Design of Interchange Loop Ramps and Pavement/Shoulder Cross-Slope Breaks (2017)

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.

Part 2: Assessment of Design Criteria for Pavement/Shoulder Cross-Slope Breaks

2 - ii ACKNOWLEDGMENT The research reported herein was performed under NCHRP Project 3-105, “Design Guidance for Interchange Loop Ramps” as Phase III – Assessment of Pavement/Shoulder Cross-Slope Breaks. While this assessment was conducted as part of a larger research project that focused on the design of interchange loop ramps, the assessment of pavement/shoulder cross-slope breaks concerned the design of cross-slope breaks on roadways in general and was not limited just to interchange loop ramps. This report was prepared by Dr. Darren J. Torbic, Ms. Lindsay M. Lucas, and Mr. Douglas W. Harwood of MRIGlobal, and Mr. Marcus A. Brewer, Mr. Akram Abu-Odeh, Ms. Elizabeth Depwe, and Ms. Kimberly Rau of the Texas A&M Transportation Institute. Ms. Karin M. Bauer, Mr. Chris A. Fees, and Mr. John J. Ronchetto of MRIGlobal, and Mr. Michael P. Pratt and Mr. Dan Walker of the Texas A&M Transportation Institute played key roles in this research. The authors wish to thank the State Departments of Transportation of Iowa and Kansas for their assistance in this research.

2 - iii ABSTRACT The objectives of this research were to 1) assess AASHTO’s current design policy for pavement/shoulder cross-slope breaks to determine whether updates in design criteria are needed, and 2) if updates are needed, to specify recommended changes to AASHTO’s current design policy. AASHTO’s current design policy states that shoulder slopes that drain away from the paved surface on the outside of a superelevated horizontal curve should be designed to avoid too great a cross-slope break—calculated as the algebraic difference between the cross-slopes of the traveled way and the shoulder—and the cross-slope break should be limited to approximately 8 percent by flattening the shoulder on the outside of the curve. Two separate work plans, one involving vehicle dynamics simulation modeling and a second involving a crash-based safety analysis, were executed to fulfill the objectives of this research. Based upon the results from the vehicle dynamics simulation modeling, it was concluded that there is no need to recommend a change to AASHTO’s current design policy for pavement/shoulder cross-slope break on the outside of horizontal curves.

ACKNO ABSTRA LIST OF SUMMA SECTION SECTION SECTION SECTION SECTION SECTION APPEND APPEND WLEDGEM CT ............ FIGURES RY ............ 1. Introd 1.1 B 1.2 R 1.3 R 1.4 O 2. Summ 2.1 C 2.2 D 2.3 C 2.4 S 2.5 S 3. Vehic 3.1 O 3.2 S 3.3 I 4. Crash 4.1 E 4.2 A 4.3 I 5. Concl 6. Refere SimuIX A. SimuIX B. ENT ......... ................... AND TABL ................... uction ........ ackground esearch Obj esearch Ap utline of Re ary of Liter ross-Slope river Behav ritical Desi ummary of ummary of le Dynamics verview of ummary of nterpretation -Based Safe xamination nalysis of C nterpretation usions and F nces .......... lation Resu lation Resul .................... .................... ES.............. .................... .................... .................... ective ........ proach ........ port ........... ature Review Break Resea ior Studies . gn Elements State DOT C Key Issues . Simulation Vehicle Dy Simulation of Vehicle ty Analysis . of FARS C rashes on H of Crash-B uture Resea .................... lts for Full T ts for Mode 2 - iv ................... ................... ................... ................... ................... ................... ................... ................... ................... and Curre rch ............. ................... and Variab ross-Slope ................... Modeling .. namics Simu Results ....... Dynamics S ................... rash Data .... orizontal C ased Safety rch Needs .. ................... raversal Sev rate Departu ................... ................... ................... ................... ................... ................... ................... ................... ................... nt Practice .. ................... ................... les .............. Break Desi ................... ................... lation Mod ................... imulation M ................... ................... urves .......... Analysis .... ................... ................... ere Departu re Vehicle T ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... gn Policy an ................... ................... eling Appro ................... odeling Re ................... ................... ................... ................... ................... ................... re Vehicle rajectories ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... ................... d Practice .. ................... ................... ach ............. ................... sults ........... ................... ................... ................... ................... ................... ................... Trajectory .. ................... ..... ii  .... iii  ......v  ... vii  ......1  ......1  ......3  ......3  ......4  ......5  ......5  ......6  ......8  ......9  ....14  ....15  ....15  ....29  ....33  ....36  ....36  ....38  ....46  ....48  ....50  . A-1  ..B-1  Part 2 Contents

2 - v LIST OF FIGURES AND TABLES Figures Figure 1. Typical Superelevated Cross-Sections with Pavement/Shoulder Cross-Slope Breaks ...........................................................................................................................2  Figure 2. Observed Maximum Steering Angles from SAE Study ...............................................7  Figure 3. General Simulation Process .........................................................................................15  Figure 4. Collision Avoidance Maneuver ..................................................................................18  Figure 5. Vehicle Paths Generated by Path Function for Collision Avoidance Maneuver (Severe Departure) .......................................................................................................19  Figure 6. Vehicle Path with 10.40 ft Lateral Displacement for Collision Avoidance Maneuver (Full Traversal Severe Departure) ..............................................................20  Figure 7. Moderate Vehicle Departure Model ...........................................................................21  Figure 8. Driver Path/Trajectory Moderate Departure on Road Radius of 250 ft .......................21  Figure 9. Roll Angle Sign Convention in TruckSim ...................................................................23  Figure 10. Roll Angle of Tractor Unit on a Tangent .....................................................................24  Figure 11. Roll Angle of Trailer Unit on a Tangent ......................................................................24  Figure 12. Illustration of Maximum Roll Angle of Tractor/Single-Van-Trailer Truck on a Tangent from TruckSim ...............................................................................................25  Figure 13. Roll Angle of Tractor Unit at 50 mph on Minimum Radius Curve with 4% Superelevation..............................................................................................................26  Figure 14. Roll Angle of Trailer Unit at 50 mph on Minimum Radius Curve with 4% Superelevation..............................................................................................................26  Figure 15. Illustration of Maximum Roll Angle of Tractor/Single-Van-Trailer Truck from TruckSim at 50 mph on Minimum Radius Curve with 4% Superelevation ................27  Figure 16. Cross-Slope Break vs. Location of Break by Roadway Type and State ......................42 

2 - vi Tables Table 1. Primary Vehicle Input Parameters Used in TruckSim .................................................16  Table 2. Specified Minimum Design Radius .............................................................................17  Table 3. Summary of TruckSim Results for a Tractor/Single-Van-Trailer Truck on a Tangent with 0% Superelevation .................................................................................23  Table 4. TruckSim Results for Tractor/Single-Van-Trailer Truck Traversing Minimum Radius Curves with 4% Superelevation .......................................................................25  Table 5. TruckSim Results for Tractor/Single-Van-Trailer Truck Traversing Minimum Radius Curves with 6% Superelevation .......................................................................27  Table 6. Summary of TruckSim Results for Tractor/Single-Van-Trailer Truck Traversing Minimum Radius Curves with 8% Superelevation ...................................27  Table 7. Tractor/Single-Van-Trailer Truck with Partial Traversal Moderate Departure ...........30  Table 8. Tractor/Single-Van-Trailer Truck with Full Traversal Moderate Departure ...............30  Table 9. Tractor/Tanker-Trailer Truck with Partial Traversal Moderate Departure (Full Tanker Trailer) .............................................................................................................32  Table 10. Tractor/Tanker-Trailer Truck with Full Traversal Moderate Departure (Full Tanker Trailer) .............................................................................................................32  Table 11. Tractor/Double-Van-Trailer Truck with Full Traversal Moderate Departure .............34  Table 12. Selected FARS Statistics of Fatal Truck and Tanker Truck Rollover Crashes ...........37  Table 13. FARS Data: Distribution of Untripped Fatal Single-Vehicle Truck and Tanker Truck Rollover Crashes by Alignment .......................................................................38  Table 14. Range of Site Characteristics of Horizontal Curves Included in Crash-Based Safety Analysis ............................................................................................................41  Table 15. Crash Counts by Crash Type, Roadway Type, and State ............................................43  Table 16. ANOVA Results—Statistical Significance Levels Associated with Factors in Negative Binomial Regression Models .......................................................................45 

