A Novel TOA-Based Adaptive Multi-AUVs Cooperative Localization Method Under Unknown NLOS/LOS Errors and Outliers

Underwater positioning accuracy is crucial for underwater Internet of Things (IoT) systems. When autonomous underwater vehicles (AUVs) encounter obstacles such as reefs or pipelines during navigation, the signal propagation path is blocked, causing nonline-of-sight (NLOS) errors. The asymmetric pulse characteristics of these NLOS errors, when mixed with the Gaussian noise of line-of-sight (LOS) conditions, significantly degrade the accuracy of traditional positioning algorithms. To address the challenges of dynamic LOS/NLOS interference switching and unknown time-varying noise statistics, this article proposes a novel multi-AUVs cooperative localization (CL) method. First, an inverse WishartGaussian-Skewed Laplace (IWMGSL) mixture distribution model is designed to describe the measurement noise. This model’s multimodal structure captures the Gaussian characteristics of LOS environments and the skewed heavy-tailed characteristics of NLOS interference. Second, a dual-noise collaborative modeling framework is developed to more accurately capture system and environmental uncertainties. Based on this, an improved robust cubature Kalman filter (CKF) for multi-AUV CL is proposed. This method combines variational Bayesian inference and robust estimation techniques, enabling real-time estimation of unknown noise covariances and automatic balancing of confidence between predicted and measured values according to environmental changes. Lake trial results demonstrate that the proposed method significantly enhances the antijamming capability and stability of navigation systems in complex underwater environments, offering a new solution for high-precision positioning in underwater IoT systems.