Robust green Wireless Local Area Networks: A matheuristic approach

Abstract

We consider the problem of designing a Wireless Local Area Network according to a Green paradigm (GWLAN), i.e. minimizing the power consumption of the network by powering-on just a subset of access points and associating users with the powered-on access points. To protect the system against fluctuations in the data rate transmission between users and access points, which naturally affect the problem because of uncertainty in measuring the exact user position and because of wireless propagation conditions, we address the robust extension of this problem, i.e. the Robust Green Wireless Local Area Networks Problem (RGWLAN), and we propose some robust matheuristics to solve it. Such matheuristics are based on two mathematical models to RGWLAN, i.e. a model based on a generalization of the classical Γ-Robustness framework, and a model exploiting the Multiband Robustness paradigm. The models are enriched by means of a preprocessing powering-on heuristic, aimed at fixing some access points as powered-on, and a reallocation algorithm, aimed at reallocating users to access points once the model solution has been computed, in order to enhance its degree of robustness. The aim is to achieve a good compromise between power saving, guaranteed level of robustness and required computational time.The results of an extensive computational analysis show that the proposed matheuristics solve the RGWLAN in a very efficient way both in terms of power consumption and computational time. Also, they are able to guarantee a high level of robustness with respect to the uncertainty in the user positions. This trend is particularly evident when the users to access points ratio is low, i.e. when the offered traffic to the wireless local area network is scarce, which is the most suitable and relevant scenario for the power consumption optimization.