Robust MPC Based Lateral Control System Design for Autonomous Vehicle Considering Localization Uncertainty

Abstract

In this paper, a lateral controller for autonomous vehicles is proposed under conditions of localization uncertainty, utilizing Robust Model Predictive Control (RMPC). Recently, localization accuracy has been largely improved through the integration of digital maps and sensor data from cameras and LiDARs. However, significant challenges remain in environments like highways, where landmarks are sparse and featureless, compromising the effectiveness of sensor data for map matching. This often leads to inaccurate vehicle positioning and, consequently, a loss of control capability. In this study, boundaries of positioning uncertainties are estimated first, and the region of uncertainty is modeled. The controller is designed to account for this uncertainty, where the constraints are adjusted to reduce chattering and biases. Simulation results demonstrate that the proposed controller effectively maintains path-tracking performance even in the presence of localization uncertainty. Additionally, this study shows that the performance can be maintained under significant localization error, indicating that it can contribute to relaxing the localization requirements.