Cone Slalom with Automated Sports Car – Trajectory Planning Algorithm

Abstract

In this paper, we present the system architecture and algorithms of an automated vehicle to perform a slalom. We demonstrate a novel trajectory planning algorithm based on optimization techniques using logic-based Benders de-composition, where an external loop optimizes the position of the nearest waypoint and an internal loop generates the optimal trajectory. The positions of the cones in this use case are unknown, but a mono camera and LiDAR detect them. They can be in a line or dispersed, have equal or unequal spacing, and the U-turns can be symmetric or asymmetric. A bicycle model is used to formulate a non-linear quadratic optimization problem aimed at optimal trajectory generation considering vehicle kinematics. Finally, the trajectory tracking control keeps the vehicle on the planned slalom trajectory while driving. The control system is interfaced with the vehicle via CAN and FlexRay buses. Much of the work was devoted to experiments with a real vehicle and fine-tuning the system parameters. During the validation of the system, interesting observations were made regarding the components' precision, frequency, and sensitivity.