Gaussian–Student’s t Heterogeneous Mixture Kernel Maximum Correntropy Adaptive Filtering for Cooperative Localization Under DVL Faults and Outliers

This paper addresses the challenge of cooperative localization (CL) in multi-AUVs (Autonomous Underwater Vehicles) systems under Doppler velocity log (DVL) faults and outliers, aiming to enhance the reliability of underwater Internet of Things (IoT) systems. We propose a Gaussian-Student’s t heterogeneous mixture (GStHM) kernel to enhance traditional correntropy, improving the robustness of signal processing in non-Gaussian noise conditions. Within the GStHM kernel correntropy framework, we reconstruct the cost function and develop a linear regression model that explicitly accounts for outliers and DVL failures. Furthermore, a GStHM kernel maximum correntropy based improved augmented extended Kalman filtering (GStHM-MCIAEKF) algorithm is proposed for multi-AUVs CL system. It provides real-time compensation for long-term DVL-fault AUV velocity and handles outliers caused by system and environmental. Moreover, the proposed method can dynamically adjust bandwidth parameter and mixture parameter in response to varying noise conditions, enhancing the adaptability and precision of positioning. Real lake trial demonstrates the algorithm’s effectiveness and superiority in handling DVL faults and non-Gaussian noise.