An unmanned aerial vehicle-aided node localization using an efficient multilayer perceptron neural network in wireless sensor networks

Abstract

Localization of sensor node is decisive for many localization-based scenarios of wireless sensor networks (WSNs). Node localization using fixed terrestrial anchor nodes (ANs) equipped with global positioning system (GPS) modules suffers from high deployment cost and poor localization accuracy, because the terrestrial AN propagates signals to the unknown nodes (UNs) through unreliable ground-to-ground channel. However, the ANs deployed in unmanned aerial vehicles (UAVs) with a single GPS module communicate over reliable air-to-ground channel, where almost clear line-of-sight path exists. Thus, the localization accuracy and deployment cost are better with aerial anchors than terrestrial anchors. However, still the nonlinear distortions imposed in propagation channel limit the performance of classical RSSI and least square localization schemes. So, the neural network (NN) models can become good alternative for node localization under such nonlinear conditions as they can do complex nonlinear mapping between input and output. Since the multilayer perceptron (MLP) is a robust tool in the assembly of NNs, MLP-based localization scheme is proposed for UN localization in UAV-aided WSNs. The detailed simulation analysis provided in this paper prefers the MLP localization scheme as they exhibit improved localization accuracy and deployment cost.