Handover Optimality in Heterogeneous Networks

This paper introduces a new theoretical framework for optimal handover procedures in heterogeneous networks by devising the novel fractional Gittins indices, which are dynamical priorities whose values can be statically associated to the decision states of evolving processes representing handover alternatives. The simple policy of activating at any time the one process currently at highest priority optimizes the bandwidth of a handover, if all other inactive processes remain idle. However, numerical evidence shows that in practice this condition can be relaxed for a wide range of handover models, because the bandwidth actually achieved by the policy never deviates for more than 12% from the optimally achievable bandwidth and remains in median within a deviation of 2% from this optimum.