Fuzzy Multiple Criteria Switch Off Method for Dense Heterogeneous Networks Anna Dudnikova * , Paolo Dini † , Lorenza Giupponi † , Daniela Panno * * Dept. of Electrical, Electronics and Computer Engineering, University of Catania, Italy anna.dudnikova@dieei.unict.it, daniela.panno@dieei.unict.it † Centre Tecnologic de Telecomunicacions de Catalunya (CTTC), Castelldefels, Barcelona, Spain lorenza.giupponi@cttc.es, paolo.dini@cttc.es Abstract—The growth in mobile traffic demand is leading to a dense heterogeneous cellular network. This massive deployment of mobile equipment (i.e. base stations) may cause a high increment of the network energy consumption and therefore operational expenditure for operators. One of the most promising techniques to save energy (and costs) is to switch off some underutilized cells during off peak hours. In this line, our focus is to optimize the number of base stations in dense LTE pico cell deployments in order to maximize the energy saving, while satisfying the Quality of Service constraints. We use a combination of Fuzzy Logic, Grey Relational Analysis and Analytic Hierarchy Process tools to trigger the switch off actions, and jointly consider multiple decision inputs for each cell. Keywords—cell switch off; LTE; energy saving; fuzzy logic; analytic hierarchy process; grey relational analysis I. INTRODUCTION The traffic volume in cellular networks has increased considerably during recent years and this growth is expected to continue at an exponential rate [1]. This forces operators to embed small cells in the traditional macrocell layout. The increased network densification allows to satisfy the traffic volume in peak hours, while leading to a significant concern about the network energy consumption. As a result, the green communication paradigm has received much attention from research projects [2] and standardization activities [3]. The Base Stations (BSs) are identified as one of the main causes for energy consumption in cellular networks [4], so that an effective method to preserve energy is to exploit the existence of low traffic periods and to switch off some underutilized capacity‒booster cells. The identification of the set of BSs to be switched off is not a trivial task though, and it is influenced by the behavior of multiple variables. The random or inappropriate switch off of the cells can seriously deteriorate the performance of the system since the BSs, which remain active, need to serve some extra traffic. Given the relevance of the problem, numerous proposals have been presented in literature. The cell‒zooming algorithm, proposed in [5], sequentially switches off BSs starting with the least loaded one, and going on till it finds the first cell that cannot be switched off because at least one of its users cannot be served by any of the neighboring cells. An improved cell‒zooming was proposed in [6], where the algorithm checks all the cells in the network for possible switch off. This algorithm permits to save more energy with respect to [5], but at the expense of an increased computational complexity. In [7] the switching off mechanism is based on a combination of traffic load and interference information. In [8] and [9] the authors propose a distance aware BS switch off strategy. The BS which has a maximal average distance from its associated User Equipments (UEs) and from those of the neighboring cells is chosen as a candidate for switch off process. If the load of this cell can be handled by its neighbors, the BS is switched off. Another BS sleeping algorithm based on UEs' location information is proposed in [10]. The UEs are classified into clusters based on their spatial features and for each cluster the closest BS is chosen as a cell to provide access service. None of these works address the problem of dense small cell deployments, where a co‒tier interference is a challenging issue, and the majority of them considers the load of the cells as the main input for decisions, while also other aspects should be taken into account, such as the coverage of the whole network, the interference perceived by the users, the load of the cells, etc. Also, it is worth mentioning that in [5‒10] the switching decision requires high computational complexity and large signaling overhead, since in [5‒7] every time that a node is selected for switch off, the network needs to evaluate whether the traffic to be offloaded can be absorbed by the network, and [8‒10] exploit user positioning information. This computation complexity can be reduced if the switch off algorithm is able to make more precise decisions. In this paper we propose a switch off method, which differs from the previous works in the following contributions. We target (i) a co‒channel dense pico cell deployment, which is one of the main trends of modern wireless networking, (ii) a multi‒criteria decision making, which allows the system to achieve higher throughput/energy tradeoff. The proposed solution implements a controller based on fuzzy theory, which is able to analyze the demand of traffic and the offer of resources in the scenario, in order to quickly evaluate whether the network is in the position to switch off any cell. This avoids any further checks on the feasibility of the switch off choice and reduces computational complexity with respect to the state of the art solutions. As for the switch off solution, we propose combination of Grey Relational Analysis (GRA) [11] and Analytic Hierarchy Process (AHP) [12] for the multi‒criteria decision making process. The combination of these tools allows us to process jointly multiple heterogeneous