Development of a path tracking controller for an autonomous vehicle

Abstract

Autonomous driving has been a fast-growing industry in the last decade with the human safety being the main concern. The improved processing capabilities of nowadays computers helped the rapid growth of such industry. In this work a computationally demanding algorithm –Nonlinear Model Predictive Control- is used to develop a path tracking controller that is capable of driving a golf kart by controlling the steering, acceleration and braking while paying attention to the passengers’ comfort. A comparison between two different NMPC schemes for path tracking is also carried out to develop an efficient controller. The controller developed is then validated in Carsim-Simulink environment. Finally the controller is evaluated against Stanley controller to show which is better suited for the task. From the results the NMPC is found to be better suited for the path tracking task as it can handle the constraints on both states and controls while following the reference track and velocities at the same time.