Guidelines for Traversability of Roadside Slopes (2019)

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 Run-off-road (ROR) traffic crashes account for almost one-third of the deaths and serious injuries each year on U.S. highways. There were 41,059 people killed in motor vehicle crashes in 2007, out of which 15,506 people (> 37 percent) were known to be killed in SVROR crashes. In addition, collisions with fixed objects and non-collisions (e.g., rollovers), which mainly occurred off road, accounted for about 19 percent of all crashes, but were responsible for 46 percent of fatal crashes. ROR crashes occurred for a variety of reasons, including • Driver inattention, • Excessive speed, • Driving under the influence of alcohol or drugs, • Collision avoidance, • Roadway condition (ice, snow, rain), • Vehicle component failure, and • Poor visibility. FARS uses the terms first harmful event (FHE) and most harmful event (MHE) in describing the events of a fatal crash. FHE is defined as the first property-damaging or injury-producing event in the crash, while the MHE is the single impact that causes the greatest trauma and damage in each crash. Using the FARS data for the 2004–2009 period, Table 2.1 illustrates that the top five MHEs (after impact with other vehicles as a harmful event) involved ditches and earth embankments as the FHE. For comparisons, the top five MHEs are also included in the table (after culvert, guardrail face, concrete traffic barrier, and utility pole as the FHE). The following observations can be made from the table: • When ditch is the FHE, the most serious trauma and damage to the vehicle (or MHE) results from rollover (53.2 percent), the ditch itself (15.8 percent), and a standing tree (15.5 percent). • After impacting earth embankments as the FHE, the consequent MHEs are rollover (51.2 percent), earth embankment itself (30.6 percent), and standing tree (9.0 percent). • Very similar to earth embankments, after striking culverts as the FHE, the consequent MHEs are rollover (50.1 percent), culvert itself (29.6 percent), and standing tree (9.0 percent). • After striking guardrail faces as the FHE, the consequent MHEs are guardrail face itself (35.6 percent) and rollover (33.6 percent). • After striking concrete traffic barriers as the FHE, the consequent MHEs are concrete traffic barrier itself (44.2 percent) and rollover (28.5 percent). • After striking utility poles as the FHE, the consequent MHEs are utility pole itself (70.8 percent) and rollover (17.5 percent). C H A P T E R 2 Literature Review

4 Guidelines for Traversability of Roadside Slopes Rollover is obviously the most consequential and deadly event for errant vehicles that involved ditches and earth embankments as the FHE. Slopes in ditches and embankments are expected to have played a key role for these rollover crashes to occur. Rollover crashes are considered a major issue on highways: they are not only a safety problem but they cause economic problems as well (1). The approximate cost of traffic crashes in 2000 was estimated to be $230 billion. According to FARS data for 2001, 10,121 occupants of light vehicles were killed in rollover crashes (Figure 2.1). Of those, 8,407 were killed in single-vehicle rollover crashes. NHTSA provided some safety strategies to improve the crashworthiness of rollover crashes with the goal of decreasing fatalities and serious injuries (1). A project team was formed by NHTSA to conduct an in-depth review of rollover mitigation, which was considered a priority area. After investigating crash data, NHTSA concluded there are four main categories to research in rollover crashes: vehicle strategies, roadway strategies, behavioral strategies, and other initiatives. Roadway strategies are the main issue with rollover crashes. In fact, 95 percent of rollover crashes were documented to occur once the errant vehicle leaves the roadway and is “tripped” by an object. FHWA and highway agencies proposed two strategies aimed at decreasing rollover crashes. These two strategies are aimed at keeping the vehicle on the road and minimizing the harm to a vehicle if it is steered off the road. A suggested way to keep the vehicle on the road was to install rumble strips on the shoulders to cause inattentive drivers to realize they are leaving the roadway and to make a correction. The quality of the roadside (slope and clear zone) was also considered to influence both severity and frequency of crashes. Thus, mitigation measures were suggested for rural roads, such as widening shoulders and improving roadsides. In 2011, FHWA released the Interactive Highway Safety Design Model (IHSDM), a software analysis tool which provides estimates of the effect of roadway geometry on expected crash frequency. FHWA encourages states to use IHSDM for two-lane road safety assessments. State highway offices are encouraged to provide traversable roadsides in order to reduce tripping mechanisms. MHE FHE Ditch Embank- ment (Earth) Culvert Guardrail Face Concrete Traffic Barrier Utility Pole Ditch 15.8% Culvert 2.6% 29.6% Embankment (Earth) 30.6% Guardrail Face 35.6% Concrete Traffic Barrier 44.2% Utility Pole 4.0% 1.7% 2.8% 70.8% Rollover 53.2% 51.2% 50.1% 33.6% 28.5% 17.5% Standing Tree 15.5% 9.0% 9.0% 6.5% 6.0% Fire/Explosion 2.5% 2.5% 2.7% 2.0% Motor Vehicle on Same Road 7.2% 9.3% 0.5% Motor Vehicle on Other Road 3.0% Immersion 1.4% All Others Combined 8.9% 6.1% 6.0% 14.6% 12.3% 3.2% Notes: Includes crashes occurred on highways with a posted speed limit of 45 mph or higher. Excludes impact with other vehicles as the MHE. Source: FARS 2004–2009. Table 2.1. Top five MHEs involving certain roadside features as the FHE.

