NEURAL-NETWORK REPRESENTATION OF TYRE CHARACTERISTICS - THE NEURO-TYRE

One of the most difficult aspects of vehicle modelling is the accurate description of tyre characteristics. A considerable amount of work has been done in the area of tyre modelling using a range of approaches, from pure theoretical modelling to fully empirical formulae fitted to measured data. In this paper, a new approach to create non-linear mapping between the input and output characteristics of tyre using Artificial Neural Networks (ANNs) is examined. A comparison is made between the results obtained using ANNs and those obtained using Pacejka's 'Magic Formula', developed in order to characterize a tyre cornering force. The applicability of ANNs to the combined braking and cornering characteristics of a tyre is also investigated, and a relatively simple way to model this complicated phenomenon is offered. In this sense, some basic problems concerning ANNs application are discussed. In addition, in order to demonstrate its useability, the developed neural network-based tyre model is used in a simulation of the directional behaviour of a straight-truck. Because the neural network performed well during the test, the authors conclude that it can be a competitive and accurate way to model a tyre for vehicle simulation purposes. The neural network representation of tyre characteristics is a first step in an ongoing project where the aim is to examine possible applications of the artificial neural network to vehicle system dynamics and control. The authors named the first model Neuro-Tyre.