Sideslip angle is one of the most important state for autonomous vehicle control. Based on new sensor configuration in autonomous vehicle, a novel method to estimate vehicle sideslip angle is proposed. A Kalman filter is adopted to estimate vehicle yaw rate measurement bias from the IMU with the GPS course angle information. By using the lateral distance information at the camera preview point, the kinematic relationship between the vehicle state and the lane line is built. The kinematic relationship between acceleration measurements by the IMU and velocity components at the point where the GPS is located are established. A Luenberger-like nonlinear observer is proposed to estimate the sideslip angle by fusing the above vehicle kinematic models. The observability and convergence of the observer are analyzed. The effectiveness and robustness of the algorithm are verified through the slalom and double lane change maneuvers on high and low friction road.