Literature Review 5 In 1995, Viner analyzed national crash data to define the nature and importance of the problem of ROR vehicles that rolled over on slopes, including ditches and embankments (2). One year of crash data from the FARS and National Automotive Sampling System’s General Estimates System (NASS GES), administered by NHTSA, were used. FARS is a census of all fatal crashes in the United States, while NASS GES is a probability sample of police-reported vehicle crashes of all severity types, from property-damage-only (PDO) crashes to fatal (Type K) crashes. In addition, data from two specialized databases was used to obtain insight on vehicle orientation and driver maneuvers in such crashes. Driver fatality counts were developed from the 1991 FARS data. Estimates of involved vehicles and A-injuries were developed from 1991 GES data. A-injuries are “incapacitating” injuries on the K-A-B-C-PDO police injury scale. Using FARS data, slope rollovers were identified by rollover cases in which FHE took place outside the shoulder, and both FHE and MHE were either overturn or ditch or embankment impacts. Slope rollovers were identified as the leading cause of ROR fatalities. Fatal ROR crashes took place mostly on rural two-lane roads—82 percent of slope rollovers and 87 percent of fixed- object impacts. Horizontal curves were seen to be especially problematic. On rural two-lane roads, about one-third of all crashes and one-half of all fatalities took place on curves for both slope rollovers and fixed-object crashes. Pickup trucks, utility vehicles, and medium/heavy trucks were found to be overrepresented in slope-rollover crashes when compared to their involvement in fixed-object crashes. This Front 25% Rollover 61% Side 10% Rear 2% Other 1% Unknown 1% Front 39% Rollover 39% Side 15% Rear 4% Other 2% Unknown 1% Front 39% Rollover 44% Side 14% Rear 1% Other 2% Unknown 1% Front 41% Rollover 22% Side 31% Cars 4,543 fatalities Pickup Trucks 2,643 fatalities Vans 793 fatalities SUVs 2,142 fatalities (a) (b) (c) (d) Rear 4% Other 1% Unknown 1% Figure 2.1. Rollover occupant fatalities by vehicle type.

6 Guidelines for Traversability of Roadside Slopes finding was consistent between the analysis results from the FARS data and from the NASS GES data. Viner found that most vehicles that rolled over on sideslopes and ditches were skidding (or non-tracking) as they left the pavement (2). Also, vehicle trajectories in slope rollovers were both different from, and more complex than, fixed-object crashes. About one-quarter of the vehicles involved in slope rollovers ran off the road, returned to the pavement, and then ran off the road a second time where the rollover occurred. In contrast, only about 6 percent of the vehicles involved in fixed-object crashes experienced more than one edge line crossing. Viner discussed the limitations of using the available slope geometrics data and location errors of police-reported crash data to study slope rollovers. He concluded that vehicle dynamics simulation models would offer the most promising approach to obtain the needed insight on rollover risks for specific slope and ditch geometric combinations. The author further suggested that data on distributions of vehicle orientations as the vehicle leaves the road and at the trip- ping point for slope rollovers, as well as the data on encroachment speed and angle distributions, would be needed when using the vehicle dynamics simulation models. Since Viner’s study, the number of registered light trucks, especially sport utility vehicles (SUVs), has increased significantly over the years. The composition of on-road vehicles has gradually shifted from passenger cars to light trucks. For example, out of the 1,181 fatal slope roll- overs identified by Viner from the 1991 FARS data, 54 percent were passenger cars, 30.2 percent were pickup trucks, 7.9 percent were utility vehicles, 4.1 percent were medium/heavy trucks, and 2.9 percent were vans. Using similar crash data screening criteria, TTI researchers identified fatal slope rollovers from FARS using data from 2004 to 2009. It was found that passenger cars represented 35 percent of the slope rollovers, pickup trucks 28.6 percent, utility vehicles 25.9 percent, medium/heavy trucks 3.5 percent, and vans 5.5 percent. There is clearly a decrease in the passenger car share from 54.0 percent to 35 percent, and a sharp increase in the share of utility vehicles from 7.9 percent to 25.9 percent. While higher centers of gravity (CG) make light trucks (e.g., pickup trucks, SUVs, vans) inherently less stable than passenger cars, light truck sales have increased dramatically over the years. As a group, light truck sales currently outpace sales of passenger cars, accounting for over 50 percent of all new passenger vehicles sold. Given this change in vehicle fleet, as well as the prevailing antilock braking system (ABS) installation on light vehicles, electronic stability control, increased seatbelt usage rate from about 60 percent in 1991 to 84 percent in 2009, and the phase-in of Federal airbag requirements since 1991, an update of Viner’s study with recent FARS and NASS GES data is needed. Glennon et al. (1983) performed a study in four states (Florida, Illinois, Ohio, and Texas) to compare accidents on rural two-lane curve and straight segments (3). Accident analyses, simulation of vehicle and driver operations using Highway Vehicle Object Simulation Model (HVOSM), field studies, and analytical studies were employed to evaluate the safety and operational characteristics of two-lane rural highway curves. Studies of accidents on highway curves demonstrated that highway curves have higher rates than highway tangents and that highway curve design involves the geometry of the curve itself as well as the design of the alignment in advance of the curve. Safety of highway curves was found to be influenced by primary elements of the cross section, such as roadway width, shoulder width, and roadside characteristics. In fact, results suggested that different roadside slope and clear-zone width requirements may be needed for highway curves than for highway tangents. For sharper highway curves (greater than 4 degrees), the outside may require flatter roadside slopes and more clear-zone width than for highway tangents. As for milder curves,

