Intelligent Emergency Stop System for Autonomous Ground Vehicles Using Minimum Jerk Trajectories and Path Tracking Linear Quadratic Regulator

Abstract

Autonomous vehicle technology is moving towards developing systems that can function without human operators. One method of improving safety in autonomous vehicles is to include redundant sensors and systems to protect against the adverse effects of hardware failure during operation. This thesis presents an emergency vehicle collision avoidance system that uses a linear quadratic regulator (LQR) to track a minimum-jerk trajectory. This system is designed to be implemented on a low-cost processor that can act as a backup to the main computing hardware on these vehicles. The main goal of these algorithms is to bring a vehicle to a stop while avoiding collisions and maintaining safe operation.

The lateral control systems on the vehicle consist of a minimum-jerk trajectory generator and an LQR path tracking controller. Minimum-jerk trajectories can be used to generate reference states for the LQR path tracker that are more dynamically conservative than other spline interpolation methods, but these trajectories do not inherently account for the dynamic limitations of ground vehicles. Improvements to the dynamic feasibility of these trajectories are made by including an intelligent velocity reference system that reduces vehicle speeds before high dynamic lateral maneuvers. Simulations were performed in Matlab to validate the performance of the lateral and longitudinal control systems. These simulations were performed using randomized vehicle model parameters in order to analyze how the algorithms perform in the absence of accurate parametric data for the vehicle. The control algorithms were also tested in live experiments using low-cost computing hardware on a Lincoln MKZ sedan, demonstrating safe operation at various speeds with lateral path tracking errors of less than 15 cm.