This paper addresses the problem of estimating the state of a nonlinear system from measurements that are perturbed by a random source of noise. The Extended Kalman Filter is a type of all-purpose filter that tries to solve this problem by dealing with a linearized version of the system. A new methodology proposed in [1], named Unscented Kalman Filter, is presented. It uses the so-called unscented transformation to better describe the stochastic evolution of the state of the system. The aim of this paper is to compare and discuss the performance of each filter when applied to state estimation of a simplified model of the DELMAC autonomous surface craft.