Literature Review 7 the outside does not seem to require as flat a slope nor as wide a clear zone as sharper curves. HVOSM studies of the cross-slope break on highway curves showed that driver control is sensitive to the shoulder slope. It was recommended that the cross slope should not exceed 8 percent for full-width shoulders. For super elevation rates between 2 and 6 percent, maxi- mum shoulder slopes from 6 to 2 percent were considered acceptable, respectively. For super elevation rates higher than 6 percent, a different kind of shoulder cross-slope design was considered necessary. Some of the main conclusions from the Glennon et al. study were the following: • The calculated average accident rate for roadway curves is significantly greater than the average accident rate for straight roadway segments. • The calculated average SVROR accident rate is significantly greater for roadway curves than for straight roadway segments. • The proportion of fatal and injury accidents increases for roadway curves (with respect to straight segments). • Roadside slope, clear-zone width, and coverage of fixed objects appear to be dominant contributors for a high-accident rate for a roadway curve. • Many existing roadway curves are significantly underdesigned when considering the posted roadway speeds. This may cause safety problems, because the drivers may not decrease their speeds to match the safe speed of a substandard roadway curve. • Roadside slope traversals on roadway curves seem to be more severe than on straight segments due to generally higher vertical decelerations and higher potential for rollover. Roadside slopes on roadway curves need to be flatter than slopes on straight segments. • Roadside slope traversals on highway curves seem to be more severe than on highway tangents. Roadside slopes on highway curves need to be flatter than slopes on highway tangents. Glennon (2007) developed an equation to define the maximum foreslopes suggested for each degree of roadway curve (4). With the assumption of a 1V:4H roadside foreslope to define the threshold of safety for tangent roadways, roadside foreslopes were analyzed to determine which ones have similar thresholds of safety for the outside of roadway curves. Considering a design speed of 60 mph and a lateral acceleration of 0.3 g (5), the following conclusions were made: • For roadways with 60 mph or higher speeds, it is undesirable to have a 1V:3H roadside foreslope. • Roadway curves sharper than 3.5 degrees seem to be undesirable on roadways with a posted speed limit of 60 mph. • When roadway curvature increases, flatter roadside foreslopes are preferred in order to provide the same level of safety as straight roadways. • Any roadside foreslope is considered to have reduced safety when having a roadway curvature of 4 degrees or more. In 2006, Peters highlighted how properties of the slope or embankment are a big reason for rollover crashes (6). Models were used to identify and research the factors that cause rollover crashes. A stability metric was used to compare different terrains. It was concluded, however, that the stability metric could have some uncertainties. 2.1 Roadside Design Guide Not enough data is available regarding the percentage of overturns versus total vehicle encroach- ments for different sideslope ratios. This is due in part to a lack of roadside data in most crash databases. According to the Roadside Design Guide, foreslopes parallel to the traffic flow may be categorized as recoverable, non-recoverable, or critical (7). Recoverable slopes are 1V:4H or

8 Guidelines for Traversability of Roadside Slopes flatter. Foreslope ratios ranging from 1V:3H up to 1V:4H are considered to be non-recoverable but traversable. Slopes steeper than 1V:3H have historically been considered critical foreslopes. The Roadside Design Guide states that such slopes “will cause most vehicles to overturn and should be treated (i.e., flattened or shielded with a barrier) if they begin within the clear-zone distance of a particular highway . . .” This guidance is based largely on studies of the relative severity of encroachments on embankments versus impacts with roadside barriers. On slopes between 1V:4H and 1V:3H, most motorists are expected to reach the bottom of the slope and not be able to stop or return to the roadway easily. The Roadside Design Guide states that fixed obstacles should not be constructed along such slopes and a clear runout area at the bottom of the slope is desirable. It suggests shielding slopes steeper than 1V:3H, as the vehicle is likely to overturn on these slopes. It also discusses alternatives that might be considered on critical parallel foreslopes. Many highway agencies construct so-called barn roof sections in embankment conditions. A relatively flat slope is provided adjacent to the roadway, followed by a steeper slope and a clear runout area at the bottom. This is more economical than a continuous flat slope and appar- ently safer than a continuous steeper slope from the edge of the shoulder. Slopes that change from negative to positive should be treated as channel sections according to the Roadside Design Guide. It also suggests that changes in slope and toes of slopes should generally be rounded to keep vehicles in contact with the ground and enhance traversability. The Roadside Design Guide identifies three regions of the roadside that are important to reducing the potential for loss of control for vehicles running off the road: top of the slope (hinge point), foreslope, and toe of the slope (intersection of the foreslope with level ground or with a backslope, forming a ditch) (Figure 2.2) (7). Some of the options provided by the Roadside Design Guide to reduce the severity of crashes due to roadside slopes are as follows: • Provide a flatter slope between the shoulder edge and the ditch bottom. • Locate ditch farther from the roadway. • Backslopes should be 1V:3H or flatter to accommodate maintenance equipment. • Backslopes steeper than 1V:3H should be evaluated with regard to soil stability and potential crash severity. • Retaining walls should be considered where there would be a slope steeper than 1V:2H. • Rate of 1V:6H or flatter on embankments should be provided where practical. • For moderate heights with good rounding, steeper slopes up to about 1V:3H can also be traversable, but are not always recoverable. Ditch Bottom Variable Width S2 1 1 S1 BackslopeForeslopeHinge Point Toe of Slope Traffic Shoulder Traffic Lane 3 5 4 21 Figure 2.2. Designation of roadside regions.

