Robust Coarse-to-Fine 3-D-Target-Localization Algorithm for Underwater-IoT-Based Networks: Design and Performance Evaluation Under Uncertain Multiparameters

Underwater Acoustic Internet of Things Networks (UAIoTNs) can furnish excellent technical support and information services for applications involving marine observation and detection, marine disaster prevention and mitigation, and maritime search and rescue, in which accurate positioning information is the fundamental requirement. The combination of high dynamics and complexity of the ocean environment to the high latency and narrowband of underwater acoustic communication are complex challenges in UAIoTNs. Due to these facts, this work investigates the received signal strength (RSS)-based three-dimensional (3D) target localization in UAIoTNs taking into account the absorption effect, uncertain transmission power (UTP), and a time-varying Path Loss Exponent (PLE). Through Taylor’s first-order expansion and certain approximations, we envision the underwater stratified acoustic propagation localization challenge as an Alternating Non-negative Constrained Least Squares (ANCLS) framework. To address the challenges posed by unknown multi-parameters, a robust coarse-to-fine localization algorithm (RCFLA) is proposed. At first, the coarse localization phase utilizes the Active Set Method (ASM), while the subsequent fine localization one employs the improved Broyden-Fletcher-Goldfarb-Sanno (BFGS) trust region method to enhance convergence towards the global optimal solution. The iterative process refines the underwater target location, UTP, and PLE, using the ASM-derived rough solution as the initial estimate. Analysis of computational complexity and derivation of the Cramér-Rao Lower Bound (CRLB) with stratified propagation and absorption effect demonstrates the superiority of RCFLA. Furthermore, Lyapunov’s second stability theorem is used to prove the stability of the RCFLA and presents a complete proof of global convergence. Numerical simulation and experimental results validate the algorithm’s optimal localization accuracy across various scenarios, showing reduced overhead compared to benchmark algorithms.