DRE2: Achieving Data Resilience in Wireless Sensor Networks: A Quadratic Programming Approach

We focus on sensor networks that are deployed in challenging environments, wherein sensors do not always have connected paths to a base station, and propose a new data resilience problem. We refer to it as DRE2: data -resiliency in $\underline{e}$xtreme $\underline{e}$nvironments. As there are no connected paths between sensors and the base station, the goal of $\mathrm{D}\mathrm{R}\mathrm{E}^{2}$ is to maximize data resilience by preserving the overflow data inside the network for maximum amount of time, considering that sensor nodes have limited storage capacity and unreplenishable battery power. We propose a quadratic programming-based algorithm to solve $\mathrm{D}\mathrm{R}\mathrm{E}^{2}$ optimally. As quadratic programming is NP-hard thus not scalable, we design two time efficient heuristics based on different network metrics. We show via extensive experiments that all algorithms can achieve high data resiliences, while a minimum cost flow-based is most energy-efficient. Our algorithms tolerate node failures and network partitions caused by energy depletion of sensor nodes. Underlying our algorithms are flow networks that generalize the edge capacity constraint well-accepted in traditional network flow theory.