Locality-Based Aggregate Computation in Wireless Sensor Networks

The design of efficient gossip-based protocols for computing aggregate functions in wireless sensor network has received considerable attention recently. A primary reason for this surge in interest is their robustness due to their randomized nature. In addition, because of the inherent simplicity in their design, gossip-based protocols are suitable to be implemented on sensor nodes with limited computing capability and battery power. In the context of energy-constrained sensor networks, it is of paramount importance to reduce the message and time complexity of gossip-based protocols. In this paper, we present DRRgossip, an energy-efficient and robust aggregate computation algorithm in sensor networks. Exploiting locality, the Distributed Random Ranking (DRR) algorithm first builds a forest of disjoint local trees over the sensor network. The height of each tree is small, so that the aggregates of the tree nodes can be quickly obtained at the root of the tree. All the roots then perform a uniform gossip algorithm on their respective aggregates to reach a distributed consensus on the global aggregate. We prove that the DRR-gossip algorithm requires O(n) messages and O( n 3/2 log1/2 n ) transmissions to obtain aggregates on a random geometric graph. This reduces the energy consumption by at least a factor of 1 log n over the standard uniform gossip algorithm. Experiments validate the theoretical results and show that DRR-gossip needs much less transmissions than other gossip-based schemes. Index Terms wireless sensor networks; aggregate computation; gossip; distributed algorithm; randomized algorithm; random geometric graph Jen-Yeu Chen is with the Department of Electrical Engineering, National Dong Hwa University, Taiwan, R.O.C. Email:[email protected]. Gopal Pandurangan is with the Department of Computer Science, Purdue University, West Lafayette, USA. E-mail:[email protected]. Jianghai Hu is with the School of Electrical and Computer Engineering, Purdue University, West Lafayette, USA. E-mail: [email protected]. 2 SUBMITTED FOR PUBLICATION TO JSAC SI: SIMPLE WIRELESS SENSOR NETWORKING SOLUTIONS, APRIL 21, 2009