Geometry-based reasoning for a large sensor network

In recent time, the study of wireless sensor networks (WSN) has become a rapidly developing research area. In a typical WSN scenario, a large swarm of small and inexpensive processor nodes, each with limited computing and communication resources, is distributed in some geometric region. Upon start-up, the swarm forms a decentralized and selforganizing network that surveys the region, communicating by wireless radio with limited range. These WSN characteristics imply absence of a central control unit, limited capabilities of nodes, and limited communication between nodes. This requires developing new algorithmic ideas that combine methods of distributed computing and network protocols with traditional centralized network algorithms. The challenge is: How can we use a limited amount of strictly local information in order to achieve distributed knowledge of global network properties? As it turns out, making use of the underlying geometry is essential. Using our toolbox Shawn [6] for the simulation of large and complex networks, we illustrate two procedures for dealing with this challenge: one identifies the boundaries of the network; the other constructs a clustering that describes the network topology. For more technical details of the underlying algorithmic side, see our recent paper [5]. Our software is freely available at www.swarmnet.de.