Literature Review 9 • When the cost of a continuous flat slope is prohibitive, the area adjacent to the roadway should be reasonably flat. • Slope combination should be selected with the objective of limiting injury to unrestrained occupants and limiting vehicle damage during the traversal. The guidelines of the Roadside Design Guide related to roadside slopes were based on studies that were conducted in the late 1960s and early 1970s and included a very limited number of full-scale embankment tests and computer simulations with passenger cars. For example, Ross et al. (1973) utilized the HVOSM to investigate the dynamic response of a passenger sedan traversing different embankment slopes at various encroachment speeds and encroachment angles (8). The encroachment conditions were selected to match the design impact conditions for guardrail. The accelerations obtained from the simulated slope traversals were used to compute a severity index that was compared to the severity of a vehicle impacting a strong post W-beam guardrail system. An “equal-severity” curve was developed as the basis for the recommended guardrail need criterion. The curve established the less severe of the two alternatives—striking a guardrail or traversing an unprotected slope. The roadside cross section investigated consisted of a 10-ft wide shoulder with a 20H:1V slope, an adjoining side slope of varying steepness and height, and a flat-bottomed ditch at the bottom of the slope. Slopes of 1V:2H, 1V:3H, and 1V:6H with heights of 10, 20, 30, and 50 ft were included in the simulation matrix. In addition, 1V:3.25H and 1V:4H slopes were evaluated at a height of 20 ft for a total of 14 roadside embankment configurations. The researchers presented criteria to help engineers make objective decisions on the need for guardrail to shield embankments. It was concluded that for encroachment angles less than 17.5 degrees and speeds up to 70 mph, impacting a strong post guardrail is higher in severity than traversing a 1V:3H embankment with a 20-ft fill height. However, as the encroachment angle increases, the severity of traversing the embankment approaches the severity of striking a guardrail. The researchers observed a considerable reduction in severity index when the embankment slope was flattened from 1V:3H to 1V:3.25H and 1V:4H. “A sharp transition was therefore found to exist in the SI [severity index] at a slope of about 1V:3H for the 20-ft embankment height.” (8) More recently, a benefit-to-cost analysis procedure was utilized to develop general guidelines for guardrail implementation in NCHRP Report 638 (9). The primary goal of this research was to identify the most appropriate guardrail test level based on highway and traffic characteristics. Specific roadway, roadside, and hazard conditions to be analyzed were assigned to a set of hazard scenarios for use in the benefit-to-cost analysis. This study included analysis of both point objects and long hazards such as steep slopes and roadside ditches, and a program called RSAP was used to analyze each hazard scenario under a wide variety of roadway and traf- fic characteristics. As part of the process, layout, construction costs, and crash severities were defined for various safety treatment options. Although this research makes a significant contribution to the state-of-the-knowledge in regard to guardrail need, the severity indices for slope traversal need to be updated just as the severity indices for guardrail impacts required updating. In the case of ditch severity, it is likely that the current severity indices for some ditch sections are too low. A recent study suggests that some roadside slope conditions that have for many years been considered traversable for passenger cars may not be traversable for light trucks (10). Under NCHRP Project 17-11, results from a comprehensive computer simulation study were weighted with real-world data derived from reconstructed SVROR crashes and used to compute rollover

10 Guidelines for Traversability of Roadside Slopes probability as a function of sideslope ratio. Generally speaking, the percentage of vehicle encroachments resulting in overturn increases as the steepness of the sideslope increases. However, a dramatic increase in overturns and decrease in stable vehicle encroachments was observed between sideslope ratios of 1V:4H and 1V:3H. Whereas the probability of rollover increased an average of 12 percent between sideslope ratios of 1V:6H and 1V:4H, the average change in rollover probability between sideslope ratios of 1V:4H and 1V:3H was 92 percent across all the roadway functional classes studied. To investigate the influence of vehicle type on encroachment stability, the data was segregated into two broad vehicle classifications: passenger cars and light trucks. While a significant increase in rollover percentage occurs for passenger cars between sideslopes of 1V:4H and 1V:3H, the corresponding increase in rollover percentage for light trucks is dramatic. Across all functional classes, the increase in rollovers for passenger cars between sideslopes of 1V:4H and 1V:3H averaged 17 percent. For light trucks, the increase in encroachments resulting in rollover between sideslopes of 1V:4H and 1V:3H averaged 112 percent. This is in contrast to an average increase in light truck rollovers of only 16 percent between sideslopes of 1V:10H and 1V:4H. For many years, a 1V:3H sideslope has been considered to be a traversable but non-recoverable slope, and constitutes the break point for when longitudinal barriers become warranted. NCHRP Project 17-11 suggests that with the steadily increasing percentage of light trucks in the vehicle fleet, the break point for what is considered to be a traversable versus critical sideslope ratio may need to be reassessed for today’s vehicle fleet. Proper assessment of slope traversability and guardrail need will help reduce the number of rollover crashes and associated fatalities. There is a need to develop new guidelines for slope traversability that consider the characteristics of the current vehicle fleet including light trucks. The guidelines should address what constitutes recoverable, traversable, and critical sideslope conditions for today’s vehicle fleet. 2.2 Encroachment Simulation Studies Over a period of many years, researchers have used HVOSM in various studies that involved the safety evaluation of a variety of roadway and roadside conditions, including embankments, ditches, driveways, and culverts (11). These studies have helped define the state-of-the- knowledge in the area of vehicular behavior (lateral distance and stability) when traversing various roadside features. In NCHRP Report 158 (1975), Weaver et al. conducted research to investigate vehicle behavior as it traversed a roadside (12). The objective of this research was to provide guidance in selecting and designing safe roadside configurations. A custom made multi-rigid body dynamics code HVOSM was used to evaluate vehicle dynamics across the roadside. Simulations were conducted for speeds from 40 to 80 mph and encroachment angles of 7, 15, and 25 degrees. Freewheeling and “return-to-road” steering inputs were applied to the vehicle. Terrain slopes were varied from 1V:3H to 1V:10H. To account for the driver’s reaction time, steering input was applied 1.5 s after the vehicle left the pavement. This was followed by a steering input that fully developed in 2 s. The vehicle was then driven for 4.5 s or until it returned to a position 10 ft outside the edge of the pavement. Thus, the vehicle was allowed to travel for a maximum of 8 s after leaving the pavement. Twenty-four full-scale vehicle tests were conducted to validate the results obtained from HVOSM. Even though HVOSM predicted slightly higher accelerations when compared to the full-scale tests, it showed a good correlation with the test data. In the tests performed with