2 - vii SUMMARY In 2011, the National Transportation Safety Board (NTSB) completed a detailed investigation of a crash involving a tractor/tanker-trailer truck that overturned and caught fire while negotiating a curved interchange ramp. Among the results from the investigation, the NTSB concluded that the transition from a positive to a negative cross-slope, as the truck moved laterally from the right lane of the traveled way onto the shoulder, reduced the speed at which the truck could negotiate the curve without rolling over. Therefore, the NTSB suggested a review of the American Association of State Highway and Transportation Officials’ (AASHTO’s) design policy for pavement/shoulder cross-slope breaks (i.e., changes in cross-slope at the outside edge of the traveled way) on horizontal curves to determine whether updates to the design criteria are needed. AASHTO’s current design policy states that shoulder slopes that drain away from the paved surface on the outside of a superelevated horizontal curve should be designed to avoid too great a cross-slope break, calculated as the algebraic difference between the cross-slopes of the traveled way and the shoulder. To avoid large pavement/shoulder cross-slope breaks, the current AASHTO design policy states that the cross-slope break should be limited to approximately 8 percent by flattening the shoulder on the outside of the curve. Based on the NTSB’s recommendation, the objectives of this research were to 1) assess AASHTO’s current design policy for pavement/shoulder cross-slope breaks on the outside of horizontal curves to determine whether updates in design criteria are needed, and 2) if updates are needed, to specify recommended changes to AASHTO’s current design policy. Two separate work plans, one involving vehicle dynamics simulation modeling and a second involving a crash-based safety analysis, were executed to fulfill the objectives of this research. In the first work plan, conducted through vehicle dynamics simulation modeling, combinations of key geometric design elements and other critical elements were evaluated to assess their impact on vehicle stability when encountering a range of cross-slope breaks between the traveled way and shoulder. The vehicle, vehicle trajectory, and geometric design factors that were varied and assessed in this study included:  Vehicle type  Vehicle speed  Vehicle trajectory  Cross-slope break rate  Superelevation rate  Extent of vehicle encroachment The primary measures of interest that were obtained as output from the vehicle dynamics simulation modeling were whether the vehicle recovered from the maneuver or rolled over and the maximum roll angles experienced by the vehicle. The previous research on which AASHTO’s current design policy for pavement/shoulder cross-slope breaks is based considered only a passenger car, since vehicle dynamics simulation models for articulated trucks were not available at that time. In the current research, three types of articulated trucks were simulated: a tractor/single-van-trailer truck, a tractor/tanker-trailer truck, and a tractor/double-van-trailer truck. For all three vehicle types, the target weight of the vehicle was approximately 80,000 lb. For the tractor/tanker-trailer truck, existing vehicle dynamics simulation models do not have the capability to simulate the dynamic effects of liquid sloshing in a tanker trailer. Therefore, only the fully-loaded trailer condition was selected for evaluation of pavement/shoulder cross-slope breaks as it was assumed that the dynamic effects of liquid sloshing should be minimal for the

2 - viii fully-loaded trailer condition. Since the dynamic effects of liquid sloshing in a tractor/tanker- trailer truck could not be simulated in this research, the tractor/tanker-trailer truck was simulated by modifying the center of gravity of the tractor/single-van-trailer truck to represent the performance of a tractor/tanker-trailer truck within the simulation model. Vehicle trajectory was also found to be a primary factor in a vehicle’s capability to recover from an encounter with a pavement/shoulder cross-slope break. In total, three vehicle trajectories were simulated: a full traversal severe departure, a partial traversal moderate departure, and a full traversal moderate departure. The full traversal severe departure vehicle trajectory represents a situation in which the driver steers to avoid an obstacle in the travel lane (i.e., a collision avoidance maneuver) while traversing a horizontal curve and all tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. This severe departure maneuver represents an emergency collision avoidance maneuver so extreme that trucks are likely to rollover in some cases even if no pavement/ shoulder cross-slope break is present. This extreme maneuver was simulated to verify that the vehicle dynamics simulation model was functioning properly and that vehicle rollovers did result in some cases. The partial traversal moderate departure vehicle trajectory and the full traversal moderate departure vehicle trajectory scenarios represent maneuvers in which a vehicle would be unlikely to rollover in the absence of a pavement/shoulder cross-slope break; these scenarios were evaluated to assure that the presence of a pavement/shoulder cross-slope break did not itself lead to a vehicle rollover. For the partial traversal moderate departure vehicle trajectory, the vehicle gradually drifts from the middle of the travel lane along a path tangent to the roadway curvature and only the passenger-side tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. For the full traversal moderate departure vehicle trajectory, the vehicle also gradually drifts from the middle of the travel lane along a path tangent to the roadway curvature but, with this scenario, all tires of the vehicle (i.e., the tires on both sides of the vehicle) traverse the cross- slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. Thus, the partial traversal moderate departure vehicle trajectory (in which only the passenger- side tires left the traveled way) represented the mildest departure scenario that was simulated, and the full traversal moderate departure vehicle trajectory (in which all tires on both sides of the vehicle left the traveled way) was more severe. The partial traversal and full traversal moderate departure vehicle trajectories represent realistic situations on which design of pavement/shoulder cross-slope breaks should be based. The appropriate design criterion is that a pavement/shoulder cross-slope break should not induce rollover by a truck in a situation in which rollovers do not occur in the absence of a pavement/shoulder cross-slope break. The full traversal moderate departure vehicle trajectory is the same vehicle trajectory used in the previous research that served as the basis for AASHTO’s current design policy on pavement/shoulder cross-slope breaks. By simulating a range of vehicle trajectories, the sensitivity of vehicle dynamics responses could be observed to assess the impact that vehicle trajectory has on a vehicle’s capability to recover from an encounter with a pavement/shoulder cross-slope break. The partial and full traversal

2 - ix moderate departure vehicle trajectories were selected as the primary maneuvers of interest for assessing the design criteria for pavement/shoulder cross-slope breaks and drawing conclusions related to the need to change AASHTO’s current design policy for pavement/shoulder cross- slope breaks. Based upon the simulation results for the tractor/single-van-trailer truck, the tractor/double-van- trailer truck, and the fully-loaded tractor/tanker-trailer truck for the partial and full traversal moderate departure vehicle trajectories, there is no evidence to suggest the need to reduce the threshold value of 8 percent as the maximum recommended cross-slope break. Of the three truck types evaluated, none rolled over in any of the simulation scenarios for the partial traversal moderate departure vehicle trajectory, nor the full traversal moderate departure vehicle trajectory, even when the cross-slope break was as high as 10 percent. It was only in the simulation scenarios for the full traversal severe departure vehicle trajectory (i.e., most extreme departure scenario simulated) that these three vehicle types did not recover in some instances when encountering a pavement/shoulder cross-slope of 8 percent or less, and this represents a situation in which a truck could rollover even in the absence of a pavement/shoulder cross-slope break. The second work plan involved a crash-based safety analysis to investigate the extent to which large cross-slope breaks may contribute to crashes near superelevated horizontal curves. There are no crash databases available that provide data pertaining to the presence, absence, or magnitude of pavement/shoulder cross-slope breaks at crash sites. However, some information about the potential role of pavement/shoulder cross-slope breaks in crashes can be determined from interpreting crash analysis results. Data from the Fatality Analysis Reporting System (FARS) were initially analyzed to gather general information on where fatal single-vehicle truck rollover crashes typically occur and the frequency of such crashes. FARS data show that untripped single-vehicle rollover truck crashes are much more likely to occur on curves than on tangents. The increased likelihood of rollover crashes on curved roadway sections, where larger cross-slope breaks are more likely to occur, suggests that a more detailed analysis of the effect of pavement/shoulder cross-slope break on crash frequency and severity was warranted. Because existing crash databases do not include quantitative information on the presence, absence, or magnitude of pavement/shoulder cross-slope breaks, a safety analysis was conducted for a small sample of horizontal curves at which the pavement/shoulder cross-slope breaks were measured in the field. Roadway design and crash data were collected for a sample of horizontal curves covering a range of superelevations and cross-slope breaks collected in four states—Iowa, Kansas, Texas, and Washington. A total of 64 curves on rural two-lane highways and 44 curves on rural freeways were included in the analysis. The safety effect of cross-slope break on curves was estimated separately for rural two-lane highways and rural freeways and for three crash types: total, single-vehicle, and single-vehicle rollover crashes. Models were estimated using a generalized linear model approach assuming a negative binomial distribution of crash counts using the combined crash data and selected roadway geometrics. No statistically significant effect of pavement/shoulder cross-slope breaks on any of the three crash types was found; however, the four states’ datasets were too small and too disparate to warrant drawing meaningful conclusions.

2 - x Based on the results from the vehicle dynamics simulation modeling, it was concluded that there is no need to recommend a change to AASHTO’s current design policy for pavement/shoulder cross-slope breaks on the outside of horizontal curves. There is some evidence to suggest that the recommended maximum cross-slope break could be increased to 10 percent. In particular, this may be possible for curves with higher superelevations, as vehicle dynamics simulation modeling showed that maximum roll angles decreased with increasing superelevation. However, this potential change to AASHTO’s design policy is not recommended as the current design policy with a maximum of 8 percent cross-slope breaks is the more conservative approach and also because the scenario for tanker trucks with sloshing liquid could not be evaluated. The research results confirm AASHTO’s current maximum value of 8 percent for design of pavement/shoulder cross-slope breaks on the outside of horizontal curves. While no changes to AASHTO’s current design policy for pavement/shoulder cross-slope break are recommended, use of mitigation measures such as edgeline or shoulder rumble strips that reduce the potential for full traversal departure onto the shoulder in high pavement/shoulder cross-slope break locations are recommended. Measures that reduce the likelihood of full traversal departures and limit encroachments onto the shoulder to partial traversal departures will reduce the maximum roll angles experienced by errant vehicles that encroach onto the shoulders. When installed, rumble strips should be placed on or close to the edgeline rather than further out onto the shoulder to reduce the likelihood of full traversal departures. Future research is recommended to more accurately incorporate tractor/tanker-trailer trucks in the vehicle dynamics simulation analyses; it would also be desirable to substantially increase the size of the dataset used in the detailed crash-based safety analysis so the relationship between cross-slope break and single-vehicle truck rollover crashes can be more fully evaluated.

