Broadcasting in Sensor Networks: The Role of Local Information

Flooding based querying and broadcasting schemes have low hop-delays of Theta(1/R(n)) to reach any node that is a unit distance away, where R(n) is the transmission range of any sensor node. However, in sensor networks with large radio ranges, flooding based broadcasting schemes cause many redundant transmissions leading to a broadcast storm problem. In this paper, we study the role of geographic information and state information (i.e., memory of previous messages or transmissions) in reducing the redundant transmissions in the network. We consider three broadcasting schemes with varying levels of local information where nodes have: (i) no geographic or state information, (ii) coarse geographic information about the origin of the broadcast, and (Hi) no geographic information, but remember previously received messages. For each of these network models, we demonstrate localized forwarding algorithms for broadcast (based on geography or state information) that achieve significant reductions in the transmission overheads while maintaining hop-delays comparable to flooding based schemes. We also consider the related problem of broadcasting to a set of "spatially uniform" points in the network (lattice points) in the regime where all nodes have only a local sense of direction and demonstrate an efficient "sparse broadcast" scheme based on a branching random walk that has a low number of packet transmissions. Thus, our results show that even with very little local information, it is possible to make broadcast schemes significantly more efficient.