Tree-Based Machine Learning for Fault Diagnosis in Autonomous Underwater Vehicles

Fault diagnosis in Autonomous Underwater Vehicles (AUVs) is important for ensuring mission success and operational reliability in dynamic marine environments. Recent advancements in machine learning have shown promising results in fault classification. However, these models often come with high computational costs and limited interpretability. This study investigates the effectiveness of tree-based learning models - Decision Tree, Random Forest, XGBoost - for AUV fault diagnosis. Using the publicly available “Haizhe” fault diagnosis dataset, we evaluate the classification performance of these models and compare them to deep learning models such as Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNNs). A Multi-Layer Perceptron (MLP) is also included as a baseline for comparison.