Roadside Features in Crash Prediction Models: Data Collection and Evaluation
Type of DegreeDissertation
MetadataShow full item record
A roadway departure (RwD) crash, comprising run-off-road (ROR) and cross median/cross centerline head-on collisions, is defined as a crash in which a vehicle crosses an edge line, a centerline, or otherwise leaves the traveled way. These types of crashes tend to be more severe than other crash types (e.g., rear end, head on, sideswipe). According to the U.S. National Highway Traffic Safety Administration (NHTSA), in 2013, 56 percent of all motor vehicle fatalities involved RwD crashes. Moreover, ROR crashes accounted for 62 percent of the total number of fatal motor vehicle crashes in the United States in that year. There are a number of reasons a driver may leave the travel lane, including, but not limited to, an avoidance maneuver, inattention or fatigue, or traveling too fast with respect to weather or geometric road conditions. There are also a number of roadway design factors that can increase the probability that driver error will result in an RwD crash (e.g., travel lanes that are too narrow, substandard curves, and unforgiving roadsides). Moreover, the probability of the severity of RwD crashes depends on the roadside features, including sideslopes, fixed-object density, offset to fixed objects, and shoulder width. The high fatality rates associated with this crash type necessitates further investigation to build a roadside inventory database, identify the factors contributing to crashes, and then to implement effective safety countermeasures. This dissertation is a collection of four papers as separate chapters. Chapter 1 evaluates the capability of existing methods for collecting roadside features vital to the effective implementation of the Highway Safety Manual (HSM) (published in 2010). Since the release of the HSM, many states have sought to tailor various safety measures and functions within the report to better reflect road safety in their specific locations. However, the widespread utilization of the HSM faces significant barriers as many state departments of transportation (DOTs) do not have sufficient HSM-required highway inventory data. A significant amount of roadside information is missing in most databases, such as roadside slope, grade, roadside fixed objects and their density, and offset to the edge of the travel way. Many techniques have been used by state DOTs and local agencies to collect highway inventory data for other purposes, but it is unknown which of these methods or combination of methods is capable of efficient data collection while also minimizing cost and safety concerns. By virtue of the fact that many state DOTs are currently redesigning their asset management plans to meet the performance requirements of the national Moving Ahead for Progress in the 21st Century Act (MAP-21), there is a need to better understand the potential applications of existing highway inventory data collection methods for gathering HSM-related roadway inventory data. Chapter 2 identifies the significant contributing factors to ROR crashes, which have accounted for the majority of RwD events, using an exploratory data analysis (EDA) technique to determine the dataset structure. To realize the vision of the FHWA’s Toward Zero Deaths, one of the challenges researchers and state DOTs face is how to identify key contributing factors within large and complex datasets in order to implement effective safety countermeasures accordingly. Chapter 3 presents an overview of cost-effective improvements for preventing vehicle departures from roadways, and it provides transportation practitioners with a good understanding of the effectiveness of RwD safety countermeasures. In order to realize the vision of the Federal Highway Administration’s (FHWA’s) Toward Zero Deaths, many safety countermeasures (e.g., signs, pavement safety, and roadside design) have recently been implemented by state DOTs and local agencies to mitigate RwD crashes. Chapter 4 presents a new reliability analysis approach to evaluating roadside safety for rural two-lane roads. Currently, the clear zone width and sideslope are used to determine the roadside hazard rating (RHR) and to quantify roadside safety for rural two-lane roadways on a seven-point pictorial scale. Since these two variables are continuous and can be treated as random variables, probabilistic analysis can be applied as an alternative method to account for uncertainty. Specifically, by emphasizing reliability analysis, it is possible to quantify the roadside safety level by treating the clear zone width and sideslope as two continuous, rather than discrete, variables and to calculate their reliability indices accordingly. As a national priority, the findings of this dissertation can prevent or mitigate the frequency and severity of RwD crashes, which will result in saving lives and reducing crash costs to society overall. It also provides guidance for all state DOTs, as a national-level resource, to obtain a better knowledge of cost-effective roadside inventory data collection methods, factors contributing to RwD crashes, and associated safety countermeasures, all of which will yield multiple national benefits.
- Roadside Features in Crash Prediction Models_Data Collection and Evaluation.pdf