Path Tracking Control of Autonomous Vehicles Using Augmented LQG with Curvature Disturbance Model

Abstract

Path tracking control is an important technology for the safety and comfort of autonomous vehicles. In tracking problems, vehicle lateral motion is highly affected by the desired path curvature, which is known as disturbance, and thus the controller performance can be additionally improved by using it in an optimal control method. This paper presents an augmented linear quadratic Gaussian (LQG) controller for reducing tracking errors and estimating accurate states. The proposed LQG is designed based on the augmented state space model, which contains lateral error dynamic model and curvature disturbance model induced from path mathematical properties. With optimal gain achieved through augmentation, the proposed method calculates the front steering wheel control input in the controller and performs state estimation in the observer by considering the tracking error and curvature simultaneously. The controller is implemented in real-time on an autonomous vehicle for driving experiments. The results show improved performance in comparison with conventional LQG without augmentation.