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.
1Project Objectives Rural curves are known to pose a significant safety problem, but the interaction between the driver and roadway environment is not well understood. Thus, the objective of this research was to assess the relationship between driver behavior and characteristics, roadway factors, environ- mental factors, and the likelihood of roadway departures on rural two-lane curves. To accomplish this, data from the second Strategic Highway Research Program (SHRP 2) Natu- ralistic Driving Study (NDS) and Roadway Information Database (RID) were used to develop relationships between driver, roadway, and environmental characteristics and the risk of a road- way departure on curves. The project focused on rural two-lane curves on paved roadways. Only paved roadways were included because the machine vision application used in the lane-tracking system does not func- tion well when lane lines or obvious discontinuities between the lane and shoulder surface are not present. Rural was defined as one or more miles outside an urban area. Additionally, only roadways posted at 64 km/h to 97 km/h (40 mph to 60 mph) were included. Research Questions Addressed This research was tailored to address four fundamental research questions: 1. What defines the curve area of influence? 2. What defines normal behavior on curves? 3. What is the relationship between driver distractions; other driver, roadway, and environmental characteristics; and risk of roadway departure? 4. Can lane position at a particular state be predicted as a function of position in a prior state? Each question addresses the problem from a different perspective. As a result, a different meth- odology was proposed for each, as described in the corresponding sections. Data Collection and Reduction Chapter 3 summarizes how Institutional Review Board (IRB) approval and data requests were completed; it also describes data reduction. The team manually identified rural curves in Florida, New York, Indiana, Pennsylvania, and North Carolina based on information about where trips were likely to have occurred. Segments were provided to Virginia Tech Transportation Institute (VTTI) staff, who identified trips through those segments. Executive Summary
2Roadway, environmental, and operational characteristics were extracted as described in Chap- ter 4. Site visits were made to the VTTI secure data enclave to reduce driver glance location and distraction for each trace. Crash Surrogates The use of crash surrogates was necessary because only one crash and three near crashes were available at the time this research was conducted. Chapter 5 discusses the rationale for selection of the identified crash surrogates. A number of potential crash surrogates were considered against the data available and the expected accuracy of relevant variables in the NDS data (e.g., lane position, forward radar, vehicle position). Lane offset was the best crash surrogate, but lane offset was not reliable in a number of traces. As a result, it was used for Research Questions 2 and 4, resulting in a smaller sample of data for those research questions. Because offset was not reliable in a number of traces, the research team determined that encroach- ments would be the best crash surrogate for Research Question 3. A right-side encroachment is defined as the right side of the vehicle crossing the right lane line, and a left-side encroachment is defined as the left side of the vehicle crossing the centerline. Results and Discussion Since four fundamental research questions were addressed, a different methodology was devel- oped specific to each, as outlined in Chapters 6 through 9. In addition to the analytical method, these chapters discuss the data sampling and segmentation approach, general variables consid- ered, results, and implications. The following sections provide a brief summary of findings and implications for each research question. Research Question 1 Answering Research Question 1 entailed understanding at what point drivers begin reacting to the presence of a curve upstream of the curve. Understanding where drivers begin to react to the curve is important for placement of traffic control and countermeasures. A better understanding can also help agencies determine optimal placement of advance signing and other counter- measures. Research Question 1 was also used to indicate the curve area of influence for Research Questions 2, 3, and 4. Time series data were modeled using regression and Bayesian analysis. Results indicate that, depending on radius of curve, drivers begin reacting to the curve 164 m to 180 m (538.1 ft to 590.6 ft) upstream of the point of curvature. These results were compared with sign placement guidelines in the Manual on Uniform Traffic Control Devices (Federal Highway Administration 2009), and it was determined that the guidelines are appropriately set based on where drivers actually react to the curve. Research Question 1 also found that drivers begin reacting to the curve