Literature Review 11 a 25-degree encroachment angle, bumper-to-terrain contact and rear overhang drag were observed in all tests above 40 mph. In the absence of a steering and braking input, very little redirection was observed in the vehicle as it traversed a ditch. It was observed that steering input applied while the vehicle is airborne greatly affects vehicle response once the vehicle lands. Sharply turned front wheels induce high side forces on landing and increase wheel digging in the terrain. The high side forces increase the potential of the vehicle to roll over. It was also observed that as a freewheeling vehicle crosses a slope hinge point, it is in a state of instability for a short while. However, once all wheels have landed, the vehicle regains stability. In the absence of humps and depressions, a tire-terrain coefficient of friction 0.6 was reason- able for soil embankments. The path of the vehicle during the return maneuver was found to depend greatly on the tire-terrain coefficient of friction. For a 0.2 friction coefficient, no return maneuvers could be performed successfully. Front slopes flatter than 1V:4H were desired to reduce bumper penetration in the terrain as the vehicle traversed the ditch. In 1975, Ross et al. of TTI performed a study to determine the dynamic behavior of a vehicle as it traversed sloped medians and different curb configurations (13). Texas A&M Transportation Institute’s version of HVOSM was used as a tool to perform parametric simulations with a 4,000 lb vehicle model. This version of HVOSM was modified to incorporate vehicle body-to- terrain contact representation. The objective of these simulations was to determine the potential of a vehicle vaulting over a barrier that was placed behind a curb or in a sloped median. Simulations were performed with 6-inch and 8-inch curbs and with medians that had slopes of 1V:11H, 1V:8H, and 1V:5H. Because of limitations in the computer code, the contact between the vehicle and the barrier was not modeled. Simulations were thus terminated when the vehicle reached the barrier. The performance of the vehicle while traversing existing curb and median configurations was compared to its performance on modified configurations. The comparison was based on vehicular accelerations and bumper height data collected from the HVOSM simulations. Bumper heights were calculated with respect to the local terrain of the vehicle. Simulations were performed assuming a “freewheeling” mode in which no steering, braking, or throttle inputs were applied to the vehicle as it encroached the roadway. It was concluded that in general, barriers should not be placed near curbs. Simulation results indicated that curbs destabilize the vehicle, causing it to vault the barrier, or in some cases impact the barrier at a lower than design impact position. It was suggested that a flat approach area to the barriers should be considered whenever possible. Under NCHRP Project 22-6, the performance of selected highway safety appurtenances and roadside features for passenger vehicles weighing less than 1,800 lb was assessed, and the limits of vehicle characteristics that can be safely accommodated through improvements in current hardware and roadside features were defined. This work is described in NCHRP Report 318: Roadside Safety Design for Small Vehicles (14). The roadside geometric features which were evaluated as part of this study included various fill and cut slopes, ditch sections, and driveways. Typical slopes, fill heights, ditch depths, and cross section combinations were selected for analysis. Embankment parameters investigated included fill slopes of 1V:6H, 1V:4H, and 1V:3H in combination with fill heights of 5, 10, and 20 ft, and cut slopes of 1V:3H, 1V:2H, and 1:1. The matrix of ditch parameters examined included foreslope and backslope combinations of 1V:6H, 1V:4H, and 1V:3H and ditch depths of 5 and 10 ft. The driveway configurations studied included various combinations of foreslopes and driveway slopes of 1V:10H, 8:1, and 1V:6H. Simulations of vehicular encroachments across these roadside features were made with the HVOSM program, which was first calibrated using the results of full-scale crash tests with small

12 Guidelines for Traversability of Roadside Slopes vehicles. In most of the computer runs, the vehicle was directed off the travelway in a tracking condition at an encroachment angle of 15 degrees and an encroachment speed of 60 mph. A limited number of non-tracking encroachment simulations were also conducted during which the vehicle was given a 15 degrees/s yaw rate as it left the travelway. At a time approximating the perception-reaction time for an average driver, the vehicle was given a steer input to simulate the response of a driver in a panic situation trying to return to the road. For each simulation run, vehicular response in terms of lateral movement, roll and pitch angles, and resultant accelerations was summarized. In addition, the behavior of the vehicle was quantified as either stable, side-slipping, spinning out, or overturning. Under NCHRP Project 17-11, TTI developed relationships between recovery area distance and roadway and roadside features, vehicle factors, encroachment parameters, and traffic conditions for the full range of highway functional classes and design speeds that can sub- sequently be used to establish clear-zone guidelines (10). It was recognized that the use of crash data for determining the statistics on the extent of lateral movement of vehicles encroaching onto the roadside is often limited by a vehicle striking a fixed object or rolling over. Therefore, any lateral extent of encroachment distribu- tion derived from crash data will be a truncated distribution, and the full effect of sideslopes and other variables on lateral extent of encroachments is only partially observed. A research approach that combined crash data analyses with computer simulation results was developed to overcome this limitation. Use of computer simulation permitted a detailed analysis of vehicle trajectory and resulting vehicle kinematics for a wide range of variables for which data was not otherwise available. When combined with real-world crash data, the results can be used to develop relationships between various encroachment parameters. Clinical reconstruction and analyses of NASS CDS data was conducted to investigate key encroachment parameters for ran-off-road crashes including encroachment speed, encroach- ment angle, vehicle orientation at encroachment (i.e., tracking, non-tracking), and driver control input (i.e., steering, braking, or both). Three years of NASS CDS crash cases from 1997 to 1999 were considered in this study. A total of 559 cases were selected for the study based on an adopted set of criteria. A supplemental data collection effort was planned and executed to obtain desired roadway and roadside characteristics associated with the crash sites. The weighted crash data was used to develop probability distributions for the key encroach- ment parameters for nine highway functional classes. The functional classes were defined in terms of roadway type, land use, and posted speed limit. The distributions were then segregated into prescribed value categories for purposes of developing probability input matrices for application to the simulation data. The computer simulation study was performed using the HVOSM computer simulation program. The variables included in the computer simulation study include: vehicle type, encroachment speed and angle, vehicle orientation, driver input, horizontal curvature, vertical grade, shoulder width, foreslope ratio, foreslope width, ditch width, backslope ratio, backslope width, and tire-terrain friction. A simulation matrix of all possible combinations of these variables would consist of literally millions of simulation runs. Because this was obviously not practical, it was necessary to prepare a selective simulation matrix. The approach utilized was to categorize the variables into two groups: baseline variables and adjustment variables. The baseline variables are control variables from which a basic set of lateral extent of encroach- ment relationships can be obtained. The adjustment variables are additional variables that are evaluated independently with additional simulation runs to provide information regarding the effects of specific roadway or roadside factors on lateral extent of encroachment or vehicle

