Skip to Main Content
The task of moving data (i.e., the routing problem) in large-scale sensor networks has to contend with several obstacles, including severe power constraints at each node and temporary, but random, failures of nodes, rendering routing schemes designed for traditional communication networks ineffective. We consider the open problem of finding optimum routes between any fixed source-destination pair in a large-scale network, such that the communication load (i.e., the required power) is distributed among all the nodes, the overall latency is minimized, and the algorithm is decentralized and robust. A recent work addressed this problem in the context of a grid topology and showed how to obtain load-balanced routing, but transmissions are restricted to be among near-neighbors and the overall latency grows linearly with the number of nodes. We show how one can route messages between source and destination nodes along random small-world topologies using a decentralized algorithm. Specifically, nodes make connections independently (based only on the source and destination information in the packets), according to a distribution that guarantees an average latency of O(log2(N)), while preventing hotspot regions by providing an almost uniform distribution of traffic load over all nodes. Surprisingly, the randomized nature of the network structure keeps the average per-node power consumption almost the same as in the case of a grid topology (i.e., local transmissions), while providing an exponential reduction in latency, resulting in a highly fault-tolerant and stable design capable of working in very dynamic environments.