Robust state estimation and fault detection for Autonomous Underwater Vehicles considering hydrodynamic effects

Abstract

Autonomous Underwater Vehicles (AUVs) are designed to operate in complex external environments, making fault detection crucial to overall safety. While a general kinetic model can be applied to describe the dynamics of an AUV, obtaining accurate parameters for the dynamic model is challenging due to the complexity of the outline structure of the AUV and external influence factors. In addition, environmental influence on model parameters is usually time-varying and stochastic, which is different from simple model uncertainties. In this paper, we model external perturbations as stochastic model uncertainties that share some characteristics with noise. We also apply an improved robust filtering method to estimate system states and alleviate the effects of stochastic system uncertainties. Based on the proposed filtering approach, a residual signal is computed based on reconstruction errors of the observations and smoothed using a sliding-window technique, which is directly applied to the fault detection problem. Finally, simulation experiments demonstrate the effectiveness of the proposed method in detecting actuator faults in several cases.