2 - 1 SECTION 1. Introduction 1.1 Background On October 22, 2009, a tractor/tanker-trailer truck overturned and caught fire while negotiating a semi-direct connection ramp between two interstate highways near Indianapolis, Indiana. The National Transportation Safety Board (NTSB) conducted a detailed investigation of the crash (NTSB, 2011). Among the results from the investigation, the NTSB concluded that the transition from a positive to a negative cross-slope, as the truck moved laterally from the right lane onto the shoulder, significantly decreased the speed at which the truck could negotiate the curve without rolling over. Furthermore, NTSB stated that guidance on pavement/shoulder cross-slope breaks in the 2004 edition of the American Association of State Highway and Transportation Officials (AASHTO) publication, A Policy on Geometric Design of Highways and Streets (commonly known as the Green Book), does not take into account low-stability, heavy trucks that are susceptible to rollover. To address this issue, the NTSB made several recommendations for consideration by the Federal Highway Administration (FHWA) and AASHTO. In particular, the NTSB suggested a review of AASHTO’s current policy for pavement/shoulder cross-slope breaks on horizontal curves to determine whether updates to the design criteria are needed. AASHTO’s current design policy states that shoulder slopes that drain away from the paved surface on the outside of a superelevated horizontal curve should be designed to avoid too great a cross-slope break, calculated as the algebraic difference between the cross-slopes of the traveled way and the shoulder (AASHTO, 2011). To avoid large pavement/shoulder cross-slope breaks, it may be desirable that all or part of the shoulder be sloped upward at about the same or lesser rate than the superelevated traveled way. Where this is not desirable due to potential adverse conditions (e.g., storm water or melting snow and ice draining onto the paved surface), AASHTO policy indicates that the cross-slope break at the edge of the paved surface should be limited to approximately 8 percent by flattening the shoulder on the outside of the curve. Alternatives for avoiding too severe of a cross-slope break include the use of a continuously rounded shoulder cross-section on the outside of the superelevated traveled way or a planar shoulder with multiple breaks in the cross-slope. The Roadside Design Guide (AASHTO, 2011) indicates the roadside should be rounded, because it reduces the chances of an errant vehicle becoming airborne and affords the driver more control over the vehicle. Figure 1 shows a typical cross-section for superelevated roadway sections with pavement/shoulder cross-slope breaks. Little change has occurred in AASHTO policy on cross-slope breaks over the years. In both the 1957 A Policy on Arterial Highways in Urban Areas (commonly referred to as the Red Book) and the 1965 A Policy on Geometric Design of Rural Highways (commonly referred to as the Blue Book), the American Association of State Highway Officials (AASHO) recommended a maximum cross-slope break of 7 percent (NTSB, 2011). Since at least the 1990 Green Book, AASHTO increased the recommended maximum cross-slope break to 8 percent, and this recommended threshold still exists in AASHTO’s current policy indicating that the algebraic difference in cross-slopes at the edge of the traveled way should not exceed 8 percent to avoid an undesirable rollover effect (AASHTO, 2011).

Figur As report criterion Highway designs b There is passenge dynamics unit com 1970s an substanti e 1. Typica ed by the N appears to b -Vehicle-Ob y testing the a natural con r cars; furth simulation bination) tru d 1980s, and ally from th l Superelev TSB (NTSB e based on a ject-Simula effects of c cern that a ermore, truc models at th cks. Vehicle the state-o e HVOSM u ated Cross- Breaks (A , 2011), the 1981 FHW tion-Model urvature, sp 1971 passen ks were not at time lack design and f-the-art of v sed in the 1 2 - 2 Sections wi ASHTO, 2 current 8-p A study (Gl (HVOSM) t eed, and pat ger car may considered ed the capab the compos ehicle dyna 980s. This c th Pavemen 011) ercent maxim ennon et al. o evaluate v h of a simu not be repre in the previo ility to sim ition of traf mics simula urrent resea t/Shoulder um cross-s , 1981), whi arious cross lated 1971 p sentative of us research ulate articul fic have cha tion modeli rch was con Cross-Slop lope break ch used the -slope break assenger ca current because veh ated (i.e., m nged since t ng has adva ducted base e r. icle ulti- he nced d on

2 - 3 NTSB’s recommendation to conduct a detailed investigation of the cross-slope break design criterion to determine whether the existing policy is appropriate for the current fleet of passenger cars and trucks. 1.2 Research Objective The objectives of this research were to 1) assess AASHTO’s current design policy for pavement/shoulder cross-slope breaks on the outside of horizontal curves to determine whether updates in design criteria are needed, and 2) if updates are needed, to provide recommended changes to AASHTO’s current design policy. 1.3 Research Approach Two separate work plans were executed to determine whether updates to AASHTO’s current design policy for pavement/shoulder cross-slope breaks on the outside of horizontal curves are needed. One work plan involved the use of vehicle dynamics simulation models to determine the effect of cross-slope breaks on the ability of drivers and vehicles to recover from traversing a cross-slope break on a curved section of roadway. Various combinations of key geometric design elements and other critical elements were defined and modeled to assess vehicle stability upon encountering a range of cross-slope breaks between the traveled way and shoulder. Critical geometric design elements and variables considered in the simulation modeling included: vehicle speed, curve radius, superelevation rate, cross-slope break rate, vehicle type, vehicle trajectory, and extent of vehicle encroachment. The primary measures of interest output from the simulations were whether the vehicle recovered from the maneuver or rolled over and the maximum roll angle experienced by the vehicle. The second work plan involved a crash-based safety analysis to determine the extent to which large cross-slope breaks may contribute to crashes near superelevated horizontal curves. Data from the Fatality Analysis Reporting System (FARS) were analyzed to gather general information on where fatal single-vehicle truck rollover crashes typically occur and the frequency of such crashes. Then, a safety analysis was conducted for a small sample of horizontal curves at which the pavement/shoulder cross-slope breaks were measured in the field. Roadway design and crash data were collected for a sample of horizontal curves covering a range of superelevations and cross-slope breaks. The vehicle dynamics simulation modeling provided clear results that directly address the appropriate design criteria for pavement/shoulder cross-slope breaks. The results of the crash- based safety analysis were of limited usefulness because crash databases lack information on the presence or magnitude of pavement/shoulder cross-slope breaks. Based upon the results of the vehicle dynamics simulation modeling, the research team made an assessment of whether updates to AASHTO’s current design policy for cross-slope breaks were needed.

2 - 4 1.4 Outline of Report This report presents an overview of research conducted to assess the need to update AASHTO’s design policy for pavement/shoulder cross-slope breaks on the outside of horizontal curves. The remainder of this report is organized as follows: Section 2—Summary of Literature Review and Current Practice Section 3—Vehicle Dynamics Simulation Modeling Section 4—Crash-Based Safety Analysis Section 5—Conclusions and Future Research Section 6—References

2 - 5 SECTION 2. Summary of Literature Review and Current Practice This section summarizes the results of a literature review and a survey to gather background information on pavement/shoulder cross-slope break issues that helped guide the direction of this research. Previous research that investigated pavement/shoulder cross-slope break issues, documents related to driver behavior during roadway departure events, documents concerning relevant design elements and study variables, and state department of transportation (DOT) design manuals were reviewed. In addition, an online survey was distributed to state design engineers asking questions related to their current cross-slope break policies and practices. The results from this review are summarized according to the following topics:  Cross-slope break research  Driver behavior studies  Critical design elements and variables  Summary of State DOT cross-slope design policies and practices  Summary of key issues 2.1 Cross-Slope Break Research Two previous studies that investigated cross-slope break design and related issues—HVOSM Studies of Cross-Slope Breaks on Horizontal Curves (Glennon et al., 1981) and HVOSM Studies of Highway Cross-Slope Design (Glennon et al., 1983)—were reviewed. In 1981, Glennon et al. studied cross-slope break issues on highway curves, using the McHenry version of the Highway- Vehicle-Object-Simulation-Model (HVOSM). Glennon et al. compared cross-slope break to design curvature, vehicle path, superelevation, and speed. AASHTO policy controlled the range of these variables and their relationship to one another. For example, for a given curvature, only the design speeds permitted by AASHTO were studied. Cross-slope break was the only variable that was studied outside of the established AASHTO guidelines. Glennon et al. used only a passenger car (1971 Dodge Coronet) as the design vehicle and did not include trucks due to HVOSM limitations and the scope of the project. Glennon et al. considered a circular design path and varied the radius of curvature. Of the many types of lane departures possible, they considered only a moderate departure, in which the vehicle was steered back to the driving lane, assuming the shoulder width was not a limiting factor. The path of vehicle departure is represented in Equation (1). ܴ௩ ൌ ଵଽ,଼ଶହ ൈ ோோାଶଷ,଴ଽ଺ (1) where Rv = radius of the vehicle path (ft) R = radius of the horizontal curve (ft) This equation represents the 95th percentile transient path (i.e., short duration trajectory when the driver changes the steering input on a vehicle) and was established based on highway operational studies conducted by Glennon and Weaver (1972).

2 - 6 Preliminary runs were conducted to establish the controlling details of the lane departure maneuver. These runs indicated that a 4-wheel departure produced more extreme responses than a 2-wheel departure and that the entry to the cross-slope break was more extreme than the exit. Comparing the performance criteria for lateral friction demand and the performance criteria for driver discomfort, Glennon et al. concluded that driver discomfort was the limiting criteria. Based on the safety-conservative design philosophy modeled by AASHTO and a study conducted by Rice and Dell’Amico (1974), it was rationalized that a maximum discomfort level of 0.3 g was appropriate. After interpolation and smoothing of the data, a single recommendation for wide shoulders was established as a maximum of 8 percent cross-slope break. From the HVOSM results, Glennon et al. determined it was most beneficial to observe the vehicle recover from an all-wheel traversal of the cross-slope break than from a partial lane departure across the cross-slope break. In a subsequent study, Glennon et al. (1983) investigated the effects of cross-slope and centerline cross-slope break on lateral tire acceleration, vehicle roll angle, and driver comfort using HVOSM and the Highway Safety Research Institute/Motor Vehicle Manufacturers Association Phase 4 (HSRI/MVMA) simulation model. In general, the study highlights the need to provide a cross-slope design that is at least the minimum value required to meet drainage requirements. 2.2 Driver Behavior Studies Three references that provided guidance on driver behavior during roadway departure maneuvers were reviewed. 2.2.1 SAE Avoidance Study In a study conducted by the Society of Automotive Engineers (SAE), Maeda et al. (1977) observed a sedan traveling at 37 mph faced with an emergency that is 1.3 seconds to collision. Because of the severity of the emergency posed, the driver was forced to perform an emergency avoidance maneuver without braking. In the test, the drivers were told to avoid the emergency by simulating a lane change through a lateral displacement of 12 ft. Steering wheel angle and vehicle position over time were recorded. The most common maximum steering angle of all the drivers was between 210 and 230 degrees. Figure 2 shows the distribution of observed maximum steering angles. This study provided no guidance regarding the behavior of a driver returning to the lane of travel. Maeda et al. also noted that braking was not applied during emergency avoidance situations. 2.2.2 IEEE Collision Avoidance Kim et al. (2005) conducted a study using a Computer Assisted Virtual Environment (CAVE) driving simulator to record driver response to an emergency that is 1.3 seconds to collision. The testing scenario was a sedan traveling at 31 mph behind a truck that suddenly stopped. The

