Modelling of tyres can be achieved by various methods ranging from purely theoretical to fully empirical fitting of curves to measured points. In previous work, the application of Artificial Neural Networks (ANNs) to modelling the steady-state cornering characteristicof a tyre was investigated and reported, from which a 'Neuro-Tyre' was developed. In this paper, ANNs were applied to create a non-linear mapping between the dynamic as well as steady-state input and output characteristics of a tyre, most notably with respect to the dynamic cornering characteristic of a tyre running on an uneven road surface. The resultant lateral forces are compared to those given by the extended version of the Pacejka's Magic Formula. The behaviour of this dynamic 'Neuro-Tyre' is also investigated by incorporating it in a computer simulation model of a single vehicle subjected to steering manoeuvre while transmitting a rough road surface. The results show that ANNs can be used to generate dynamic tyre models with reasonable accuracy and high computational efficiency, which portends their use for vehicle dynamics simulation. (A)
Abstract