Trajectory-planning using set-based motion-primitives considering model uncertainty and controller tracking errors

In this paper a trajectory planning approach based on set-based motion primitives is presented. Through the set-based motion primitives model uncertainty and the control error of a trajectory controller, due to imperfect tracking is considered. Due to the linearization approach of the vehicle dynamics, two main improvements compared to the state of the art are achieved. First, the resulting reachset of the controlled vehicle dynamics is linear and therefore more efficient to calculate compared to a nonlinear method. Secondly, the sequential composition condition, which has to be fulfilled when switching between two set-based primitives, can be satisfied efficiently online. This leads to more switching possibilities between the motion primitives during planning, and hence improves robustness of the collision avoidance behaviour in a dynamically changing environment. The trajectory planning approach is successfully demonstrated within an intersection scenario with a dynamic obstacle in a simulation framework.