Unscented Kalman Filter Based Single Beacon Underwater Localization with Unknown Effective Sound Velocity

The extended Kalman filter (EKF) is the most widely applied state estimation algorithm for single beacon underwater localization systems. However, since the assumptions of local linearity are often violated during underwater surveys, the EKF is difficult to tune and often gives unreliable estimates. Compared with the EKF, the unscented Kalman filter (UKF) can achieve improved estimation performance for nonlinear systems, while its computational efforts is the same order of magnitude as the EKF. This paper investigates the estimator performance and computational complexity while implementing the UKF based on a novel single beacon localization model with unknown effective sound velocity. Furthermore, the estimator performance while using the iterated EKF, which is an alternative way to reduce the linearization error of the EKF, is also investigated. In addition, the Rauch-Tung-Striebel (RTS) smoother is applied for postprocessing. Through numerical examples of using simulated data, both the filter and RTS smoother results show that while implementing the UKF, the localization accuracy, as well as the estimation of effective sound velocity, can be improved.