Comparative Study of Modeling Road-Vehicle Dynamic Behaviour Using Different Data Based Techniques

The current work provides data based modeling of passive automotive suspension dynamics of a half-car model, by implementing different machine learning techniques, which model the passive dynamics. The passive vertical and pitch suspension dynamics, using vector auto-regressive with exogenous input (VARX), recurrent neural network (RNN), and artificial neural network (ANN), and their accuracy will be compared. The training data is obtained from a half car model simulated by Matlab/Simulink which intended to mimic suspension sensors readings. It is found that the VARX and ANN models showed very high identification accuracy, added to the superior computational efficiency realized by the VARX technique.