High speed cornering of an off-road vehicle poses considerable challenges to the development of an autonomous vehicle due to the non-linear dynamics of the tyre road interface as well as those of the vehicle as a whole during high lateral accelerations. Most driver models are developed for low speed applications using linear control methods under the assumption of linear vehicle dy- namics. The dynamics of a vehicle however become highly non-linear as the lateral acceleration increases, thus rendering these linear models unusable during high speed manoeuvres. In this study, two robust driver models for use in an autonomous vehicle capable of path following at both low and high speeds are presented. Both models make use of the relationship between the yaw acceleration and steering rate to control the yaw angle of the vehicle. The first driver model is derived from the simulation of a full non-linear vehicle model in ADAMS. The Magic Tyre Formula is used to model the relationship between the vehicle's yaw acceleration and steer rate as a function of vehicle speed. The second driver model is a mathematical model which incorporates a form of sliding control. The model includes the lateral tyre dynamics as modelled by the Pacejka '89 tyre model. Both driver models are coupled with a gain scheduling proportional derivative controller to reduce the cross-track error. The two driver models were implemented on a Land Rover Defender and experimentally validated by performing a double lane change manoeuvre at speeds up to 80km/h. The vehicle remained stable even though the lateral accelerations experienced were 80% of the vehicle limits. The result is a robust controller capable of path following at various speeds and at high lateral accelerations. Copyright
Dissertation (MEng)--University of Pretoria, 2011.