PEAS: A Robust Energy Conserving Protocol for Long-lived Sensor Networks

Small, inexpensive sensors with limited memory, computing power and short battery lifetimes are turning into reality. Due to adverse conditions such as high noise levels, extreme humidity or temperatures, or even destructions from unfriendly entities, sensor node failures may become norms rather than exceptions in real environments. To be practical, sensor networks must last for much longer times than that of individual nodes, and have yet to be robust against potentially frequent node failures. This paper presents the design of PEAS1, a simple protocol that can build a long-lived sensor network and maintain robust operations using large quantities of economical, short-lived sensor nodes. PEAS extends system functioning time by keeping only a necessary set of sensors working and putting the rest into sleep mode. Sleeping ones wake up now and then, probing the local environment and replacing failed ones. The sleeping periods are self-adjusted dynamically, so as to keep the sensors’ wakeup rate roughly constant, thus adapting to high node densities. PEAS shares the basic technique of turning off nodes with other energy-conserving protocols in wireless ad hoc networks (GAF[4], SPAN[2], AFECA[3]) and sensor networks (ASCENT[1]). However, existing work is intended for much more powerful mobile computers and/or a relatively stable network environment, whereas PEAS design targets at a much harsher or even hostile working environment, where (1) node failures should be considered norms rather than exceptions, (2) due to ad hoc deployment and the need for long-last operations, the nodes’ deployment density can be several times or even a magnitude higher than the minimum required for normal functioning, and (3) nodes are too constrained in memory and computing resources to afford relatively complex protocols. None of the related work addresses the issue of robust operations in the harsh environmental setting that PEAS design has assumed. PEAS achieves highly robust operation by randomizing the sleeping times of sleeping nodes to detect and replace failed nodes actively. It eliminates per-