ADAPTIVEDISTRIBUTED ALGORITHMSFORPOWER-EFFICIENT DATAGATHERING
Baltasar Beferull‐Lozano · 2005
Inthis work, weconsider theproblem ofdesigning adaptive distributed processing algorithms inlarge sensor networks that areefficient intermsofminimizing thetotal power spent forgathering thespatially correlated datafromthe sensor nodestoasinknode.Wetakeinto account both thepowerspent forpurposes ofcommunication aswellas thepowerspent forlocal computation. Ourdistributed algorithms arealsomatched tothenature ofthecorrelated field, namely, forpiecewise smooth signals, weprovide two distributed multiresolution wavelet-based algorithms, while forcorrelated Gaussian fields, weusedistributed prediction based processing. Inbothcases, weprovide distributed algorithms that perform network division into groups ofdifferent sizes. Thedistribution ofthegroup sizes within the network istheresult ofanoptimal trade-off between the local communication inside eachgroup needed toperform decorrelation, thecommunication needed tobring theprocessed data (coefficients) tothesink andthelocal computation cost, which growsasthenetwork becomes larger. Our experimental results showclearly thatimportant gains in powerconsumption canbeobtained with respect tothecase ofnotperforming anydistributed decorrelating processing.