Low-Cost Precise Vehicle Localization Using Lane Endpoints and Road Signs for Highway Situations

Abstract

In highways, lane markings are undoubtedly the most widely used landmarks for vehicle localization. However, they have a drawback in that they lack information on longitudinal position estimation and ego-lane identification. To alleviate this drawback, this paper presents a practical vehicle localization system for highways. The proposed system utilizes lane endpoints to enhance longitudinal position accuracy and road signs to improve the ego-lane identification accuracy. This system efficiently fuses the lane markings, lane endpoints and road signs along with a digital map and other low-cost sensors in a particle filter framework. Since it only uses low-cost sensors such as a monocular front-view camera, in-vehicle wheel speed and yaw rate sensors, as well as a low-end global positioning system (GPS), it is ready to mount on mass-produced vehicles. In the experiment, the proposed system was quantitatively evaluated using a dataset obtained while driving on 40 km stretch of highway, and outperformed previous approaches by showing a lateral position error of less than 0.12 m and a longitudinal position error of less than 0.25 m in terms of root mean square error (RMSE).

Keywords

• Highway autonomous driving
• lane endpoint
• particle filter
• road sign
• sensor fusion
• vehicle localization