Handover Mitigation in Dense HetNets via Bandit Arm Elimination

Dense heterogeneous network (HetNet) has become one of the key architectural features in the 5G communication network. In this paper, we aim at solving the frequent handover (FHO) problem in dense HetNets. We first present an unified framework suitable for various optimization objectives in wireless communication networks. A novel bandit algorithm is then proposed to obtain better results compared with the existing works. The corresponding bounds for the regret and the number of handovers are rigorously proved accordingly. The results are further generalized by taking into account the imperfect feedbacks. Extensive performance comparisons with different bandit algorithms are provided in order to highlight the advantages of the proposed algorithm and its potential for pursuing a general solution.