience (QoE) of a UE.