Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
3 CHAPTER 2 Assessing the Relationship between Driver, Roadway, Environmental, and Vehicle Factors and Lane Departures on Rural Two-Lane Curves: An Investigation Using the SHRP 2 Naturalistic Driving Study Shauna Hallmark (Center for Transportation Research and Education at Iowa State University) and Dan McGehee (Public Policy Center at the University of Iowa) Background The Federal Highway Administration estimates that 58% of roadway fatalities are lane departures (FHWA 2009). Addressing lane-departure crashes is therefore a priority for national, state, and local roadway agencies. Horizontal curves are of particular interest because they have been correlated with overall increased crash occurrence. Curves have approximately three times the crash rate of tangent sections (Glennon et al. 1985). Preston (2009) reported that 25 to 50% of severe road departure crashes in Minnesota occurred on curves, even though they only account for 10% of the system mileage. Around 76% of curve-related fatal crashes are single vehicles leaving the roadway and striking a fixed object or overturning (AASHTO 2008). Curve- related crashes have a number of causes including roadway and driver factors. Degree of curve or radius of curve is the roadway factor most cited in the literature as having an impact on crash risk (Luediger et al. 1988; Miaou and Lum 1993). Other factors that have been correlated to the frequency and severity of curve-related crashes include length of curve, type of curve transition, lane width and shoulder width (Zegeer et al. 1981), preceding tangent length (Milton and Mannering 1998), presence of spirals (Council 1998), grade (Fink and Krammes 1995) and required speed reduction between the tangent and curve. Driver error on horizontal curves is often due to inappropriate speed selection, causing an inability to maintain lane position. FHWA estimates that approximately 56% of run-off-road fatal crashes on curves are speed related. Distracting tasks such as radio tuning or cell phone conversations can draw a driverâs attention away from speed monitoring, changes in roadway direction, lane keeping, and detection of potential hazards (Charlton 2007). Other factors include sight distance issues, fatigue, or complexity of the driving situation (Charlton and DePont 2007; Charlton, 2007). McLaughlin et al. (2009) evaluated run-off-road events in the Virginia Tech Transportation Institute (VTTI) 100-car study and found that distraction was the most frequently identified contributing factor along with fatigue, impairment, and maneuvering errors.
4 Project Objectives Because rural curves pose a significant safety problem, and the interaction between the driver and roadway environment is not well understood, the objective of this research is to assess the relationship between driver behavior and characteristics, roadway factors, environmental factors, and the likelihood of lane departures on rural two-lane curves. In order to accomplish this, SHRP 2 NDS and RID data were used to develop initial models that explore how drivers interact with the roadway environment and what conditions are present when a driver does not successfully negotiate a curve versus what conditions are present when successful negotiation occurs. This includes the conditions of the driver, roadway, and to a limited extent, the environment. The project will gain insight into where a driver's attention is focused during curve negotiation and what roadway cues are the most effective in keeping drivers within their lane. Most highway agencies are proactive in implementing a range of countermeasures to reduce lane departures on curves and other areas. However, agencies have only limited information about the effectiveness of different countermeasures. The results of this research will provide more information about which specific roadway features are correlated to increased risk of lane departure. It will also provide valuable information about how drivers interact with roadway features and the impact that that has on the effectiveness of countermeasures. This will allow agencies to make better decisions about countermeasure selection. Research Questions Addressed This research was tailored to address three fundamental research questions: 1. What defines normal curve negotiation? 2. What is the relationship between driver distraction, other driver characteristics, roadway characteristics, environmental characteristics and risk of lane departure? 3. What roadway cues and countermeasures are the most effective in getting a driverâs attention and how do they affect driver response? Each question addresses the problem from a different perspective and as a result a different methodology was proposed for each as described in the corresponding section. Data Collection and Reduction In Phase 1, Florida data collection was the most advanced so the team manually identified rural curves in Florida based on information about where trips were likely to have occurred. Geographic buffers around those curves were submitted to VTTI and a total of 90 useable traces (trips) were identified in the NDS data. Roadway, environmental, and operational characteristics were extracted. A site visit was made to the VTTI secure data enclave to reduce driver glance location and distraction for each trace. In general it was determined that the data were similar to
5 what had been expected and that all of the variables necessary with a few minor exceptions would be available for Phase 2. Considering the results for Phase 1, the sample size was determined for Phase 2 based on what would be needed to fully answer the research questions balanced with what could economically and feasibly be conducted in Phase 2. Although each question has slightly different data needs, the same NDS and RID data can be used to answer all three questions. The sample size for Phase 2 was based on research question 2 since it has the most covariates and will require the most data. It was estimated that around 1,000 traces would be sufficient. The driver and roadway characteristics initially determined to be the most relevant were used to construct a sampling plan with a specific focus on roadway countermeasures. Phase 2 data will focus on North Carolina, Indiana, New York, and Pennsylvania which have the most rural driving. Crash Surrogates Crash surrogates are necessary to conduct both Phase 1 and Phase 2 of this research. A number of potential crash surrogates were considered against what data were available in Phase 1 and are expected to be available in Phase 2. Lane deviation, or position within the lane, was selected as the type of crash surrogate that was the most feasible given the available data and resources. Crash surrogates will be ordered by risk and boundaries developed between thresholds which will include normal driving, safety critical lane departures, near crashes, and crashes. The team used initial datasets to determine the boundary between normal driving and a lane departure event. Analysis Plan Because three fundamental questions were addressed, a different methodology was developed specific to each. Early assessments of the data indicate that the necessary data can be extracted in Phase 2 to answer each research question and initial model results indicate that the selected models are appropriate. Answering research question 1 entails developing a conceptual model of curve driving. The model will help identify zones where driver workload and behavior differ and assess how this affects successful curve negotiation. The model will also be used to define boundaries between lane departure events and normal driving. Times series data, at the level collected from each vehicle, will used as the data input. The conceptual model evaluates changes in driver attention and response through curve negotiation expressed as changes in vehicle kinematics. Initial results, using a qualitative assessment, indicate that forward acceleration, accelerator pedal position, and speed showed marked changes within 200 meters upstream of a curve suggesting that this area can be used as the curve area of influence. The relationships between curve negotiation and vehicle kinematic variables, such steering position, were assessed as shown in Figure 2.1.
