Abstract: The hydraulic actuator in this work is an unobservable nonlinear system， for which a normal observer is infeasible. The extended Kalman filter (EKF) was thus adopted to estimate pressures in piston and rod chambers of the cylinder. The 4th order nonlinear statespace model was developed and its parameters were obtained by the GaussNewton method. The state of the hydraulic actuator was estimated using joint angle based on the EKF. The simulation results indicate that the proposed method can estimate the cylinder pressures accurately. The experimental results show that the estimated joint torque is close to the measured one， which verifies the validity of estimating hydraulic states by the EKF. A reference for statebased fault detection and control of a hydraulic actuator is thus provided.

Key words: hydraulic actuator, nonlinear system, extended Kalman filter, state estimation, GaussNewton method