Trajectory Following of an Autonomous Ground Vehicle

Researchers: Nicole Chan, Eric Westman

Abstract: While autonomous vehicles are quickly making their way into the market, a major hurdle that remains is the inability to drive comfortably like humans do. Much of this discomfort is a result of rapid corrections and adjustments to the steering of the vehicle made by the steering controller. Research has been conducted on limiting the velocity of the vehicle based on turning angle using data gathered from human drivers. However, the steering controllers remain at the core of the issue of passenger comfort. In this paper, we present data gathered from an autonomous vehicle testbed, the CATVehicle, that implements a widely used steering controller. This data is used to create a state-space model of the steering controller’s behavior. Based on this model, various techniques were investigated and applied to improve the smoothness of the steering output. The result is a balance between following closely to the desired path and achieving longer, continuous steering adjustments (rather than many short bursts of corrections). Our analysis and results are demonstrated in simulation as well on the CATVehicle testbed.

Keywords: Control systems, simulation, optimal control, trajectory following