A robust data delivery protocol for large scale sensor networks

Recent technology advances in low-cost, low-power chip designs have made feasible the deployment of large-scale sensor networks. Although data forwarding has been among the first set of issues explored in sensor networking, how to reliably deliver sensing data through a vast field of small, vulnerable sensors remains a research challenge. In this paper we present GRAdient Broadcast (GRAB), a new set of mechanisms and protocols which is designed specifically for robust data delivery in spite of unreliable nodes and fallible wireless links. Similar to previous work [1], GRAB builds and maintains a cost field, providing each sensor in the network the direction to forward sensing data. Different from all the existing approaches, however, GRAB forwards data along an interleaved mesh from each source to the receiver. The width of the forwarding mesh is controlled by the amount of credit carried in each data message, allowing the degree of delivery robustness to be adjusted by the sender. GRAB design harnesses the advantage of large scale and relies on the collective efforts of multiple nodes to deliver data, without dependency on any individual ones. As demonstrated in our extensive simulation experiments, GRAB can successfully deliver above 90% of data with relatively low energy cost even under adverse conditions of up to 30% node failures compounded with 15% link packet losses.