Open AccessNeural Network Based Model for Friction Potential Estimation under Longitudinal and Lateral Excitations
Author(s) -
Smiljana Todorovic,
Andreas Wagner,
Sven Müller,
Jens Neubeck
Publication year - 2022
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2234/1/012005
Subject(s) - artificial neural network , work (physics) , friction coefficient , estimation , automation , computer science , relation (database) , engineering , artificial intelligence , data mining , materials science , mechanical engineering , systems engineering , composite material
Information about the maximum tire-road friction coefficient secures safety, increases the performance of ADAS functions, and enables higher levels of automation. An important parameter representing the relation between maximal longitudinal or lateral forces and normal force is friction potential. This paper presents a new approach for friction potential estimation based on Neural Networks by using data of 4WD vehicle. The regression model built in this paper can estimate the friction potential under both, longitudinal and lateral excitations. The goal was to build an estimation model, which does not require a tire model and can work under different types of excitations. Furthermore, this paper confirms, that the neural network approach requires lower excitations for friction potential estimation in comparison to traditional approaches. Data for training and validation was collected using a research vehicle called FlexCar, which was designed, built, and automated in the project with the same name. Experiments were conducted on three different friction levels.