Road longitudinal slope estimation based on the fusion of machine vision and vehicle dynamics

In the case of unknown road information, estimating the slope by the dynamic modeling method is difficult because it is difficult to get relevant driving data before the vehicle enters the ramp. Therefore, this paper provides a solution to improve the accuracy of road slope estimation that integrates vehicle dynamics and vision. Firstly, a model is constructed for the longitudinal dynamics of the vehicle, and the road slope is estimated by a fuzzy adaptive unscented Kalman filter. Subsequently, a road slope estimation method utilizing both machine vision and Kalman filter techniques has been devised, effectively mitigating the issue of significant initialization errors impacting estimation precision. Ultimately, both simulation and actual vehicle trials confirmed the proposed approach’s estimation precision and robustness.