Vehicle-type Specic Headway Distribution in Freeway Work Zones: A Nonparametric Approach

摘要

Vehicle Headway is critical to trac ow control and operation. Signicant research has been conducted on this topic. Previous work focused primarily on parametric mod- els which are based on certain assumptions, thus its reliability is still debated. This paper employs a nonparametric distribution model with Gaussian kernel functions to investigate freeway work zone scenarios. Without prior assumptions of the possible distribution model, a Gaussian kernel model is capable of capturing the intrinsic fea- tures from empirical work zone headway data to describe the headway distribution.The nonparametric model would be more favorable in various scenarios. In addition, this paper aims on the vehicle-type specic model: car-car, car-van, car-truck, van-car, van-van, van-truck, truck-car, truck-van, and truck-truck. The K-S test conrmed the performance of the nonparametric model. All K-S statistics indicate that the non-parametric model with Gaussian kernel model outperforms parametric models such as the lognormal distribution. Experiments were further conducted on the nine types of headways to provide visual evidence. The Gaussian kernel model shows robust capa- bility in describing the probability density function and cumulative density function, the relative error is rather small and can be considered to be negligible. The lognormal distribution is compared against the Gaussian kernel model when tting the empirical headway data. The results show Gaussian kernel model performs better in approximat- ing empirical headway data in work zones. As a result, the relative error is consistently smaller than the lognormal distribution which has a large initial uctuation. The re- sults suggest that the nonparametric distribution model with Gaussian kernel functions has a better goodness-of-t in the vehicle-type specic work zone scenario.