Robust Lane-keeping Control System for Automated Driving

Abstract

Recently, much attention has been paid to the development of automated driving that can improve the safety and enhance driver’s comfort and convenience. Thus, various methods have been developed to reduce the driver's load and fatigue with development of environment recognition sensors such as camera and radar. Despite their efforts, some problems are remained such as slow update and robust failed sensor characteristics of the camera sensor for acquiring the lane information, and a steering system for controlling the lateral movement of the vehicle is a non-linear characteristic and precise control due to the wide range of the disturbance. The main topic of this dissertation is robust lane-keeping system(LKS) for automated driving. For the LKS, three methods are proposed as follows: 1) multi-rate lane-keeping control for lateral dynamics control, 2) predicted virtual lane (PVL) for lane detection, and 3) torque overlay based robust steering wheel angle control for electric power steering. The robust multi-rate lane-keeping control scheme is proposed to improve the lane-keeping performance. This method consists of a multi-rate Kalman filter (KF) and a state feedback controller. The KF is developed to resolve the problems caused by slow lane detection due to the vision processing system. Using the data measured from the camera sensor at a slow rate, this multirate KF estimates the vehicle states at a fast rate of the microprocessor. To improve lane-keeping performance on a curved road, the integral of lateral offset error is added to the state feedback controller. Utilizing the estimated states, the linear quadratic state feedback control operates at the same fast update rate of the ECU. The main contribution of the proposed PVL method is that the virtual lane can be substituted for lane detection using a camera sensor when the camera image processing fails to detect the lane. The proposed method generates the predicted virtual lane using the relative movement between a vehicle and a lane. To predict the lane, a third-degree polynomial function of the longitudinal distance is used as the lane model. Each coefficient of the lane polynomial function at the next sampling time is geometrically calculated using the relative movement of the vehicle, the lanes, the longitudinal velocity and the yaw of the vehicle at present time. Then, the predictive virtual lane at the next sampling time is obtained in the absence of lane information from the camera sensor at the next sampling time. The torque overlay based robust steering wheel angle control of electric power steering is developed for lateral control in autonomous vehicles. External disturbances, system dynamics, and input function uncertainty are regarded as disturbances. In order to estimate the full state and disturbance, an augmented observer is designed. A nonlinear damping controller is developed via backstepping to suppress position tracking errors using input-to-state stability (ISS) property. Since the proposed method is designed based on torque overlay, it is easier for the driver to take over the steering wheel control in an autonomous vehicle when an emergency occurs than under the previous methods that were designed based on angle overlay. The developed methods were experimentally evaluated using test vehicles. The lateral control performance was improved by the combination of three methods. The multi-rate lane-keeping controller and the steering wheel angle controller reduced the lateral offset error on both straight and the curved road tests. The multi-rate based control scheme reduced the ripple in the yaw rate. Furthermore, the combination of the multi-rate lane-keeping control scheme and the PVL can make the LKS operate when the lane information is momentarily unavailable. It was observed that the LKS with the developed methods offers robustness in the presence of parameter uncertainties and external disturbance.