Identification of an Autonomous Underwater Vehicle hydrodynamic model using three Kalman filters.

The controllability and maneuverability of an Autonomous Underwater Vehicle (AUV) in practical applications need to be properly validated and assessed before the prototype is finalized for manufacturing. With regard to mathematical system model of the AUV, hydrodynamic coefficients have a dominant effect on the quality of vehicle pre-testing and evaluation, which is crucial to be estimated with adequate accuracy to curb the uncertainty from modeling simplifications. The standard time domain discrete Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) algorithms are two promising numerical approximation approaches for hydrodynamic identification technique with favourable computational complexity and acceptable estimation precision. In this paper, a study with respect to time domain discrete optimized UKF (OUKF) algorithm based on recursive tuning rule and update gradient descent lemma for a typical prototype known as NPS AUV II will be proposed to enhance adaptability and prediction performance of the identification approach with appropriate verification. In addition, Auto Regressive Moving Average (ARMA) noisy model needs to be included into three Kalman Filter (KF) algorithms to further improve the estimation precision and perturbations inhibiting performance. Pre-processing numerical model validation and non-dimensional viscous linear damping coefficient identification based on three KF algorithms will be implemented respectively to provide an assistant pre-assessment for the vehicle. In accordance with comparable outputs and estimation-experiment errors, the OUKF identification algorithm is certified to be more precise and superior compared with EKF and UKF approaches in the presence of ARMA noisy model. • Augmented system model for typical prototype NPS AUV II will be mathematically hypothesized and demonstrated. • Single estimation circle of Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) will be explained in detail. • Optimized UKF (OUKF) approach is proposed to elevate the adaptability and prediction performance of identification process. • ARMA noisy model is included in Kalman Filter (KF) to improve estimation precision and perturbations inhibiting performance. • The output of validations and identifications based on three KFs and ARMA model with sensitivity analysis will be presented. [ABSTRACT FROM AUTHOR]