LSTM-DL: A Counting Optimization Approach for LSTM-Based DV-Hop Localization Model in Wireless Sensor Networks

Abstract

The quest for precise localization of unknown nodes in Wireless Sensor Networks (WSN) garners significant academic attention, further intensified by the burgeoning domain of meta-heuristic algorithms. Traditional localization methods, such as the Least Squares Method (LSM), inherently yield lower accuracy, a limitation that this paper aims to transcend. We introduce a robust deep learning model grounded in Long Short-Term Memory (LSTM-DL) designed to enhance localization processes by effectively minimizing distance measurement errors in network topologies. This approach is augmented with an improved evaluation function, meticulously tailored to mitigate topological inaccuracies. Through rigorous experimental validation, the LSTM-DL model demonstrates a remarkable performance leap, surpassing the conventional DV-Hop method and outperforming existing meta-heuristic algorithms in accuracy and reliability. The findings decisively suggest that meta-heuristic approaches, and specifically the advanced LSTM-DL, hold substantial promise over traditional DV-Hop in the sphere of node localization within WSNs.