| Summary: | 博士 === 國立臺灣大學 === 土木工程學研究所 === 99 === Due to inappropriate intergreen interval, the dilemma zone (DZ) has been identified as a critical issue in vehicle safety, because it might lead drivers to make incorrect decisions and causes potential risks of rear-end or right-angle accidents.
In practice, the Gazis, Herman and Maradudin (GHM) model is commonly used by traffic engineers to design intergreen to prevent the formation of the DZ. According to the yellow change interval formula proposed in the ITE handbook, the designed yellow change interval is supposed to allow an approaching vehicle moving at a constant speed to either safely stop or clear the intersection. Thus, the DZ is not supposed to happen. In reality, the DZ is difficult to be eliminated due to its dynamic features of location and length, resulting from a variety of drivers’ behavior and characteristics. Hence, current implementation of yellow change intervals still has a potential to form a DZ if a vehicle approaches an intersection at a speed higher than that identified in the ITE formulation. As a result, the static ITE approach might be less helpful for removing DZs.
The traditional concept for inter-green interval design would force traffic engineers to face the trade-off between the vehicle safety and operational efficiency of traffic signals at signalized intersections. Besides, the dynamic feature of dilemma zones cannot be effectively captured by the traditional approaches. A new approach that can improve the vehicle safety without diminishing existing efficiency is desired. Therefore, based on the concept of VII (Vehicle Infrastructure Integration)/connected vehicle, this study aims to develop a series of warning algorithms (WAs) for onboard advanced dilemma zone warning system to handle the dilemma zone problem in real-time. These algorithms are designed for the connected (VII-ready) vehicle to dynamically determine dilemma zones and provide drivers with proper warnings. To increase detection accuracy for dilemma zone, inputs from roadways, drivers and vehicles are considered. Besides, in order to capture the dynamical feature of DZ, a genetic systematic rule improving the prediction capacity of the Kalman filter is integrated into the warning algorithms.
The algorithms are tested by simulation with a variety of scenarios to demonstrate system reliability and performance. The results show the reliability and operating practicability of the proposed algorithms. Furthermore, the warning accuracy contour plot designed in this study also provides an effective way to assist users or system designers in determining the critical parameters of the algorithms for future implementations.
|
|---|
