An extended Kalman filter is used to tackle the problem of parameter identification of an autonomous underwater vehicle (AUV). Since a large number of parameters must be identified, two alternative sub-approaches are investigated. A classical approach seems to have a better performance in cases where a specific parameter will be identified. However, a modified approach guarantees a better identification when dealing with multiple parameters. A sway/yaw mathematical model for an AUV is obtained considering motions in the horizontal plane. Circular and zig-zag maneuvers are tested to assess the identification approaches employed. Results are presented and analysed. © 2009 IFAC.