6 Figure 2.1. Steering wheel position for curve 503 WB. Research Question 2 will address how driver distraction, speeding, and other driver behaviors in conjunction with roadway and environmental factors affect the likelihood of a crash or lane departure on rural two-lane curves. Multivariate logistic regression was used to model the probability (odds) of a given type of lane departure based on driver, roadway, and environmental characteristics. Data were aggregated to the event level. A logistic model was developed in Phase 1 using 90 events. A model was only developed for right-side lane departures since there were only five left-side lane departures in the Phase 1 data. Results indicate that the odds of a right-side lane departure are 7.1 times higher for drivers negotiating the outside of the curve than for drivers on the inside of a curve. In this model, curve direction was the only statistically significant variable. Initial results were not as good as had been expected. However, the range of characteristics evaluated was limited. Only a few curves were available in the dataset, and of the 12 drivers, only three drivers produced over 70% of the lane departures. This suggests that the likelihood of a lane departure is highly correlated to driving style and the over-representation of only a few drivers may have overwhelmed the model. This underscores the need to ensure that a large sample of drivers over a range of geometric conditions is available for Phase 2. The threshold was increased to -0.3 m beyond the right edge line and a simple odds ratio found that the amount of distraction a driver was engaged in upstream of the curve was also significant. Research Question 3 focuses more specifically on driver response to changing roadway characteristics and will be used to determine what roadway cues and countermeasures affect driver response. A time-series model will be developed which will incorporate the dynamic process of information acquisition and response as a driver negotiates a curve. This will allow the team to determine how roadway cues and countermeasures affect driver response. For example, drivers on a roadway with rumble strips may exhibit better lane keeping than drivers on a curve with similar characteristics without rumble strips.
7 A dynamic linear model was used to develop an initial model, which related lane offset to curve characteristics. The dynamic model form was selected because it allows for the flexibility of modeling several time series at once which will allow the team to develop estimates of what ânormal drivingâ looks like, allowing differences in covariates or behavior when lane departures occur to be evaluated. Additional Considerations for Phase 2 A review of initial data for Phase 1 indicated that, for both the NDS and RID data, overall data quality, resolution, and format were similar in most cases to what was expected. Review of the data also suggested that more stringent thresholds need to be set for lane departure because a higher than expected number of lane departures resulted from the initial threshold of -0.1 meters beyond the lane line. Lessons Learned A number of lessons were learned during Phase 1 which informed plans for Phase 2. The most important lesson was that the data request should be carefully constructed so that the unusable data are filtered out. For instance, it was realized that a filter should be used to exclude traces in which desired data, such as lane position, are missing. A protocol to identify unusable traces was developed to ensure that sufficient samples are available for Phase 2. Translating Research to Practice The research will provide results that will aid transportation agencies in understanding the relationship between driver distraction and other characteristics and curve negotiation. The results will allow agencies to better understand which curve treatments result in fewer and less severe lane departures. Stakeholders who are expected to use the results include safety researchers, AASHTO, FHWA, state Departments of Transportation, counties, and cities. The research will also result in information about driver distraction that can be used by policy makers. Most highway agencies actively implement a range of countermeasures to reduce lane departures on curves and other areas. However, only limited information is available about the effectiveness of different countermeasures. The results of this research will provide more information about which specific roadway features are correlated to an increased risk of lane departure. It will also provide valuable information about how drivers interact with roadway features and the impact that their interaction has on the effectiveness of countermeasures. This will allow agencies to make better decisions about countermeasure selection. Understanding how drivers approach the task of curve negotiation will also provide invaluable information about why certain countermeasures work. The research has implications for roadway design, selection of sign type and placement, sight distance, and selection and application of countermeasures. It is expected that more appropriate application of countermeasures to mitigate run-off-road or head- on crashes on curves will result in fewer fatal crashes.
8 Steps will be taken to ensure that, when appropriate, results from Phase 2 can be translated to practice. Tech briefs will be developed to summarize information about which factors are correlated to lane departures. Results about the effectiveness of different countermeasures will be included in a toolbox of curve countermeasures which the team has developed for state and county agencies. The team will communicate any information about placement or effectiveness of existing traffic control devices to the National Committee on Uniform Traffic Control Devices.
