Accurate modeling of underwater vehicle dynamics is essential for simulation, stability analysis, and control system design. In this context, hydrodynamic coefficient identification plays a crucial role due to its dependence on vehicle geometry, forward speed, and operating conditions. In this study, a Robust Extended Kalman Filter (REKF) is employed for the online estimation of key hydrodynamic coefficients in the diving dynamics model of the REMUS autonomous underwater vehicle. An augmented state vector formulation is used to simultaneously estimate the dynamic states q, θ, and z, together with the hydrodynamic coefficients M_q, M_q ̇ , and M_(δ_s ). To evaluate the influence of operating conditions, three scenarios are considered with forward speeds of 1.5, 1.75, and 2 m/s and rudder angles of 5.5°, 7.5°, and 10°. The Kalman gain K_kis recursively updated at each time step based on the prediction error covariance, measurement matrix, and measurement noise covariance, following the standard REKF formulation. Simulation results indicate that the proposed method achieves stable estimation of hydrodynamic coefficients and effectively reproduces the diving dynamics of the REMUS vehicle across all scenarios. Furthermore, increased forward speed and rudder deflection enhance system excitation, leading to improved estimation accuracy and faster convergence behavior within the considered operating range. Finally, the comparison between the responses obtained using estimated and reference coefficients demonstrates the capability of the proposed framework for accurate system identification and potential support for control design under varying operational conditions.