A variety of control concepts for autonomous driving have been developed during the past years. Beginning with linear lateral guidance and velocity controllers, nowadays nonlinear control designs have been applied several times successfully to real world examples under research conditions. The most promising approach to this is model predictive control. This control method allows the explicit handling of coupling in longitudinal and lateral dynamics of the vehicle.
One of the main difficulties of this approach is the high reliance of the control accuracy and recursive feasibility on accurate prediction models. The great difficulty is the inevitable uncertainty in the dynamics, resulting mostly from the tire models, the vehicle mass, and external disturbances like wind. To employ model predictive control in road conditions, they need appropriate robustification. This includes the robustness of constraint satisfaction which has to be ensured in traffic conditions.