On the Information Lifetime and the Localization Cost in Sensor Networks with Random Topologies

Abstract

Sensor networks (and to some extent ad hoc networks) are networks of nodes with resource constraints, such as a limited battery life, a limited bandwidth or a limited processing capacity. These constraints create a well documented trade-off: each node needs to participate in the network to perform its role or duty, while this participation will deplete the node's resources. Here we consider a particular aspect of this trade-off: the storage of information in the network. Due to the resource constraints, each single node carries an incentive to limit the amount of data it contains. This leads to the expiration of the data carried by the node after a period of time. On the other hand, some data is critical to the functioning of the whole network, and should be found in the network at all times. In C. Westphal (2005), we introduced the trade-off between the finite lifetime of a piece of information at each node, and the survival of this information indefinitely within the network. We then studied this trade-off in a lattice topology and a tree topology. Here, we extend the results of C Westphal (2005) in two directions: we consider random topologies for the underlying network, and we take into account the cost of exchanging information in the network. We show that the maximum number of hops in a request for information broadcast is a critical parameter to ensure the survivability of any information within the network indefinitely. We identify the parameter which minimizes the load on the network for a network graph satisfying the Poisson boolean model. We also show how to minimize the cost of the dissemination on the network, so as to keep this cost decreasing asymptotically to 0