A Hybrid Localization Algorithm for Underwater Nodes Based on Neural Network and Mobility Prediction
W. Pu, Wei Zhu, Yang Qiu
Abstract
Due to the harsh marine environment, precise node positioning has become a great challenge for underwater wireless sensor networks (UWSNs), as various errors (e.g., ranging errors) may be induced in positioning. Thus, many localization algorithms (a.k.a. positioning algorithms) have been proposed for UWSNs, but most of them exhibit limitations in accuracy due to the difficulty of complicated system modeling when considering heterogeneous errors. In this article, we propose a hybrid localization algorithm based on convolutional neural network (CNN) and mobility prediction (HLCM) when considering various kinds of errors in positioning. Different from previous location algorithms, the proposed HLCM algorithm trains a CNN-based localization model to establish the positional relationship among underwater environmental factors, anchor nodes, and ordinary nodes, which aims to alleviate the uncertainties caused by variable sound speed, with reduced various errors in the ranging process, and thus enhance localization accuracy. Besides, the proposed HLCM algorithm considers the node drifting induced by the ocean current during its positioning and predicts ordinary nodes speeds via weighted superposition of anchor nodes speeds, which helps compensate for the positional deviation generated in positioning. The simulation results and comparative analysis indicate that the proposed algorithm obtains high localization accuracy and extensive localization coverage with high fault tolerance.