Motion estimation of an autonomous surface craft using an unscented Kalman filter

This paper addresses the problem of navigation algorithms in the context of high accuracy motion control for a unmanned surface crafts. The effectiveness of an unscented Kalman filter algorithm in combination with a detailed nonlinear 3DOF dynamical model for motion estimation is shown. Water currents are regarded as slowly varying parameters estimated by the filter which has been augmented by two states for that purpose. The performance is demonstrated by comparing two control tasks in a real world environment with and without motion filtering.