maneuve Similar to Figu the emerg three driv of travel. simulatio 2.2.3 Te The Texa recomme (Texas D  D t  D y  B y y a Other sta handbook r was simula the Maeda re 2. Obser ency and th ers were ex This inform ns conducte xas Comm s Commerc ndations for epartment o o not apply urning. If th o not turn a ou turn, the e prepared ou have pas ou will not nd counters tes have sim s. ted on a str et al. (1977 ved Maxim e necessity amined whi ation was u d as part of ercial Mo ial Motor V truck drive f Public Saf the brake w at happens, ny more tha greater the to counterst sed whateve be able to do teering as tw ilar steering aight road in ) study, brak um Steering for avoidanc ch included sed in defin this research tor Vehicl ehicle Drive rs under the ety, 2010) a hile you are you may ski n needed to chances to s eer, that is, t r was in yo it quickly o parts of o and brakin 2 - 7 an urban ar ing was det Angles fro e rather tha data for the ing vehicle t (see Sectio e Driver H r Handbook subheading s follows: turning. It d out of con clear whate kid or rollov o turn the w ur path. Unl enough. You ne driving a g recommen ea with a fr ermined to b m SAE Stu n stopping. return of the rajectories f n 3). andbook (CMVD) o “How to Tu is very easy trol. ver is in you er. heel back in ess you are should thin ction. dations in th iction coeffi e zero due dy (Maeda The driving vehicle to or the vehic utlines emer rn Quickly to lock you r way. The the other d prepared to k of emerge eir respecti cient of 0.8 to the severi et al., 1977 behaviors o the original le dynamics gency steer and Safety” r wheels wh more sharpl irection, onc countersteer ncy steerin ve CMVD . ty of ) f lane ing ile y e , g

2 - 8 2.3 Critical Design Elements and Variables Several documents that provided insight into key design elements and variables were reviewed. 2.3.1 Review of Truck Characteristics as Factors in Roadway Design In NCHRP Report 505, Review of Truck Characteristics as Factors in Roadway Design, Harwood et al. (2003) summarized results of research conducted to assess whether geometric design criteria for highways and streets can reasonably accommodate the dimensions and performance characteristics of the current and future truck fleet using the U.S. highway system. Relevant findings and/or key points from the research include:  The current Green Book criteria for cross-slope breaks and vertical clearances appear to be appropriate for the current truck fleet.  The minimum rollover threshold for trucks is generally in the range from 0.35 to 0.40 g. This minimum rollover threshold generally applies to trucks fully loaded with uniform density cargo.  When a vehicle moves through a curve at higher speed, the rear axles of the vehicle tend to move outward. This tendency to move outward is called high-speed offtracking. It acts in opposite direction to low-speed offtracking, so that the two phenomena tend to counteract each other. At lower speeds, low-speed offtracking predominates; as the speed increases, the net offtracking is reduced. At sufficiently high speeds, the two phenomena cancel each other out, resulting in no net offtracking. High-speed offtracking is usually not a significant factor in roadway design, compared with low-speed offtracking. In another study, Garcia et al. (2003) analyzed the dynamic response of a tractor/single-van- trailer truck carrying loads of varying weights under actual operating conditions as the vehicle traveled along roadway curves of varying radii. Data were collected over a total of 690 mi of highway under three load configurations: empty, loaded with less-than-truck-load (i.e., partially- loaded), and fully loaded with bottled water packed in boxes. For the empty unit, a rollover threshold in the order of 0.54 was estimated. For the partially-loaded unit, the rollover threshold was on the order of 0.34; and for the fully-loaded unit, the rollover threshold was on the order of 0.45. Thus, among their conclusions Garcia et al. concluded that, of the three load configurations tested, the partially-loaded vehicle displayed the highest propensity to rollover. 2.3.2 Superelevation Criteria for Sharp Horizontal Curves on Steep Grades During research to develop superelevation criteria for sharp horizontal curves on steep grades, Torbic et al. (2013) conducted a series of field studies and vehicle dynamics simulations to investigate the combination of horizontal curve and vertical grade design criteria. Vehicle types considered in the research included passenger cars and trucks. In addition, pavement-tire friction data were collected at eight locations, representative of pavement surface conditions on multi- lane, divided highways. Relevant findings from this research include:

2 - 9  For passenger cars, mean maximum wet-tire friction values ranged from approximately 0.91 to 0.82, and mean skidding wet-tire friction values ranged from approximately 0.67 to 0.58 in the longitudinal (braking) direction.  For trucks, mean maximum wet-tire friction values ranged from approximately 0.82 to 0.78, and mean skidding wet-tire friction values ranged from approximately 0.59 to 0.54 in the longitudinal (braking) direction.  The margins of safety against skidding were slightly higher for the tractor/double-van- trailer truck (i.e., double trailer trucks) compared to the tractor/single-van-trailer truck.  Assuming the center of gravity (CG) height and track width of the trailers for both vehicles are the same, tractor/double-van-trailer trucks (i.e., double trailer trucks) have very similar rollover margins of safety as compared to tractor/single-van-trailer trucks. 2.4 Summary of State DOT Cross-Slope Break Design Policy and Practice The research team conducted a review of state department of transportation (DOT) design manuals to assess the state of practice in the 50 states related to pavement/shoulder cross-slope break design. The review focused on identifying policies and practices that differ from what is specified in the 2011 Green Book, which provides the following guidance on cross-slope break design (AASHTO, 2011): On tangent or long-radius curved alignment with normal crown and turf shoulders, the maximum algebraic difference in the traveled way and shoulder grades should be from 6 to 7 percent. Although this maximum algebraic difference in slopes is not desirable, it is tolerable due to the benefits gained in pavement stability by avoiding stormwater detention at the pavement edge. Shoulder slopes that drain away from the paved surface on the outside of well-superelevated sections should be designed to avoid too great a cross-slope break. For example, use of a 4 percent shoulder cross-slope in a section with a traveled way superelevation of 8 percent results in a 12 percent algebraic difference in the traveled way and shoulder grades at the high edge of the traveled way. Grade breaks of this order are not desirable and should not be used (Figure 4-2A). It is desirable that all or part of the shoulder should be sloped upward at about the same rate or at a lesser rate than the superelevated traveled way (see the dashed line labeled Alternate in Figure 4-2A). Where this is not desirable because of stormwater or melting snow and ice draining over the paved surface, a compromise might be used in which the grade break at the edge of the paved surface is limited to approximately 8 percent by flattening the shoulder on the outside of the curve (Figure 4-2B). One means of avoiding too severe of a grade break is the use of a continuously rounded shoulder cross-section on the outside of the superelevated traveled way (Figure 4-2C). The shoulder in this case is a convex section continuing from the superelevation slope instead of a sharp grade break at the intersection of the shoulder and traveled way slopes. In this method, some surface water will drain upon the traveled way; however, this disadvantage is offset by the benefit of a smoother transition for vehicles that may accidentally or purposely drive upon the shoulder. It should also be noted that convex shoulders present more

2 - 10 difficulties in construction than do planar sections. An alternate method to the convex shoulder consists of a planar shoulder section with multiple breaks in the cross slope. Shoulder cross-slopes on the high side of a superelevated section that are substantially less than those discussed above are generally not detrimental to shoulder stability. There is no discharge of stormwater from the traveled way to the shoulder and, therefore, little likelihood of shoulder erosion damage. Highlights from selected state policies as they relate to pavement/shoulder cross-slope break design are shown below. Where information for a state is not provided, it is either because state policy is consistent with the AASHTO Green Book or no design manual or guidelines were obtained for review for the particular state.  Colorado DOT’s Roadway Design Guide (2011) chapter on Cross Section Elements indicates that shoulder cross-slope break may be 5 percent. For greater cross-slope break, rounding the outside of the travel lane is required.  Florida DOT’s Green Book (2011) recommends a 5 to 8 percent maximum cross-slope break at the crossover line for curves with design speeds under 30 mph and 4 to 5 percent for curves with design speeds of 35 mph and above.  Kentucky Transportation Cabinet’s (KYTC) Geometric Design Guidelines recommends that for the cross-slope break between superable and nonsuperable shoulder, the cross- slope break should be less than 12 percent.  Louisiana Department of Transportation and Development’s (DOTD) Road Design Manual (2009) recommends that the cross-slope break between main lanes be limited to 5 percent and the cross-slope break between the main roadway and shoulder should not exceed 7 percent.  Mississippi DOT’s Roadway Design Manual (2001) recommends cross-slope breaks of 5 to 8 percent for turning roads with design speeds of 20 mph or less, 5 to 6 percent for turning roads with design speeds between 25 and 30 mph, and 4 to 5 percent for turning roads with design speeds 35 mph or more.  Nevada DOT’s Roadway Design Manual (2010) recommends that cross-slope should be checked for maximum allowable cross-slope break for entrance or exit ramps. For example, when the freeway mainline cross-slope is 2 percent and the merging on-ramp is an 8 percent superelevation, this is an algebraic difference of 6 percent, which exceeds the maximum allowable cross-slope break of 5 percent.  New Jersey DOT’s Roadway Design Manual (2011) recommends that too great of a cross-slope break should be avoided at the high side of superelevated sections. The manual states that the shoulder cross-slope should be 4 percent where the superelevation rate is 3 percent or less; for superelevation rates greater than 3 percent and less than 5 percent, a maximum cross-slope break of 7 percent will be used to establish the shoulder cross slope; and when superelevation rates range from 5 percent to 6 percent, the shoulder cross-slope will be 2 percent. In addition to reviewing state DOT design manuals, the research team conducted a brief online survey asking state design engineers about their current cross-slope break policies/practices. Responses to the cross-slope break survey are summarized below. The survey was administered

