On The Structure of Barrier Graphs

Wireless sensor networks are commonly used to monitor various environmental conditions. Possible geometries for the region covered by a sensor include disks, wedges, and rays, among others, depending on the function of the sensor. In this paper we consider a network consisting of ray sensors deployed to detect intruders traversing a path, not necessarily straight, from a source α to a destination β. The coverage of the network with respect to α and β can be modeled by a tripartite graph, the barrier graph of the network. While all barrier graphs are tripartite, the converse is not true (for instance, C5 is not a barrier graph). The main result of this paper is a rigidity theorem on the structure of barrier graphs that results from constraints imposed by the geometry of the network. This allows us to show that almost all bipartite graphs are not barrier graphs, despite the fact that various classes of bipartite graphs, including trees, cycles of even length, and Km,n are barrier graphs. Furthermore, vertex cover of a barrier graph corresponds to a set of sensors whose removal allows a clear path from α to β. While all bipartite graphs with small vertex covers are barrier graphs (a fact we prove for sizes less than 4), the rigidity property also implies that graphs with vertex covers bigger than a certain constant are not barrier graphs.