sooner for curves with larger radii than for curves with smaller radii. Drivers may not be able to gauge the sharpness of the curve, or sight distance issues may be a concern for sharper curves. This suggests that use of countermeasuresâsuch as chevrons or raised pavement markings (RPMs)âthat better delineate the curve may provide better advance information for drivers. It should be noted that the model only identified the point at which drivers reacted to the curve. This research question did not attempt to answer whether the reaction point was sufficient for drivers to successfully negotiate the curve. Research Question 2 Research Question 2 developed conceptual models of curve driving to assess changes in metrics as the driver negotiates the curve. Understanding how a driver normally negotiates a curve
3 provides insight not only into how characteristics of the roadway, driver, and environment potentially influence driving behavior, but also into areas that can lead to roadway departures. Knowing how much drivers normally deviate in their lane, as well as how they choose their speed, could potentially have implications on policy or design. Data for several positions upstream and along the curve were sampled from the time series data. Models were developed for lane position and speed for both inside (right-hand curve from the perspective of the driver) and outside (left-hand curve from the perspective of the driver). Lane position was modeled as the offset of the center of the vehicle from the center of the lane. Models were developed using generalized least squares. Results indicate that lane position within the curve is influenced by lane position upstream of the curve. The models developed for offset of lane centerline found that drivers who were dis- tracted or who glanced away from the roadway tended to shift away from the center of the lane. When driving on the inside of the lane, a driver who was distracted at a particular point within the curve tended to shift 0.14 m to the right by the next point in the curve. When driving on the outside (left-hand curve), a driver who engaged in a non-roadway-related glance at a particular location within the curve was expected to move to the left, or toward the centerline by 0.13 m at that same point. This confirms the role of distraction in lane keeping. Additionally, the models found that drivers on the inside of a curve tended to move more to the right at the center of curve, while drivers on the outside of curves were at the furthest point from the lane centerline at the beginning of the curve. As a result, drivers may be particularly vulnerable to roadway departures at certain points in the curve negotiation process. These results suggest that countermeasures such as rumble strips, paved shoulders, and high-friction treat- ments may ameliorate the consequences of variations in lane position through the curve. Additionally, the lane offset models indicate that age and nighttime driving are factors in driver lane position. The model for speeding in the curve found that if drivers are speeding in the upstream, they will also speed in the curve. Drivers of sport utility vehicles (SUVs) and pick-up trucks traveled on average 2.1 km/h (1.3 mph) faster than drivers of passenger vehicles. Speeds were predicted to be 0.9 km/h (0.5 mph) lower for each additional 10 years in age for a driver, and drivers engaged in a non-roadway-related glance are expected to travel 5.3 km/h (3.3 mph) slower than drivers who do not engage in a non-roadway-related glance. This suggests that drivers whose attention is focused away from the roadway do not maintain longitudinal control. The results indicate that distractions/nonroadway glances affect lateral and longitudinal control. Although drivers are more likely to travel at slower speeds, they are more likely to vary within their lane. The models also confirm that speed plays an important role in curve negotiation. This suggests that effective speed management countermeasures will have an impact on curve negotiation. Speed management countermeasures include better delineation of the curve (i.e., chevrons, edge lines, post-mounted delineators) so that drivers can better estimate the sharpness of the curve. Other measures, such as transverse speed markings or speed feedback signs, target drivers who are traveling over the speed limit. Research Question 3 Research Question 3 addressed how driver behavior in conjunction with roadway and envi- ronmental factors affect the likelihood of a roadway departure on rural two-lane curves. Four different models were developed using multivariate logistic regression. Two models evaluated the probability (odds) of a right-side or left-side encroachment based on driver, roadway, and environmental characteristics. Two additional models evaluated the probability that a driver would exceed the advisory speed if present or posted speed limit if not present at the curve entry by 8 km/h and 16 km/h or more (5 mph and 10 mph or more). Data were aggregated to the event level.