2 - 11 electronically to all 50 U.S. state highway agencies; highway agencies from 27 states responded to the survey. Responses to categorical questions are summarized by showing both the percentage of the responses and the frequency/number of responses shown in parentheses. For those questions that asked agencies to further explain an issue, a summary of individual responses is provided. Not all respondents answered all of the questions. 1. Does your agency have a written policy for design of pavement/shoulder cross-slope break? YES: 92.6% (25) NO: 7.4% (2) If you answered YES to Question 1, what is your agency’s policy on design of pavement/shoulder cross-slope breaks: The same policy in the AASHTO Green Book: 50.0% (13) Different from the policy in the AASHTO Green Book (please explain): 50.0% (13)  Responses ranged from a maximum cross-slope break of 4 percent to 8 percent. Others indicated they maintain the superelevation rate for the full width of the high side shoulder. 2. For a shoulder on the outside of a superelevated horizontal curve, does your agency: Normally provide a shoulder that slopes away from the pavement: 44.4% (12) Normally provide a shoulder that slopes toward the pavement: 55.6% (15) 3. If your agency normally uses a shoulder that slopes toward the pavement on the outside of a superelevated horizontal curve, is the shoulder cross-slope typically: The same as the superelevated pavement cross-slope: 81.8% (18) Different from the superelevated pavement cross-slope (if so, please explain your typical practice): 18.2% (4)  Responses included: (a) If pavement superelevation rate is 4 percent or less, the shoulder slope is 4 percent downward (normal tangent cross-slope) away from the traveled way. If the pavement superelevation rate is greater than 4 percent, but less than or equal to 6 percent, the shoulder slope is 2 percent downward away from the traveled way. If the pavement superelevation rate is greater than 6 percent, the shoulder slope is 1 percent toward the traveled way; (b) For superelevation rates between 3 and 5 percent, the pavement/shoulder cross-slope break must not exceed 7 percent. For superelevation rates between 5 and 7 percent, the shoulder cross slope must be 2 percent away; (c) 6 percent maximum cross-slope break. For -1, 0, 1, and 2 percent pavement cross-slopes, 4 percent sloped paved shoulders are used for 10 ft wide and above paved shoulders; and (d) We only allow shoulder widths less than 4 ft to slope toward the pavement at the superelevation rate.

2 - 12 4. What is your agency’s typical shoulder cross-slope on tangent sections on rural highways?  2% (8 agencies; 31% of respondents)  4% (8 agencies; 31% of respondents)  2-4% (3 agencies; 12% of respondents)  4-6% (2 agencies; 8% of respondents)  2-6% (1 agency; 4% of respondents)  3-8% (1 agency; 4% of respondents)  4.2% (1 agency; 4% of respondents)  4-8% (1 agency; 4% of respondents)  5% (1 agency; 4% of respondents) 5. What is your agency’s typical shoulder cross-slope on the outside of superelevated horizontal curves, if it is not based on the superelevation rate of the curve?  Maximum cross-slope break is used to calculate cross-slope of shoulder (4 agencies;25% of respondents)  Based on superelevation of the curve (4 agencies; 25% of respondents)  4% (3 agencies; 19% of respondents)  Algebraic difference not to exceed 8% (2 agencies; 13% of respondents)  Algebraic difference not to exceed 7% (2 agencies; 13% of respondents)  2% (1 agency; 6% of respondents) 6. Does your agency’s policy on maximum pavement/shoulder cross-slope break differ between facility types (e.g., rural highways, urban highways, freeways, ramps)? YES (please explain): 18.5 % (5)  Responses included (a) maximum cross-slope break of urban highways is generally limited to 4 percent; (b) specific methods are used for interchange ramps; (c) different rates are used for rural vs. urban and for divided vs. undivided; and (d) different rates are used based on shoulder type. NO: 81.5% (22) 7. Does your agency’s policy on maximum pavement/shoulder cross-slope break differ due to posted speed changes? YES: 7.7% (2) NO: 92.3% (24) 8. Does your agency’s policy on shoulder width or maximum pavement/shoulder cross- slope break vary as a function of truck volume or truck percentage? YES: 18.5% (5) NO: 81.5% (22)

2 - 13 9. Are you aware of any crashes on highways/roadways within your state that might be related to maximum pavement/shoulder cross-slope break issues? YES: 7.4% (2) NO: 92.6% (25) 10. What records does your agency have available that might help in identifying field sites with various magnitudes of pavement/shoulder cross-slope breaks? As-built plans: 77.4% (22) Roadway alignment databases including horizontal curve superelevation and shoulder cross-slope: 6.5% (2) Lidar scans: 0% (0) Other (please explain): 16.1% (6)  Responses included: (a) construction and as-built plans, (b) field examination, and (c) design exceptions 11. Has your agency encountered any problems concerning pavement/shoulder cross- slope breaks? YES (please explain): 7.7% (2)  Responses included: (a) one instance of improper construction, (b) physical constraints on an exit ramp created a problem in designing adequate superelevated section, and (c) melting snow can cause problems when the shoulder is sloped toward the travel lanes. NO: 92.3% (24) 12. Are you aware of any sites/locations within your state where pavement/shoulder cross- slope break exceeds 8 percent, and that may be suitable for field study (e.g., possibly one or two major sites that come to mind or based on a policy that permits pavement/shoulder cross-slope breaks greater than 8 percent on certain facility types or ramps)? YES: 0% (0) NO: 100% (27) 13. Could you explain your state design experience in regards to drainage on the high edge of a superelevated roadway (i.e. design a v-ditch to capture sloped runoff)?  Common responses included: (a) we maintain a standard ditch throughout, and (b) a V-ditch is used to capture sloped runoff. 14. From a construction standpoint, is it practical for a state agency to implement rounded shoulders as described in the AASHTO Green Book? YES (please explain): 29.6% (8)  Responses included: (a) we have constructed rounded shoulders in various pavement types, including concrete, asphalt, aggregate, and turf; (b) yes for gravel shoulders but we found this degrades over time; (c) this is simple to do with asphalt paving, which we commonly use; (d) we routinely round shoulder breaks; and (e) we round shoulders when cross-slope break exceeds 8 percent.

2 - 14 NO (please explain): 70.4% (19)  Responses included: (a) construction is too difficult and complex; (b) constructing rounded shoulders is too costly; (c) construction is too complex, and over time the section becomes rounded anyway due to weathering; (d) our construction administration finds it difficult; (e) difficult to construct and difficult to inspect whether construction was done as designed; (f) the amount of rounding that could be done would be too small to quantify; (g) new pavers and screeds would be required; and (h) too difficult to place and achieve adequate compaction. 15. Is there guidance on pavement/shoulder cross-slope break design not currently included in your state policy or the AASHTO Green Book that would be useful? YES (please explain): 13.7% (3)  Responses included: (a) situational guidance/criteria where lesser break is more appropriate; (b) we allow up to 4 percent grade break in the gore areas for interchanges and up to 8 percent break in the flare area of guardrail; and (c) a summary of crash experience with slope breaks versus plane cross-slopes. NO: 86.3% (24) 2.5 Summary of Key Issues Based on the review of literature and current policies and practices, several key issues relevant to this research are as follows:  AASHTO’s current policy states that to avoid an undesirable rollover effect, the algebraic difference in cross-slopes at the edge of the traveled way should not exceed approximately 8 percent. This threshold value of limiting the cross-slope break to 8 percent is based upon previous HVOSM simulations for a passenger car.  As expected, most agencies’ policies/practices are consistent with AASHTO policy. However, some agencies have stricter criteria; for example, they limit the cross-slope break to a maximum algebraic difference of 6 to 7 percent, and a few states limit the cross-slope break to 4 or 5 percent. Also, in some states, the limiting threshold is a function of speed.  Several resources were reviewed that provided information to help guide vehicle dynamic simulation activities. For example, information was found to help define vehicle trajectories and driver response upon roadway departure. In addition, previous research suggests that the rollover characteristics of tractor/double-van-trailer trucks are similar to those of tractor/single-van-trailer trucks.  Different load configurations should be tested during simulation, as load configuration impacts a vehicle’s rollover propensity.

