Variable-speed autonomous path tracking of a vehicle via robust linear model predictive controller

A Model-predictive-controller (MPC) can deal with path-tracking problems in autonomous vehicles due to its predictive ability. However, changes in vehicle speed have a significant impact on the linear controller performance. In this study, a novel robust linear MPC (RL-MPC) is proposed to obtain the required front wheel steering angle to keep a vehicle on reference trajectories in the NATO double-lane-change maneuver. The proposed controller is based on a one-track, 2-DOF linear vehicle model, and controller parameters are tuned to ensure successful trajectory tracking at different vehicle speeds. The tuning process is handled practically by artificial neural network (ANN) structure. The simulated results revealed that the proposed RL-MPC can improve the path-following performance, and provide flexibility to operate at different vehicle speeds, compared to the linear-quadratic-integral (LQI) and I controller (LQI+I) set.