Literature Review 13 stability. For example, the effect of horizontal curvature can be quantified by comparing the lateral extent of encroachment obtained for a curved section of roadway with that for a tangent section of roadway having similar roadside characteristics. The effects of these adjustment variables can be presented in the form of factors that can be applied to a basic set of relationships rather than requiring the development of numerous sets of relationships. The baseline variables and their selected values included: • Vehicle type (820-kg passenger cars, 1500-kg passenger sedans, 2000-kg pickup trucks, and small utility vehicles), • Encroachment speed (50, 70, 90, and 110 km/h), • Encroachment angle (5, 15, and 25 degrees), • Driver control response (steering and combined steering and braking), • Foreslope ratio (flat, 1V:10H, 1V:6H, 1V:4H and 1V:3H), • Roadside coefficient of friction (longitudinal/lateral–0.5/1.2), and • Vehicle orientation (tracking and non-tracking with yaw rate of 15 degrees/s). Note that under the baseline conditions, the roadway was assumed to be straight and level and the foreslope was assumed to extend indefinitely (i.e., no ditch or backslope). Each adjustment variable was analyzed individually, not in combination with other adjust- ment variables. Selected values for each adjustment variable were evaluated with the same set of encroachment parameters used for the baseline simulations. This permits a more direct comparison of the baseline and adjustment variable simulation results as well as the use of the same probability distributions. The adjustment variables and their values included: • Horizontal curvature (3, 6, and 9 degrees), • Vertical grade (3 and 6 percent downgrade, and 3 and 6 percent upgrade), • Shoulder Width (0.6, 1.8, and 3.6 m), • Ditch Configuration – Foreslope width (4, 8, and 12 m), – Ditch Width (1.0 and 3.0 m), – Backslope ratio (1V:6H, 1V:4H and 2.5:1), and – Backslope Width (6 and 12 m). The ranges used for these variables are generally intended to comprise current design practice for the classes of roadways being investigated. In order to establish functional relationships from the discrete simulation data points, the probability distributions developed from the weighted NASS CDS data were applied to each encroachment parameter to obtain a probability for each value category of that parameter used in the simulation matrix. The combined probability for a given simulation with a unique set of encroachment condi- tions was determined by multiplying the individual probabilities assigned to the value of each encroachment parameter. The probability that an encroaching vehicle will have a certain range of lateral movement is simply the probability sum of simulated encroachments that have lateral movements within that range. In this way, exceedance curves were developed and used to create lateral extent of movement relationships that combine simulation and real-world crash data such that they are a function of multiple encroachment parameters. Following this procedure, relationships for lateral extent of movement were developed for each functional class in terms of foreslope ratio. For a given functional class and foreslope ratio,

14 Guidelines for Traversability of Roadside Slopes the exceedance curves can be used to determine the percentage of encroachments that will exceed a certain lateral distance. The simulation runs conducted for the adjustment variables were analyzed in the same manner as the baseline runs. The exceedance curves developed for each value of the adjust- ment variables were compared to those of the baseline simulations to develop adjustment factors. The adjustment factors can be applied to the baseline exceedance curves to account for the effect of adjustment variable on lateral extent of encroachment. The simulation results were also analyzed for purposes of developing encroachment severity relationships for the different roadway and roadside variables of interest. Such relationships can be used for determining accident costs associated with roadside encroachments. Vehicular resultant accelerations and angular displacements were captured as output from the simulations and used to help assess encroachment severity. The simulation results were used to compute rollover probability as a function of sideslope ratio. The injury severity for the rollover encroach- ments was estimated using six years of NASS CDS data. Non-rollover encroachments were treated separately as a function of maximum resultant vehicle accelerations obtained from the computer simulations. The relationships developed under this study provided data and relationships from which recovery area guidelines can be developed for selected highway functional classes. The researchers recommended that future guideline development processes will involve some form of cost- effectiveness procedure and/or utilize a benefit-cost analysis program such as RSAP. The lateral extent of encroachment relationships developed under this study can be used to update accident frequency or rate as a function of recovery distance and sideslope. The severity relationships can be used to update accident severity in the RSAP model expressed in terms of injury probability and vehicle stability (i.e., rollover probability). This approach for developing guidelines for roadside clear zones is currently being carried out under NCHRP Project 17-11(2). The research work plan is organized into two phases with a total of twelve tasks. The first phase primarily involves the development of relationships, analysis tools, and data for subsequent use in Phase II. Phase II of the project focuses on development of clear recovery area guidelines using a benefit-cost analysis approach. In a study for the Texas DOT, TTI researchers developed supporting data from which clear- zone guidelines could be established as a function of embankment side slope for rural arterial types of roadways (15). In this study, a set of design encroachment conditions was selected using data from pole- and bridge-related accidents (16, 17, 18). The impact speed and angle data from the accident studies were used to establish 85th percentile design encroachment condi- tions. Because it is not possible to select a unique percentile value for a combination of variables such as speed and angle, the effects of both variables were taken into account by considering the lateral component of vehicular velocity as shown below. Vlat = V sinq where V = encroachment speed, and q = encroachment angle. The lateral component of vehicular velocity not only provides a unique percentile value, but is considered to be a good indicator of the potential for lateral movement of an errant vehicle. In order to take unreported accidents into account, the percentile of reported accidents was adjusted using a conservative 1 to 1 ratio of reported to unreported accidents. Thus, the