SECT Vehic This sect effect of slope bre combinat assess the the travel dynamics followed 3.1 Ov Critical g included  V  V  V Figure 3 available buses, an details on involved for analy scenarios ION 3. le Dyna ion presents cross-slope ak when nav ions of key ir impact on ed way and simulation by the simu erview o eometric de : ehicle type ehicle spee ehicle traje illustrates th software to d articulated the method selecting th sis and cond to be simul mics Sim the vehicle breaks on a igating a cu geometric d vehicle sta shoulder. T modeling to lation result f Vehicle sign elemen d ctory e general ap ol for simula vehicles, w ology used e vehicle typ ucting preli ated. Figu ulation dynamics si vehicle’s ab rved roadw esign eleme bility when his section f investigate s, and a sum Dynami ts and varia proach to th ting and an as used in t to define the es, design s minary simu re 3. Gene 2 - 15 Model mulation mo ility to recov ay. Through nts and othe encounterin irst presents the effect o mary of key cs Simul bles conside  Cro  Sup  Ext e simulation alyzing the v his research parameters peeds, roadw lations to re ral Simulat ing deling cond er from enc vehicle dyn r critical ele g a range of the general f cross-slop findings. ation Mo red in the si ss-slope bre erelevation ent of vehic process. T ehicle dyna . The follow for the sim ay geomet fine the me ion Process ucted to de roachment amics simu ments were cross-slope approach to e break on v deling A mulation mo ak rate rate le encroachm ruckSim, a c mic behavio ing sections ulation mod ry, and vehi thodologica termine the over the cro lation mode evaluated to breaks betw the vehicle ehicle stabil pproach deling ent ommerciall r of trucks, provide fur eling. This cle trajector l approach a ss- ling, een ity, y ther ies nd

2 - 16 3.1.1 Vehicle Types and Models Three vehicles were selected for use in the simulations: a tractor/single-van-trailer truck, a tractor/tanker-trailer truck, and a tractor/double-van-trailer truck. The basic input parameters for the vehicles were selected based on vehicle dimensions from the AASHTO Green Book, and mass and center of gravity values from the AASHTO Manual for Assessing Safety Hardware (MASH) (AASHTO, 2009). Default TruckSim models were modified with the dimensions and weights of each of the vehicle types. For most simulations, the target weight of the vehicle was approximately 80,000 lb, the current federal maximum standard for commercial vehicle operations on the Interstate Highway System. For a few test scenarios, partially-loaded vehicles were simulated that were less than 80,000 lb. Consideration was given to modeling the performance of a sport utility vehicle (SUV) using CarSim, a commercially available software tool for simulating and analyzing the vehicle dynamic behavior of passenger cars, light trucks, motorcycles, and specialty vehicles. However, since SUVs have a lower propensity for rolling over than the three truck types included in the simulation study, it was determined that modeling the vehicle dynamics of an SUV for the same testing scenarios as the three trucks would provide no additional information relevant to cross- slope break policies. Table 1 shows the primary vehicle input parameters selected for use in TruckSim. Dimensions for the tractor/single-van-trailer truck and tractor/tanker-trailer truck were based on the WB-19 design vehicle (AASHTO, 2011), and the dimensions for the tractor/double-van-trailer truck were based on the WB-20D design vehicle. The heights for the centers of gravity of the trailer ballasts were determined from Table 4-2 in MASH (AASHTO, 2009). Table 1. Primary Vehicle Input Parameters Used in TruckSim Vehicle inputs Tractor/single-van-trailer truck Tractor/tanker-trailer truck (full tanker) Tractor/double- van-trailer truck Tractor type Sleeper Cab Sleeper Cab Day Cab Total mass of vehicle1 (lb) 80,158 80,161 80,151 Total mass of tractor (lb) 18,629 18,629 12,698 Total mass of trailer(s) (lb) 15,999 15,999 17,262 (two trailers + dolly) Trailer ballast (laden) (lb) 45,358 45,358 48,610 (both) Height of center of gravity of trailer ballast (in) 73 81 73 Offset of center of gravity from middle of trailer (in) 0 0 0 Total length of vehicle (ft) 69 69 72 Total length of trailer to back axle (ft) 44.2 44.2 24.1 Distance from front axle to center of tandem of tractor (ft) 18.5 18.5 11.5 1 Weight of vehicle model verified by summing the tire forces underneath the vehicle driving along a flat road. For the tractor/single-van-trailer truck and tractor/double-van-trailer truck, one loading configuration was simulated for each truck type. For the tractor/tanker-trailer truck, existing vehicle dynamics simulation models do not have the capability to simulate the dynamic effects of liquid sloshing in a tanker trailer. Therefore, only the fully-loaded trailer condition was selected

2 - 17 for evaluation of pavement/shoulder cross-slope breaks as it was assumed that the dynamic effects of liquid sloshing should be minimal for the fully-loaded trailer condition. (Note: The dynamic effects of liquid sloshing in a tractor/tanker-trailer truck could not be simulated in this research. Therefore, the tractor/tanker-trailer truck was simulated by modifying the center of gravity of the tractor/single-van-trailer truck to represent the performance of a tractor/tanker- trailer truck within the simulation model). Appendix A provides results for additional combinations of tractor/tanker-trailer trucks represented and simulated in this research to learn more about the possible differences in stability between a fully-loaded tanker trailer and a partially-loaded tanker trailer. However, as the dynamic effects of sloshing liquid within a tractor/tanker-trailer truck could not be simulated, the validity and accuracy of the simulation results for the partially-loaded tractor/tanker-trailer trucks could not be verified. Therefore, results from simulation scenarios for the partially-loaded tractor/tanker-trailer trucks were not considered when making decisions regarding design criteria for pavement/shoulder cross-slope breaks and are only provided in Appendix A for presentation purposes. 3.1.2 Select Design Speeds and Roadway Geometries for Simulation The roadway geometries defined for simulation purposes were selected based upon combinations of vehicle design speeds, superelevation rates, and cross-slope break rates. Table 2 shows the primary design speeds and superelevation rates selected for consideration in the analysis, including design speeds of 30, 50, 60, and 70 mph and superelevation rates of 4, 6, and 8 percent. Based upon the design speed and superelevation, Table 3-7 in the AASHTO Green Book specifies the minimum design radius for a horizontal curve. Table 2 summarizes the minimum design radii for the design speeds and superelevation rates selected for simulation. All simulation scenarios were based upon minimum radius curves and vehicles traveling at the design speed of the curve. A 12-ft travel lane was assumed for all simulations, along with a continuous-width paved shoulder. Table 2. Specified Minimum Design Radius Design Speed Minimum Radius (ft) for a Superelevation of: 4% 6% 8% 30 mph 250 231 214 50 mph 926 833 758 60 mph 1,500 1,330 1,200 70 mph NA1 2,040 1,810 1 A 70-mph design speed with a 4-percent superelevation is not an applicable design scenario in the AASHTO Green Book A variety of pavement/shoulder cross-slope breaks (i.e., algebraic differences in cross slope) were investigated for each of the design curves. Five cross-slope break rates were considered in the assessment: 0, 4, 6, 8, and 10 percent. For the simulations, a dry surface coefficient of friction of 0.8 for tire rolling resistance was assumed. This value was chosen based on recent studies (Kim et al., 2005; Torbic et al., 2013).

All simul curve and vehicle th 3.1.3 Ve To simul vehicle tr  F  P  F For partia only the p shoulder traversal vehicle) t The full t attempts curve and departure departs th corrects t The defin based up ment for ated geome encroached at first trav hicle Dep ate driver be ajectory pat ull traversa artial traver ull traversa l traversal d assenger-si . For full trav moderate de raverse the raversal sev to steer a pa then steer b maneuver w e travel lan he steering ed steering on data gath a sedan enc Figure 4. C tries were cu onto the ou ersed the cro arture Tra havior when hs were def l severe dep sal moderat l moderate d eparture sce de tires of th ersal depar parture scen cross-slope ere departur th to avoid a ack into the as only tes es until all t to return bac input for thi ered by Kim ountering an ollision Av rves to the tside should ss-slope bre jectories encounteri ined for veh arture (for m e departure eparture narios (i.e., e vehicle tr ture scenario arios), all ti break and en e trajectory n obstacle i travel lane ted with a fu ires traverse k to the trav s corrective et al. (2005 emergency oidance Ma 2 - 18 left. In other er, it was a ak. ng the pave icles to follo odel testing partial trave averse the c s (i.e., full t res of the ve croach onto was defined n the travel upon encro ll traversal the cross-sl el lane. collision av ). Kim et al at a speed o neuver (ad words, as v lways the pa ment/should w along the only) rsal modera ross-slope b raversal sev hicle (i.e., t the shoulde based upon lane while t achment ont departure sc ope break, a oidance ma . observed t f 31 mph. apted from ehicles trac ssenger-sid er cross-slo simulated g te departure reak and enc ere departu ires on both r. the situatio raversing th o the should enario in wh t which poin neuver (see ime versus l Kim et al. ked around e tires of the pe break, th eometries: scenarios), roach onto re and full sides of the n where a d e horizontal er. The sev ich the veh t the driver Figure 4) w ateral displa [2005]) the ree the river ere icle as ce-

To apply were adju However complete used to e avoidanc speed of path func strating t et al. Figure Although driver at displacem of the veh fully cros displacem centerlin Figure 6 the vehic “10.40ftL that were collision braking w The prim that the v during se combinat the data to a sted to refle , Kim et al. the return t xtrapolate th e paths for a 50 mph was tion. The da hat the gene 5. Vehicle these gener different spe ent from th icle to inter s onto the s ent of 10.4 e of the vehi presents the le dynamics atDisp_60m modeled in avoidance, t as not cons ary purpose ehicle dyna veral study ions of simu vehicle mo ct the targe only recorde o the travel e path funct vehicle trav then checke ta proved to rated path fu Paths Gene ated paths r eds, they do e centerline act with the houlder, the 0 ft from the cle onto the scaled path simulation ph”, and “1 to TruckSim he SAE stud idered in the for simulati mics simula scenarios an lation scena ving at 50, t speed. Hen d the prelim lane, the slo ion to full r eling at 50, d against th be compati nction corre rated by Pa (Sever epresent the not constit of the trave pavement/s paths were centerline shoulder w s for a full tr model. The 0.40ftLatDi for the sim y (Maeda e constructio ng the full t tion model w d recovered rios to be te 2 - 19 60, and 70 m ce, a speed inary return pe from the ecovery. Fig 60, and 70 e original d ble and show lated well w th Functio e Departur collision av ute a full lan l lane was 9 houlder cro uniformly sc of the travel as 4.40 ft. aversal seve paths labele sp_70mph” ulations. Co t al., 1977), n of these p raversal sev as function during othe sted in this ph for use and time pat behavior o last two dat ure 5 presen mph. The ge ata gathered ed favorab ith the veh n for Collis e) oidance ma e departure .18 ft, which ss-slope bre aled to acco lane. There re departure d “10.40ftL represent th nsistent wit and CMVD aths. ere departur ing properly rs) and to as research. Th in this resea h function w f the driver/ a points was ts the gener nerated path by Kim et a le resembla icular path o ion Avoidan neuver exec . The maxim is not adeq ak. To allow mmodate a fore, the dis that was u atDisp_50m e vehicle de h the study b manual (Tx e trajectory (i.e., vehic sess the pri is vehicle tr rch, the data as develop vehicle. To averaged an ated collisio for a vehic l. to validat nce, demon- bserved by ce Maneuv uted by the um lateral uate for all the vehicle total lateral placement o sed for testin ph”, parture path y Kim et al DPS, 2010) was to verif les rolled ov mary ajectory ed. d n le e the Kim er tires to f the g s . on , y er

