In this paper, we first construct a vehicle kinematics model using longitudinal acceleration and implement real-time ramp angle estimation by Kalman filter algorithm. Subsequently, it is improved to a dynamics estimation method with extended Kalman filter, and the results of the two estimation nodes are data fused by analyzing the shortcomings of the dynamics and vision schemes in order to achieve redundant and complementary results. The average absolute error and root-mean-square error of this paper’s model, VB-EKF, are reduced by at least 50% compared with the comparison model, which reduces the oscillation error of the estimation results when the vehicle enters and exits the ramp; and the estimation accuracy of the VB-EKF model in special working conditions is significantly improved. The real-vehicle test of roadway lateral slope estimation based on VB-EKF shows that the average absolute error and root-mean-square error of the method are reduced by more than 60% compared with the comparison model, and its estimation error is controlled within 2%, with better robustness and stability of the model. This paper provides a new and more accurate method for dynamic slope detection and automatic tracking and positioning accuracy estimation.
