On area coverage reliability of wireless sensor networks

In this paper we consider wireless sensor networks (WSNs) whose nodes can fail independently of each other during normal operation. For successful operation, a network is required to provide sensing coverage of an aggregate area that exceeds a given application-specific coverage requirement. We formulate an area coverage reliability problem that quantifies the likelihood that the network can be in an operating state where the coverage condition is satisfied. Existing results show that the problem is #P-hard even when the WSN is restricted to a rectangular grid with a single corner-located sink. Thus, an exact solution is unlikely to exist for many WSN topologies. We present a framework for computing reliability lower bounds by decomposing an arbitrary WSN into subnetworks that share only the sink node and computing coverage probabilities for each subnetwork. We show that the framework yields an efficiently computable lower bound when the subnetworks are chosen to be trees and cycles. Our design utilizes a 3-state node reliability model that has been shown to improve over the conventional 2-state (operate/fail) model. The obtained results investigate some design issues of the framework, the quality of the obtained lower bounds, and discuss an application of the obtained lower bounds in improving the design of WSNs.