Performance evaluation of region estimation with reactive routing in wireless sensor networks

Abstract

Sensor networks with mobile nodes often employ reactive routing strategies to transfer events to the gateway of the network. Estimating the region of presence of a GPS-less mobile sensor node is non-trivial as infrastructural assumptions for the network are time-dependent. Our contribution has two parts. Firstly, we propose an anchor node placement algorithm, ANCHREG, which establishes the theoretical basis for maximising region estimation performance with reactive routing in sensor networks. Region of a mobile node is estimated by detection if the region is anchored and prediction if the region is unanchored. ANCHREG calculates the minimum number of anchor nodes required and generates their optimal deployment configuration to maximise region estimation accuracy. Secondly, we simulate reactive routing over optimal ANCHREG-generated deployment configurations and compare the performance of region prediction for mobile nodes with machine learning models, namely - artificial neural networks ANN and support vector machines SVM.