Evolutionary design of wireless sensor networks based on complex networks Andre S. Ruela #1 , Raquel S. Cabral *2 , Andre L. L. Aquino #3 , Frederico G. Guimaraes #4 # Department of Computer - Federal University of Ouro Preto Ouro Preto, MG, Brazil 1 andrebardo@gmail.com 3 alla@iceb.ufop.br 4 frederico.g.guimaraes@gmail.com ∗ Department of Electrical Engineering - Federal University of Minas Gerais Belo Horizonte, MG, Brazil 2 raquelcabral@gmail.com Abstract—This work proposes a genetic algorithm for design- ing a wireless sensor network based on complex network theory. We develop an heuristic approach based on genetic algorithms for finding a network configuration such that its communication structure presents complex network characteristics, e.g. a small value for the average shortest path length and high cluster coefficient. The work begins with the mathematical model of the hub location problem, developed to determine the nodes which will be configured as hubs. This model was adopted within the genetic algorithm. The results reveal that our methodology allows the configuration of networks with more than a hundred nodes with complex network characteristics, thus reducing the energy consumption and the data transmission delay. I. I NTRODUCTION Wireless Sensor Networks (WSNs) represent an emerging technology that allows the monitoring and control of physical and environmental variables and conditions, such as tempera- ture, sound, light, vibration, pressure, movement and pollution. A WSN consists of a great number of wireless autonomous devices, called sensor nodes, which work in a cooperative way to perform many different functions. These characteristics make the WSNs a promising technology in a wide range of application domains, for instance, biotechnology, industry, public health, and transportation ([1], [2]). Despite its potential applicability, a WSN has several re- source restrictions, such as low computational power, reduced bandwidth, and limited energy source. Therefore, algorithms for WSNs need to be carefully designed. Thus, sending a large amount of data can be problematic, causing excessive delay in response time, invalidating the data. Moreover, a large traffic on the network can diminish its lifetime. Due to these restrictions, in some cases, it is necessary to adopt specific infrastructure designs to balance network requirements keeping its functionality. Generally, the phenomenon monitored is reported through to sink considering a specific routing strategy. Examples of routing strategies are depicted in Figs. 1–3. A simple naive, but inefficient, way of propagating information through the network is flooding (Fig. 1). In this case, the information is flooded to all sensors until it reaches the sink node [3]. This strategy causes unnecessary communication, consequently, a large energy consumption and a high response time to deliver the data. Sink Sensor Hub Fig. 1. Flooding propagation. A common alternative to flooding is tree routing (Fig. 2), a simple and low-overhead routing protocol. Using a tree routing, each sensor is configured to send its data only to a specific sensor node, denoted father node. The choice of which node will be the father depends on the policy established by the application, in general, the shortest path policy is used [4]. The major drawback of tree routing is the increased hop counts as compared with more sophisticated path search protocols. However, there is a significant energy consumption because the link is kept, i.e., all non father nodes perceive the propagated data, this situation can be seen as light gray links in Fig. 2. Sink Sensor Hub Fig. 2. Routing tree propagation. Additionally, considering applications that use thousands of