This paper proposes an actuator fault detection based on unknown input observer for simultaneously state estimation and fault estimation applied to lateral vehicle dynamics. The lateral dynamics is modeled by a linear parameter varying model where the longitudinal velocity is considered as parameter varying. The observers gains are based on the resolution of linear matrix inequalities. The observers convergence is analyzed by using the Lyapunov theory. The proposed fault detection system is finally applied to the detection and estimation of steering angle fault and the estimation of the yaw velocity, yaw angle and the trajectory of the vehicle. Simulation study verifies the effectiveness of the proposed approach.