Optimization Models for Determining Performance Benchmarks in Wireless Sensor Networks Valeria Loscr´ ı, Enrico Natalizio, Carmelo Costanzo, Francesca Guerriero, Antonio Violi DEIS - University of Calabria, Rende, ITALY {vloscri,enatalizio,ccostanzo,guerriero}@deis.unical.it avioli@si.deis.unical.it Abstract In this paper we propose some “innovative” optimiza- tion models for Wireless Sensor Networks. The models are chosen depending on the task the network is called to exe- cute and they focus on the optimization of some specific per- formance objectives. Indeed, starting from a generic config- uration, the optimal solution defines a specific sensors dis- placement, which allows the network to achieve high per- formance, in terms of energy consumption and travelled dis- tance. Controlled mobility of the nodes is used for reaching the wanted displacement. The behaviour of the proposed models has been evaluated in comparison with a distributed algorithm on the basis of an extensive computational study and by considering different scenarios. 1. Introduction Wireless sensor networks (WSNs) are deployed to an area of interest to sense phenomena, process sensed data, and take action accordingly. Typical applications of WSNs include environmental control such as fire fighting or ma- rine ground floor erosion, but also sensors installations on bridges or buildings to understand earthquake vibration pat- terns and various surveillance tasks such as intruder surveil- lance on premises. In this paper, we focus on a multi- objective, self-organizing WSN that is able to perform dif- ferent tasks. Hence, the network consists of independent, collaborative, mobile nodes that do not only sense, process and exchange data [6], [5], [9], [12] and [14], but also move towards positions that are considered “optimal”. The opti- mality criterion is chosen by considering the specific task the network has to achieve. In this work the tasks are mod- elled through optimization models. In particular, five inno- vative optimization models, that differ in the specific perfor- mance objective to be achieved, are presented. The mathe- matical formulation of the optimization models and the def- inition of the optimal placement are computed by a cen- tral computation unit. In order to assess the behaviour of the proposed models, in terms of efficiency and robustness, an extensive computational study is carried out. The re- sults achieved by applying the proposed centralized solution strategy are compared with those obtained by a distributed algorithm, taken from the literature [7]. The remainder of the paper is organized as follows. We discuss related works in Section 2. Section 3 describes the steps of the centralized placement algorithm. Section 4 introduces the optimiza- tion models and some possible tasks or applications based on them. Section 5 recalls the distributed algorithm and presents the experimental results of the models. Finally, the paper ends with conclusions in Section 6. 2. Related Works As a result of significant advances in micro- electromechanical systems (MEMS), low power, small form factor radio transceivers and compact digital circuit designs, self-organizing wireless sensor networks are becoming a reality in residential, commercial, medi- cal, industrial and military applications. With the term self-organization, we mean the process of autonomous for- mation of connectivity, addressing, and routing structures. Self-organization of wireless sensor networks is challeng- ing because of the tight constraints on the bandwidth and energy [2] resources available in these networks. In the last few years the concept of self-organization and adaptive network has been strictly related to the concept of mobility. For example, a mobile sensor network, that is capable of changing its layout has already been put in use in [10]. In [13], the concept of artificial potential fields was introduced to guide the movement of a robot to a particular location through a field filled with obstacles. Another approach of increasing network flexibility is to allow in-network reprogramming. Two technologies that allow this are Mat´ e [11], and SensorWare [4]. In our work we consider an