This paper discusses the impact of changes in driving dynamic quantities on the lateral driving quality of a vehicle when its lane keeping assistance (LKA) is activated. Moreover, this paper will pay attention to changes in driving accuracy when implementing a 2-degree-of-freedom (DOF) steering model in the simulation. Attention will be paid on the design and implementation this model. A detailed description of the used approach will be given and the system will be assessed by a handling test. All simulations are carried out with activated lane keeping system, while the changes in driving dynamic quantities concerns the wheelbase, vehicle mass and tire oblique stiffness. Simulations are carried out with a defined driving maneuver and defined road that corresponds to a real one.

The quality of the vehicles lateral movement will be measured by key performance indicators (KPI). The maximum overshoot in lateral distance to center line for a trajectory return will be considered. All results are compared with a simulation without any LKA usage and no implementation of the steering system model. In this basic scenario only the vehicle dynamics which can be attributed to the linear bicycle model does play a role. Comparing the simulation results with activated LKA to the basic scenario a higher deviation of the KPI could be found. Besides, higher maximum values for the KPI appear. When implementing the steering system model and activating the LKA even higher overshoots and deviations of the KPI could be found.

The variation of the vehicle's wheelbase shows the highest impact on the quality of lateral control. Miner impacts could be shown when variate vehicle mass and oblique stiffness. A two-parameter variation is carried out that does show no unexpected behavior when changing vehicle mass and tire oblique stiffens at the same time. Moreover, it shows that changes in vehicle mass have higher impacts when running simulations with small tire oblique stiffens. In addition, when applying small values for the tire oblique stiffens while the steering model is activated the previously seen higher deviation in the KPI do disappear.

Mr. Julian Staiger, TU Delft, NETHERLANDS