A Novel Multiobjective Cell Switch-Off Framework for Cellular Networks

TL;DR

This paper presents a multiobjective cell switch-off framework that reduces energy consumption in cellular networks by optimizing resource allocation, minimizing transitions, and executing heavy computations offline, all while maintaining QoS.

Contribution

It introduces a novel multiobjective optimization approach for cell switch-off that minimizes transitions and handovers, with offline computation for practical implementation.

Findings

Achieves significant energy savings in small cell networks.

Maintains QoS without heavy real-time processing.

Effectively handles non-uniform service demand.

Abstract

Cell Switch-Off (CSO) is recognized as a promising approach to reduce the energy consumption in next-generation cellular networks. However, CSO poses serious challenges not only from the resource allocation perspective but also from the implementation point of view. Indeed, CSO represents a difficult optimization problem due to its NP-complete nature. Moreover, there are a number of important practical limitations in the implementation of CSO schemes, such as the need for minimizing the real-time complexity and the number of on-off/off-on transitions and CSO-induced handovers. This article introduces a novel approach to CSO based on multiobjective optimization that makes use of the statistical description of the service demand (known by operators). In addition, downlink and uplink coverage criteria are included and a comparative analysis between different models to characterize intercell interference is also presented to shed light on their impact on CSO. The framework distinguishes itself from other proposals in two ways: 1) The number of on-off/off-on transitions as well as handovers are minimized, and 2) the computationally-heavy part of the algorithm is executed offline, which makes its implementation feasible. The results show that the proposed scheme achieves substantial energy savings in small cell deployments where service demand is not uniformly distributed, without compromising the Quality-of-Service (QoS) or requiring heavy real-time processing.