Modeling TimeSynchronization Uncertainty Sources inWireless Sensor Networks

This paper deals with athorough analysis oftheuncer- II.MODEL DESCRIPTION tainty sources affecting theinter-node timesynchronization inWire- less Sensor Networks (WSNs). Suchananalysis relies onastochasticTheuncertainty sources affecting thetimesynchronization model describing thevarious uncertainty contributions aswell ason betweenWSN nodescanberoughly grouped intotwomain themainfeatures ofreal WSNnodearchitectures. Someexperimen-categories: talresults confirm thecorrectness ofthetheoretical approach. The .Clock-related contributions, i.e. duetomanufacturing tol- achieved results will enable theimplementation ofasimulator that erances, aging andchanging operating conditions ofthe could beusedtocompare theaccuracy performances ofdifferent syn- oscillators embedded onthenodes; chronization protocols. *Inter-node communication latencies occurring whentime- stamped packets areexchanged between pairs ofnodes. Wireless SensorNetworks (WSNs),timesynchronization, Inthefollowing, both sourcesofuncertainty aredescribed ac- Keywords cording toproper mathematical models. measurement uncertainty, communication latency. ~~~~~~~~~~more accurate time measurements, anexternal, i.e. on-board, collected bydifferent nodes tend todrift significantly fromone crystal oscillator (XO)oflower resolution andhigher stabil- another (2), (3). Thisinturnmaycausemisleading interpre- ity canbeusedasapossible timereference toclock theMCU tations about thedatasensed bydifferent nodes, orevencom- munication problems (e.g. whenvarious sleeping nodes wake timers. Generally, therandomprocess f(t) modeling thein- upatdifferent times). Inorder toassure aproper inter-node stantaneous frequency ofanoscillator isdefined by(12): synchronization, manydifferent protocols havebeenproposed f(t) = fo(1+vs +pt+vr(t))