Approaching the cell switch-off problem in 5G ultra-dense networks with dynamic multi-objective optimization

Abstract

Energy efficiency is a major issue in the fifth generation (5G) of cellular networks as they require an ultra-dense deployment of small base stations (SBSs) to meet the forecasted traffic demands. Switching off cells is a widely recognized strategy to reduce the power consumption of these networks during off-peak conditions but, as times goes by, these demands change, thus requiring the activation or deactivation of a different set of cells that provide the users with a minimum QoS. In this context, the optimization problem of selecting which cells have to be switched on/off in each period of time has been approached from the dynamic multi-objective evolutionary (MOEA) domain, by proposing a novel restart method that enables the algorithms to react to changes in the traffic demands. The newly devised operator, named Adjacent Cell Restart (ACR), is based on exploiting the spatial continuity of the mobile users in the network. The experiments over a set of 9 ultra-dense networks with increasing densities of both users and base station has shown the enhanced capabilities of the ACR-enabled dynamic MOEAs to better approximate the newly induced the Pareto fronts in consecutive periods of time.