Literature Review 15 85th percentile design encroachment conditions for both reported and unreported accidents were equivalent to the 70th percentile impact conditions for reported accidents. Using this approach, the lateral speeds for 53 accidents on rural arterial highways were calculated and the resulting data was fitted with a logistic normal distribution. Based on this theoretical distribution, the 70th percentile of lateral speed for reported accidents was found to be 9.74 mph, from which design encroachment conditions were determined for speeds of 45, 50, and 55 mph. For these design encroachment conditions, the HVOSM computer simula- tion program was used to determine the response of large and small cars for a selected range of roadside slope conditions and a selected set of driver steer inputs. The results were used to help define appropriate clear-zone distances for rural arterial roadways with embankment slopes of 1V:6H, 1V:4H, and 1V:3H. In 1993, Ross and Bligh conducted a research study for the Minnesota DOT to investigate the benefits of slope rounding the hinges at the intersections of shoulders and side slopes (19). Analysis was performed to investigate vehicle dynamics of a vehicle as it traversed different roadside shoulder and side slope configurations. HVOSM was used to perform this simulation analysis. Vehicle encroachment speeds of 45 and 65 mph, at encroachment angles of 5, 15, 25, 35, and 45 degrees, were investigated. Side slopes of 1V:6H, 1V:4H, and 1V:3H were investigated with a shoulder and roadway cross slope of 21V:5H and 50H:1V, respectively. All simulations were performed with a Honda Civic vehicle with a “return-to-road” steering input of 8 degrees at the wheels. It was determined that the body-to-terrain contact was an important factor for this analysis. Therefore, a TTI version of HVOSM (V-3), which provided this type of contact, was used. Curved roadways and V-ditches were not considered in this research and all roadside ditches were assumed to have flat bottoms. Vehicle overturns were predicted for rounded and unrounded side slopes for several combi- nations of encroachment speeds, angles, and ditch configurations. It was determined that high cornering forces due to panic steering, in combination with body-to-terrain contact, resulted in significant vehicular instability. For many cases, this instability led to vehicle overturns. Occupant risk was evaluated for each simulation and a severity index was assigned for the benefit/cost analysis. The benefit/cost analysis was used to develop guidelines for identifying scenarios when slope rounding becomes beneficial. Thomson and Valtonen conducted a series of tests to document the performance of passenger car vehicles once they enter a V-ditch at various conditions (20). The conduction of these tests was part of a joint Swedish-Finnish test study aimed at investigating the influence of impact angle and speed and the effects of driver steering on vehicle dynamics after entering a V-ditch. Entering angles between 5 and 20 degrees and speeds between 80 and 110 km/h were chosen for the test matrix. The ditch configuration used for the full-scale crash tests was a 1H:3V foreslope and a 1H:2V backslope. The ditch width was 5 m for all except two tests, and the backslope was extended 1 m above level ground. Two tests were conducted with modified ditch bottoms: for one test, a U-shaped ditch bottom was chosen, while for the other test a low concrete barrier (40 cm high) was placed on the backslope at about 50 cm from the V-ditch bottom. In one test, a pneumatic actuator was used for steering rotation input and the steering wheel was locked at the requested position. Thomson and Valtonen preferred to maximize the number of tests performed by limiting the instrumented vehicle number. According to them, the dynamic of the vehicle entering the ditch could have been reasonably and sufficiently observed from the videos. Vehicle trajectory (including any rollover event) was collected for later comparison and calibration method for computer simulations.

16 Guidelines for Traversability of Roadside Slopes A total of four rollovers were observed for the impact conditions investigated: two rollovers happened with a free running vehicle, one was the result of steering input while the vehicle was in the ditch, and one was caused by impact with the low barrier placed in the bottom of the ditch. From the tests, it was observed that small cars rolled over when entering the ditch at 10 degrees and 80 km/h. Also, it appeared that the major factor inducing vehicle rollover in the V-ditch was vehicle contact with the backslope. Simulations of the vehicle impacts were not fully capable of reproducing the vehicle motion through the ditch observed in the tests. The authors suggested inclusion of the frame-to-ground contact in the simulation model to better replicate the vehicle motion in the ditch. Also, they identified the influence of castor on the vehicle steer motions to be another reason for the simulations not to replicate the vehicle motion recorded with the tests. This apparently caused simulations to under-predict the lateral vehicle motion in the ditch. Critical issues identified by the researchers from these tests were the vehicle steering and suspension characteristics which lead ultimately to castoring of the vehicle, and the need to adequately numerically model the tire-soil contact, for a more realistic understanding of the tire lateral loads. The ongoing NCHRP Project 16-05 has the objective of developing guidelines for cost- effective treatments of roadside ditches in order to reduce the severity of ditch crashes. As part of this project, the researchers conducted a survey of state DOTs and Canadian provincial transportation agencies to gather information about typical roadside ditch configurations and to identify innovative treatments that may not have been previously documented. A total of 42 states and four Canadian provinces responded to the survey. Results of this recent survey provide valuable insight on the current state of practice in roadside ditch design and will help the TTI research team in selecting simulation analysis parameters such as typical roadside foreslope widths and surface type. According to the preliminary results of the survey, the most common foreslope is 1V:6H and the least common is 1V:10H or flatter. The use of 1V:4H and 1V:3H foreslopes is also prevalent. The use of 1V:2H slope is greater than the 1V:10H slope. Figure 2.3 presents the breakdown of foreslope usage. The survey results also indicated that the most common surface treatment on the roadside foreslopes is vegetation, followed by turf. Such information will be used by the researchers in determining the surface properties of the foreslope during simula- tion analysis. The survey also indicated that slope rounding is not a common practice among the states and provinces. Only 21 percent round the top break point and 26 percent round the bottom of the ditch more than 50 percent of the time. The researchers of NCHRP Project 16-05 also performed an extensive crash database analy- sis using FARS, NASS CDS, NASS GES, NCHRP Report 665, and Highway Safety Information System databases. This analysis focused on identifying any trends in the type and severity of roadside ditch-related crashes and their relationship to ditch geometry, roadway characteristics, vehicle type, presence of appurtenances, and other relevant characteristics. While the crash data analysis proved useful in identifying several trends in crashes related to ditches, the databases did not contain sufficient information on the roadside ditch configurations. Because of this lack of information on ditch design, various ditch design parameters could not be reliably connected to vehicle crashes or rollovers on slopes. Currently, the researchers of the ongoing NCHRP Project 16-05 study are working on an extensive simulation study using CarSim. The results of the simulation analyses will be used to evaluate a large number of roadside ditch parameters, in combination with various driver inputs and terrain conditions.

