Optimizing Underwater Passive Maneuvering Target Tracking With Innovative Radial Basis Function Deep Neural Paradigm

Tracking underwater passive maneuvering targets poses significant challenges due to high maneuverability and noisy ocean environments, particularly when target follows large and sharp turns continuously. This study presents a novel approach to improve the accuracy of tracking such targets by employing the potential of radial basis function (RBF) deep neural network. The proposed method integrates state–space dynamic modeling and measurement error design, employing low and high numerical values of standard deviations of Gaussian measured noise. By leveraging RBF neural network, the passive maneuvering target tracking system computes and adjusts the distance of hidden neurons through weight adaptation, thus enhancing the accuracy of trajectory, position and velocity estimation of target. The proposed model effectively reduces the root-mean-square error of target's dynamic features, particularly under complex maneuvering conditions. Simulation findings demonstrate improvements in trajectory tracking, turning rate graphs, position, velocity, and error distribution analysis. The performance of RBF deep neural paradigm is compared with state-of-the-art generalized pseudo-Bayesian tracking algorithm, such as interacting multimodal unscented Kalman filter as well as existing neural computing techniques, such as scaled conjugate gradient neural intelligence and intelligent Bayesian regularization backpropagation neurocomputing. The findings suggest that proposed RBF-based neural paradigm offers a robust solution for passive target tracking in noisy underwater environments, enhancing overall performance in challenging scenarios.