represent results fr making d maneuve Fig The path Glennon where: R R While tra travel alo moderate constant of the veh trajectori the pavem This met Partial an type, spe simulated ed the most om the full t ecisions reg r was consid ure 6. Vehic of moderate et al. (1981) v = radius = radius versing the ng a path ta vehicle dep with respect icle. 8 disp es illustrated ent/should hod was use d full mode ed, encroach for a small extreme ma raversal sev arding desig ered too ext le Path wit Maneuv vehicle dep in their stu of the vehic of the horizo horizontal c ngent to the arture. The to the partia lays this con in 8 repres er cross-slop d to define t rate departu ment, and g er subset of neuver that ere departur n criteria fo reme to serv h 10.40 ft L er (Full Tr arture was d dy of cross- ܴ௩ le path (ft) ntal curve ( urve, the ve roadway cu lateral displ l and full d cept as appl ent the later e break bou he geometri re scenarios eometry. Fu scenarios fo 2 - 20 was simulat e simulation r pavement/ e as a basis ateral Disp aversal Sev efined base slope break ൌ ଵଽ,଼ଶହ ൈ ோோାଶଷ,଴ଽ଺ ft) hicle/driver rvature. Fig acement (i.e eparture par ied to a hor al displacem ndary (i.e., es of the veh were simula ll traversal r model test ed as part of s were not d shoulder cro for determi lacement fo ere Depart d on the veh in the early begins leavi ure 7 illustra ., delta) sho ameters esta izontal curv ent of the c 0.00 ft on th icle paths f ted for man severe depar ing and for this researc irectly cons ss-slope br ning design r Collision ure) icle trajecto 1980’s. ng the midd tes the defi wn in Figur blished rela e with radiu enterline of e vertical ax or input into y combinat ture scenari defining tes h; however idered when eaks as this policy. Avoidance ry used by le of the lan ned path for e 7 is consid tive to the w s of 250 ft. T the vehicle is of the fig TruckSim. ions of vehi os were ting scenario , the (2) e of a ered idth he from ure). cle s.

Fig Fig In summ slope bre  P t ure 7. Mode ure 8. Driv ary, two veh aks: artial traver ravel lane al rate Vehic er Path/Tr icle trajecto sal moderat ong a path t le Departur ajectory Mo ries were se e departure: angent to th 2 - 21 e Model (ad derate Dep lected for ev The vehicle e roadway c apted from arture on R aluation of gradually d urvature and Glennon e oad Radiu pavement/sh rifts from th only the pa t al., 1981) s of 250 ft oulder cros e middle of ssenger-sid s- the e

2 - 22 tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. This vehicle trajectory represented the mildest departure scenario simulated as part of this research.  Full traversal moderate departure: The vehicle gradually drifts from the middle of the travel lane along a path tangent to the roadway curvature and all tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. This vehicle trajectory was somewhat more severe than the partial traversal moderate departure trajectory as Glennon et al. (1981) reported that when all tires (i.e., tires on both sides of the vehicle) traverse the cross-slope break it produced more extreme responses than when only the passenger-side tires traverse the cross-slope break. This vehicle trajectory served as the basis for AASHTO’s current design policy on pavement/shoulder cross-slope break. A third vehicle trajectory was used for testing the vehicle dynamics simulation model to verify that realistic results were being obtained and to assess the primary combinations of simulation scenarios to be tested in this research:  Full traversal severe departure: The steering inputs represent the situation where the driver steers to avoid an obstacle in the travel lane (i.e., a collision avoidance maneuver) while traversing a horizontal curve and all tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. This vehicle trajectory represents an emergency collision avoidance maneuver so extreme that trucks are likely to rollover in some cases even if no pavement/shoulder cross-slope break is present. This vehicle trajectory represented the most extreme maneuver simulated as part of this research. The results from the full traversal severe departure simulations were not directly considered when making decisions regarding design criteria for pavement/shoulder cross-slope breaks as this maneuver was considered too extreme to serve as a basis for determining design policy. 3.1.4 Preliminary Simulations Preliminary simulations of a tractor/single-van-trailer truck encroaching onto the right shoulder and encountering various cross-slope breaks were conducted to assess the primary combinations of simulation scenarios to be tested in this research. First a tractor/single-van-trailer truck traversing a tangent section of roadway with zero percent superelevation and zero percent cross- slope break was simulated. In this scenario, because the superelevation of this tangent section of roadway was zero, the initial roll angle of the vehicle was also zero. The initial roll angle of a vehicle is correlated to the cross-slope of the roadway. The initial (steady-state) roll angle of a vehicle is not zero for inclined or superelevated roadways (superelevation > 0). The initial roll angle increases with an increase in superelevation, but the roll angle does not equal the superele- vation because the vehicle suspension compensates partially for the cross-slope in the road surface. A positive roll angle is a roll toward the passenger side of the vehicle, and a negative roll angle is a roll toward the driver side of the vehicle (see Figure 9).

Table 3 s tractor/si avoidanc vehicle s from the angles ex tractor un screen ca maximum Table In the sim Exceptio positive r 2.47 degr moderate hows the pr ngle-van-tra e model) tra peed input in maneuver o perienced b it, and Figu ptures show roll angle 3. Summa ulated scen ns were for oll angle fo ees. The tra departure. Figure 9. R imary measu iler truck te jectory. The to TruckSim r rolled over y the vehicl re 11 shows n in Figure of either uni ry of Truck T Superelevatio Cross-Slope B Speed (mph) Recover (Y/N) Max Neg Roll Max Roll Ang arios, the m cases with m r the tractor ctor unit con oll Angle S res of inter sted for the f table specif . The table (N), and it e. Figure 10 a graph of t 12 were tak t. Sim Result angent with n reak (CSB) Angle (- deg) le (+ deg) aximum roll aneuver ev unit in the f trolled only 2 - 23 ign Conven est output fr ull traversa ies the supe also indicat displays the shows a gra he roll angl en at the tim s for a Trac 0% Super 50 Y -5.374 6.846 angle almo ents that had ew cases in a small por tion in Tru om TruckSi l severe dep relevation, c es whether maximum n ph of the ro es experienc e when the tor/Single- elevation 0% 0% 60 Y -6.684 7.48 st always oc insignifica which the tr tion of case ckSim m for the sim arture (i.e., c ross-slope b the vehicle r egative and ll angles exp ed by the tr vehicle expe Van-Traile 70 Y -7.137 8.425 curred in th nt roll angle actor unit co s with the pa ulations of ollision reak, and ecovered (Y positive rol erienced by ailer unit. Th rienced the r Truck on e trailer uni s. The maxi ntrolled wa rtial travers a ) l the e a t. mum s al

Figure 10. (0% Sup Figure 11. (0% Sup Roll Angle erelevation Roll Angle erelevation 2 - 24 of Tractor and 0% Cro of Trailer U and 0% Cro Unit on a T ss-Slope Bre nit on a T ss-Slope Bre angent ak) angent ak)

Figure Table 4 s tractor/si curvature values of under wh cross-slo shows gr Figure 14 influence maneuve Table 12. Illustrat hows the pr ngle-van-tra for design 0, 4, and 6 ich the vehi pe break of aphs of the r shows the of the cross r. 4. TruckSim Supere Cross- Speed Curve Recove Max Ne Max Ro ion of Max (0% Sup imary measu iler truck te speeds of 50 percent were cle did not r 4 percent; at oll angles e roll angles o -slope break Results fo Radi levation Slope Break ( (mph) Radius (ft) r (Y/N) g Roll Angle ll Angle (+ de imum Roll Tangent erelevation res of inter sted for the f and 60 mph tested. The ecover. For 50 mph, the xperienced b f the trailer at prescrib r Tractor/S us Curves w CSB) 50 926 Y (- deg) -6.16 g) 8.42 2 - 25 Angle of Tr from Truck and 0% Cro est output by ull traversa and supere simulation a speed of 6 non-recove y the tracto unit. The sc ed times dur ingle-Van- ith 4% Su 0% 60 1,500 Y 8 -8.702 9 11.812 2 actor/Singl Sim ss-Slope Bre TruckSim l severe dep levation of s were run u 0 mph, the v ry threshold r unit for th reenshots in ing the full Trailer Tru perelevatio 4% 4% 50 60 926 1,500 Y N -9.121 1.479 e-Van-Tra ak) for the simu arture path a 4 percent. C ntil a scenar ehicle faile was 6 perc e simulation Figure 15 s traversal sev ck Travers n 6% 50 6 926 1,5 N iler Truck o lations of a nd horizont ross-slope b io was creat d to recover ent. Figure s at 50 mph how the ere departu ing Minimu 0 00 n a al reak ed at a 13 , and re m

F F igure 13. R igure 14. R oll Angle o oll Angle o f Tractor U with 4% f Trailer U with 4% 2 - 26 nit at 50 m Supereleva nit at 50 mp Supereleva ph on Mini tion h on Minim tion mum Radiu um Radiu s Curve s Curve

