Robust Constraint Satisfaction for C/GMRES bassed on NMPC with Parameter Uncertainties

Automated driving is an emerging technology in the automotive industry. Steady development of assistance systems relieves and supports the driver in a multitude of traffic situations. Next to the advantage of increased comfort,the safety aspect plays a significant role. Current traffic statistics show that 90% of all traffic accidents are caused by human error and underline the importance of pushing research in this field even further. In this work a Highway Pilot is proposed. As an introduction Adaptive Cruise Control (ACC) is described and its problems with changing lanes are discussed. The idea of the Highway Pilot is to expand the longitudinal control with a transverse component (i.e. lane change). Based on a derived mathematical vehicle model two approaches to solve the problem are proposed that allow time, acceleration and/or jerk optimal driving through traffic. In order to manage computational burden, several simplifying assumptions are made, including modelling lane changes in a first step as discrete jumps. Several cost functions are analyzed and the computation times of both approaches are compared. In a second step, the former discrete lane jumps are transformed into drivable trajectories. Next to clothoids, which allow a jerk free connection between straight lines and curves, as used in road and rail-road construction, n3-splines are proposed. These allow smooth interpolation between start and end point with respect to the tangential vector, curvature and change of curvature. Furthermore, it is shown that the lateral trajectory can be solved by means of a Quadratic Program (QP).