In this work, two cooperative navigation solutions based on the extended Kalman filter are described. One of these is a centralized solution and the other is fully decentralized, taking full advantage of the benefits that come with decentralization, such as scalability and robustness. Simulations are performed for a formation of autonomous underwater vehicles with a fixed measurement topology. The vehicles are assumed to be equipped with sensors that allow them to take measurements of their depth and bearing angles to their neighbors. Two different topologies are considered, one acyclical and one cyclical. The transient and steady-state errors of both solutions are analyzed resorting to Monte Carlo simulations. In particular, the mean error and the root-mean-squared-error (RMSE) of the navigation estimates is presented.