4The model for right-side encroachments indicates that the probability increases as drivers spend less time glancing at the forward roadway. The results also indicate that a right-side lane departure is 6.8 times more likely on the inside of a curve compared with the outside of the curve. Lane departures are slightly more likely (1.3 times) for curves with any type of curve advisory sign (including W1-6). It is unlikely that the presence of a warning sign itself increases the prob- ability. Rather, it is likely that advisory signs are more likely to be present on curves of a certain type (i.e., those with sight distance issues, sharper curves), and encroachments are also more likely for those road types. Additionally, the results suggest that the simple presence of curve warning signs may not mitigate roadway departures. A statistically significant but small correlation exists between radius of curve and probability of a right-side encroachment. Drivers were 0.33 times less likely to have a right-side encroach- ment on roadways with a guardrail. Presence of a guardrail decreased the probability of a right- side encroachment. The purpose of a guardrail is to mitigate the consequences of a driver leaving the roadway rather than to keep the driver from leaving the roadway. Consequently, a guardrail in and of itself does not mitigate roadway departures. The presence of a guardrail may suggest to the driver that roadway conditions are less safe, resulting in better driver attention. Addition- ally, few delineation countermeasures (e.g., chevrons) were present in the curves included in the analysis. As a result, a guardrail may provide some delineation of the curve, which provides feedback to the driver about the sharpness of the curve. The model for left-side encroachments indicates that males are more than four times more likely to have a left-side lane departure, and drivers traveling on the inside of the curve are 0.1 times less likely to have a left-side encroachment than drivers traveling on the outside of the curve. The impact of radius was statistically significant but minor. The probability that a driver will be 8 km/h (5 mph or more) over the posted/advisory speed is higher when the driver is younger and has a higher average speed upstream and when edge line markings are obscured or not present. The amount of time a driver spends following another vehicle, presence of lower visibility conditions, and presence of paved shoulders and RPMs decrease the probability that he or she will enter the curve 5 mph or more over the posted/ advisory speed. The probability that a driver will be 16 km/h (10 mph or more) over the posted/advisory speed is higher when the driver has a higher average speed upstream. The probability is lower when the average glance at roadway-related tasks is longer and when paved shoulders and RPMs are present. Results from the right-side encroachment and speed models suggest that better curve delinea- tion may allow drivers to better gauge upcoming changes in roadway geometry, resulting in better speed selection and decreased risk of a roadway departure, and may help decrease speed. Delineation countermeasures include chevrons, the addition of reflective panels to existing chev- ron posts, reflective barrier delineation, RPMs, post-mounted delineators, edge lines, and wider edge lines. The speed models suggest that driver age and upstream speed have a significant impact on driversâ speed within a curve. As a result, speed management countermeasures that affect tangent speed will also decrease curve speeds. The results also indicate that speed management is appro- priate to get driversâ attention before entering a curve. Countermeasures specifically targeted to reduce speed on curves include dynamic speed feedback signs, on-pavement curve warning signs, and flashing beacons. Research Question 4 Research Question 4 focused more specifically on driver response to changing roadway charac- teristics and traffic conditions. Time series models were developed to incorporate the dynamic process of information acquisition and response as a driver negotiates a curve. The analysis evaluated the influence of roadway geometries or traffic conditions on driversâ lane-keeping
5 behavior. For example, drivers on a rural two-lane roadway tend to have larger lane deviation from the centerline when there is an oncoming vehicle. Two types of dynamic linear models (DLMs) were built in this study to describe and explain the curve negotiation process: DLM with intervention analysis and DLM with autoregression and moving average (ARMA). The DLM with intervention analysis was mainly used for explana- tory purposes, relating lane offset to curve characteristics and traffic conditions. The DLM with ARMA was mainly used for forecasting purposes, which could be used for a roadway departure warning system. Only limited data were used in the analyses, given that the objective was to demonstrate the utility of the approach. Results indicate that lane position can successfully be modeled as a func- tion of vehicle position in a prior state and as a function of other characteristics such as position within the curve or presence of oncoming vehicles. The methodology shows promise for use in development of roadway departure crash warning systems.
