L.H. Xu, Q. Luo, L.Y. Zhang, Y.G. Huang

Pages: 37-48

Abstract
In this paper, a new approach, which is called the variable virtual lane boundary (VVLB) for lane departure warning systems is proposed based on the idea of future offset distance (FOD) algorithm. There are two goals for this approach: giving more warning time and triggering fewer nuisance alarms. Firstly, we obtain the standard deviation of lateral position and velocity taken from driver’s history data. Then, we give three types of driver, including tight, normal and loose, based on the analysis of the standard deviation of data. After that, the virtual lane width is confirmed considering about driver's type, local (real-time) lateral distance and speed by using a fuzzy logic inference method. The lane width is allowed to vary according to different drivers and real-time driving situations. When the vehicle is predicted to exceed the virtual lane boundary, an alarm will be triggered. Finally, we do the compared experiments on the performance of VVLB, Car’s Current Position (CCP), Roadside Rumble Strip (RRS) and Time to Lane Crossing (TLC) method based on the real driving data. The compared results show that there are fewer nuisance alarms with the VVLB method than with the TLC method, and there is much more warning time with the VVLB method than with the CCP and RRS method.

Keywords: lane departure; driver’s type; VVLB; fuzzy logic inference