Eco-driving and road curvature estimation: Retrospective analysis of experimental data of a fully electric coach

This paper extends the eco-driving optimal control problem, a single-vehicle driver assistance system that provides speed recommendations to help drivers operate their vehicles more energy efficiently. The state-of-the-art model is improved by incorporating road curvature as an additional input variable, alongside elevation and drivetrain dynamics. The framework is enhanced to account for centripetal forces and vehicle slip constraints, resulting in a more realistic representation of vehicle behavior. Road curvature is estimated using the osculating circles method and required derivatives are approximated using the method of undetermined coefficients. The proposed method is validated against human driving data obtained in free flowing traffic, showing that the velocity recommendations in and around corners are more realistic than those from eco-driving methods that neglect cornering. Furthermore, the cornering-aware eco-driving approach is more energy-efficient, achieving energy savings of up to 6% in specific corners and 25% overall compared to uninstructed human drivers. Finally, it provides significantly more accurate energy consumption estimates than the case that omit cornering effects.