Literature Review 17 2.3 Vehicle Testing with Remote Control Driver Input Texas A&M Transportation Institute has conducted two projects for NHTSA and the U.S. DOT where remote controlled input was applied for vehicle control and maneuvering. In the late 1960s, NHTSA developed a cooperative program with TTI aimed at comparing the handling and dynamic stability characteristics of selected small passenger cars sold during the period of 1960 through 1967 (21). Five test driving maneuvers were chosen with the intent of providing a basis for comparing the handling characteristics of six automobiles included in the program. The driving maneuvers were designed to incorporate steering, steering plus braking, and steering on a rough surface. In four of the five tests selected, a live professional driver was used to apply the driving inputs. In one of the maneuvers, a specialized controller was used to remotely apply driving inputs at precise timings. A summary of each applied maneuver is reported below. • Steady turn braking, smooth road. This test consisted of braking a vehicle which initially was in a steady-state turn of 0.3 g at 30 mph. This type of maneuver simulated an emergency situation in which a vehicle being driven through a curve must be brought to a stop with little or no deviation from its intended path. • Steady turn, rough road. In this test the vehicle was placed in a 0.3 g steady-state turn at 30 mph before encountering a series of bumps simulating a rough road. The purpose of this test was to simulate the real-world situation where a driver who is cornering his vehicle at a fairly high lateral acceleration suddenly encounters a series of bumps or other random road disturbances. The test is to be carried out at constant throttle to simulate a condition when the driver does not have adequate time to react before striking the disturbances. • Step steer. This maneuver consisted of suddenly applying a steering input to a vehicle traveling in a straight line at 30 mph. This test was repeated with progressively larger steering inputs to the point of maximum steering wheel displacement as dictated by the mechanical steering Figure 2.3. Survey results of typical roadside foreslopes.

18 Guidelines for Traversability of Roadside Slopes properties of the vehicle. This simulated a real-world avoidance maneuver situation where an extreme turn is required suddenly. • Reverse steer input. In this maneuver the vehicle was put into a steady-state turn which was interrupted by a sudden steering input applied in the opposite direction. It represented an evasive maneuver using a second steering input to change the lane suddenly. • Drastic steer, drastic brake. This maneuver utilized an automatic vehicle controller rather than a driver because the performance of the maneuver depends upon the critical timing of events based on measured response parameters as well as programmed inputs. The controller was developed at the Highway Safety Research Institute. The Highway Safety Research Institute of the University of Michigan was the primary developer of the Input-Response testing methods. Passive devices were incorporated to con- strain driver steering and breaking inputs for relatively simple step type control inputs. In addition, an automatic vehicle controller was developed and utilized to provide precise control inputs. TTI refined the test apparatus from the University of Michigan to ensure that precise steer- ing and braking inputs were applied during severe maneuvering and to provide a record of selected vehicle response variables. The remote controlled automatic vehicle controller included an adjustable steering stop device to permit precision limiting of steering stroke and an adjust- able brakeline pressure limiting system to permit precision limiting of braking input. The automatic vehicle controller used in this study was designed to examine vehicle response to complex control input time histories. The controller had different operational modes to pro- vide inputs to the steering wheel, brake pedal, and accelerator, and was designed to be installed in any passenger car used in the study. A block diagram of the automatic vehicle controller used in this study is reported in Figure 2.4. Figure 2.4. Block diagram of automatic vehicle controller.

Literature Review 19 In the early 1980s, TTI conducted a series of embankment traversal tests as part of a project aimed at evaluating the performance limits of guardrails, median barriers, and embankments for different classes of vehicles and impact conditions (22). Prior to the performance of the tests, simulations of embankment traversals for several types of vehicles and traversal conditions were performed using the HVOSM computer program. The test conditions and embankment geometry used in the full-scale tests were selected on the basis of the simulation results. The scope of this part of the study was to identify and select an embankment slope that would be at or near the limit of steepness for traversable slopes and to demonstrate the behavior of vehicles traversing such a slope. The tests were conducted on a two-lane highway with paved shoulders where the main portion of the side slope was a 1V:3H slope and the height of the embankment was approxi- mately 15 ft. Three full-scale embankment traversal tests were performed using a 1979 pickup, a 1979 3/4-ton van and a 1979 Honda Civic passenger car. Each vehicle was controlled by a proportional radio control system which allowed the vehicle to be driven toward the test site at a speed of 50 mph and to be maneuvered by remote actuation of steering, accelerator, and brakes. A programmable maneuver generator was available in the test vehicle. However, manual remote-controlled input was preferred to allow for compensation for crosswind, road crown, and longitudinal position. In each test, steering input was remotely applied to the vehicle so that it would leave the edge of the roadway at 15 degrees and a second steer input was then remotely initiated with the intent to drive the vehicle back to the roadway. For each test, vehicle tire paths were recorded through the slope to allow for reconstruction of vehicle dynamics. This study concluded that smooth, well-compacted roadside slopes as steep as 1V:3H can be traversed safely. However, vehicle stability on such slopes is vulnerable and small terrain discontinuities can have an impact on the vehicle’s trajectory.