The purpose of the paper is to describe an auto- matic estimation method of an underwater vehicle trajectory from ultrashort baseline system measurements. The standard Viterbi algorithm is then used to find the most likely trajectory of the dynamic system observed through the noisy measure- ments with a model of nonconstant speed as state variable. A decoupled trajectory is achieved by considering separately each coordinate of the point. Aberrant measurements are au- tomatically discarded. For very long trajectory, a temporal window is shifted along the data set, and an automatic adjust- ment of the algorithm parameter is performed.