A nonlinear model predictive controller for autonomous driving

Abstract

Self driving cars have been the focus of several research groups due to the improvement in processing capabilities, and the feasibility to run sophisticated control algorithms in real time. We introduce two NMPC implementations able to guide the vehicle autonomously through acceleration and steering to track a predefined path with reference velocities accurately while paying attention to the comfort of the passengers and moreover capable of running in real time. The two controllers are compared in terms of control effort, accuracy and iteration times, Carsim is then used to show the ability of the controller to drive the vehicle.