An Efficient 3-Dimensional Node Localization Using Recurrent Neural Networks in Unmanned Aerial Vehicle-Assisted Wireless Networks

Accurate localization of nodes is essential for various applications, such as environmental monitoring, disaster management, and asset tracking. A novel approach to achieve precise and efficient node localization in wireless sensor networks (WSNs) is by utilizing unmanned aerial vehicles (UAVs). These UAVs facilitate data exchange with nodes through stable air-to-ground communication links. Traditional localization techniques based on received signal strength (RSS) measure wireless signal strength but are significantly affected by environmental factors like obstacles, multipath interference, and signal attenuation. In contrast, neural network (NN)-based methods can learn complex patterns in the received signals, enabling them to handle these variations and deliver more reliable localization results effectively. Among NNs, the multilayer perceptron (MLP) and recurrent neural network (RNN) models are particularly popular due to their versatility and effectiveness. Moreover, RNNs have an advantage over MLPs by incorporating previous network outputs during training. The simulation results in this paper demonstrate that UAV-assisted node localization using NNs achieves superior accuracy compared to existing methods.