Figure 1 Table 5 s tractor/si curvature and 70 m Table Table 6 s tractor/si curvature superelev encounte Table 6 Several i combinat  V d p t 5. Illustrat TruckSim hows the pr ngle-van-tra for design ph, the non- 5. TruckSim Supereleva Cross-Slop Speed (mph Curve Radi Recover (Y/ Max Neg Ro Max Roll An hows the pr ngle-van-tra for design ation, at spe ring cross-sl . Summary Superelevatio Cross-Slope Speed (mph) Curve Radius Recover (Y/N Max Neg Rol Max Roll Ang nsights draw ions of simu ehicle stab ependent up aths defined rajectory (i.e ion of Maxi at 50 mph o imary measu iler truck te speeds of 60 recovery cr Results fo Radi tion e Break (CSB ) us (ft) N) ll Angle (- de gle (+ deg) imary measu iler truck te speeds of 60 eds of 60 an ope break th of TruckSi Minimum n Break (CSB) (ft) ) l Angle (- deg) le (+ deg) n from the p lation scena ility followin on the defin for analysi ., full trave mum Roll A n Minimum res of inter sted for the f and 70 mph oss-slope br r Tractor/S us Curves w ) 0 60 1,330 Y g) -9.791 10.441 res of inter sted for the f and 70 mph d 70 mph, t resholds of m Results f Radius Cu 0% 60 1,200 Y -10.879 9.057 reliminary rios tested i g an encou ed vehicle d s in this rese rsal severe d 2 - 27 ngle of Tr Radius C est output by ull traversa and a 6 pe eak threshol ingle-Van- ith 6% Su % 70 2,040 Y -10.102 8.974 est output by ull traversa and 8 perc he truck exp 4 and 6 per or Tractor/ rves with 8 70 1,810 1 Y -11.24 7.588 simulations n this resear nter with a c eparture tra arch, only t eparture) w actor/Singl urve with 4 TruckSim l severe dep rcent supere ds were 4 an Trailer Tru perelevatio 6% 4% 60 7 1,330 2,0 N Y -11 25. TruckSim l severe dep ent superele erienced ro cent, respect Single-Van % Superele 8% 4% 60 70 ,200 1,810 N Y -12.5 22.91 which helpe ch were as ross-slope b jectory. Of he most sev as initially s e-Van-Trai % Superel for the simu arture path a levation. Fo d 6 percent ck Travers n 6 0 60 40 1,330 .221 631 for the simu arture path a vation. As w llover event ively. -Trailer Tr vation 6% 60 1,200 4 4 d to select t follows: reak is likel the three ve ere vehicle d imulated to ler Truck f evation lations of a nd horizont r speeds of , respectivel ing Minimu % 70 2,040 N lations of a nd horizont ith a 6 perc s when uck Traver 70 1,810 N he final y highly hicle departu eparture collect data rom al 60 y. m al ent sing re to

2 - 28 further define the final combinations of simulation scenarios for testing in this research. It is anticipated that the less severe vehicle departure trajectories will have less of an impact on the vehicle dynamics when testing the effect of cross-slope break on vehicle stability.  The observed roll angle trends will help in interpreting results even if the vehicle successfully completed the maneuver without rolling over.  From the preliminary investigations, there is no clear indication concerning the relationship between the interaction of vehicle speed and cross-slope break and vehicle stability. For the simulations conducted with 4 percent superelevation, the vehicle lost stability as both vehicle speed and cross-slope break increased. For example, at 4 percent cross-slope break, the vehicle completed the maneuver at 50 mph but rolled over at 60 mph, and at 6 percent cross-slope break the vehicle rolled over at 50 mph. However, at 6 percent and 8 percent superelevation, the tractor/single-van-trailer truck rolled over at a lower speed of 60 mph at a 4 percent cross-slope break while successfully completing the same maneuver at 70 mph with a cross-slope break of 4 percent. Appendix A includes results from all full traversal severe departure scenarios simulated for testing of the model and assessment of test scenarios for this research. The results in Appendix A for the full traversal severe departure scenarios show that trucks in emergency collision avoidance maneuvers rolled over in several cases without the presence of a pavement/shoulder cross-slope break and in several cases where a pavement/shoulder cross-slope break is present. It is also worth noting that several of the scenarios for the tractor/single-van-trailer truck for the full traversal severe departure vehicle trajectory simulated conditions very similar to those at the interchange ramp where the 2009 tanker truck rollover crash occurred in Indiana, supporting the validity of the simulation results. By contrast, the design situations for which simulation results are presented in Section 3.2 resulted in no truck rollover in any situation considered. 3.1.5 Range of Critical Geometric Design Elements and Variables Simulated In summary, to evaluate design criteria for pavement/shoulder cross-slope breaks, simulation scenarios were tested for values of the following critical geometric design elements and variables (although not all possible combinations were simulated):  Cross-Slope Break - 4 percent - 6 percent - 8 percent - 10 percent (as necessary)  Superelevation - 4 percent - 6 percent - 8 percent

2 - 29  Vehicle Type - Tractor/single-van-trailer truck (~80,000 lb), CG of ballast at 73 inches - Tractor/tanker-trailer truck (~80,000 lb), CG of ballast at 81 inches  Fully-loaded tanker (i.e., the dynamic effects of liquid sloshing assumed to be minimal) - Tractor/double-van-trailer truck (~80,000 lb), CG of ballast at 73 inches  Simulated selected scenarios to determine if behavior is consistent with tractor/single-van-trailer truck  Vehicle Speed - 30 mph - 50 mph - 60 mph - 70 mph (as necessary)  Vehicle Departure - Partial traversal moderate departure vehicle trajectory - Full traversal moderate departure vehicle trajectory 3.2 Summary of Simulation Results A total of 106 simulation scenarios were tested involving the partial and full traversal moderate departure vehicle trajectories, including 75 simulation scenarios for the tractor/single-van-trailer truck, 20 simulation scenarios for the tractor/tanker-trailer truck, and 11 simulation scenarios for the tractor/double-van-trailer truck. For each of the simulations, the stability of the vehicle was assessed based on the maximum positive (i.e., toward the passenger side of the vehicle) roll angle and whether the vehicle recovered after its maneuver. When rollover occurred in the simulations, the vehicle always rolled in the positive direction. Therefore, only maximum roll angles in the positive direction were considered when evaluating the stability of the vehicle. Complete roll angle results are provided in Appendix B, but the maximum roll angles for each design vehicle and departure trajectory are provided in the following sections, which summarize the recovery results for each vehicle simulated. 3.2.1 Tractor/Single-Van-Trailer Truck Simulations Table 7 shows that the tractor/single-van-trailer truck recovered from the departure maneuver and returned to the travel lane successfully for all simulation scenarios conducted with a partial traversal moderate departure vehicle trajectory. The corresponding maximum positive roll angles of the partial traversal moderate departure were benign; the largest roll angle of 4.5 degrees was recorded for 4 percent superelevation and 10 percent cross-slope break at 50 mph. Trends in the data showed that the maximum roll angle increased with an increase in the pavement/shoulder cross-slope break and the maximum roll angle decreased with an increase in the superelevation of the roadway. Since the maximum roll angles were benign, trends were difficult to pinpoint between stability and vehicle speed.

2 - 30 Table 7. Tractor/Single-Van-Trailer Truck with Partial Traversal Moderate Departure CSB 0% 4% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 S u p e r e l e v a t i o n 4% Y 0.7 Y -1.0 Y -1.2 Y 2.0 Y 2.5 Y 2.3 Y 2.5 Y 3.4 Y 3.3 Y 3.0 Y 4.5 Y 4.3 6% Y -1.5 Y -2.2 Y -2.4 Y -2.7 Y -1.5 Y -2.2 Y -2.4 Y -2.8 Y 1.2 Y 2.3 Y -2.5 Y -2.9 8% Y -2.7 Y -3.4 Y -3.6 Y -4.0 Y -2.7 Y -3.4 Y -3.7 Y -4.1 Y -2.7 Y -3.4 Y -3.8 Y -4.2 Shaded Cell – Indicates the design combination is not applicable per the AASHTO Green Book Empty Cell – Indicates the design combination was not simulated Y – Indicates the vehicle recovered from the departure maneuver and subsequent return to the travel lane Value in the cell indicates the maximum negative or positive roll angle experienced by the vehicle. A positive roll angle is a roll toward the passenger side of the vehicle, and a negative roll angle is a roll toward the driver side of the vehicle. Table 8. Tractor/Single-Van-Trailer Truck with Full Traversal Moderate Departure CSB 0% 4% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 S u p e r e l e v a t i o n 4% Y 0.9 Y -1.1 Y -1.3 Y 5.6 Y 4.8 Y 4.5 Y 7.1 Y 6.4 Y 6.3 Y 8.5 Y 8.1 Y 7.9 6% Y -1.5 Y -2.2 Y -2.4 Y -2.7 Y -1.5 Y -2.2 Y -2.4 Y -2.8 Y 1.2 Y 2.3 Y -2.5 Y -2.9 Y 6.3 8% Y -2.9 Y -3.6 Y -4.0 Y -5.0 Y 3.1 Y -4.0 Y -4.3 Y -5.5 Y 4.5 Y 4.4 Y -4.5 Y -5.6 Y 6.1 Y 6.1 Shaded Cell – Indicates the design combination is not applicable per the AASHTO Green Book Empty Cell – Indicates the design combination was not simulated Y – Indicates the vehicle recovered from the departure maneuver and subsequent return to the travel lane Value in the cell indicates the maximum negative or positive roll angle experienced by the vehicle. A positive roll angle is a roll toward the passenger side of the vehicle, and a negative roll angle is a roll toward the driver side of the vehicle.

2 - 31 Table 8 displays the recovery results for the tractor/single-van-trailer truck with a full traversal moderate departure vehicle trajectory. The largest maximum roll angle for the tractor/single-van- trailer truck with a full traversal moderate departure was 8.5 degrees for 4 percent superele- vation, 10 percent cross-slope break, and 30 mph. Similar trends existed between the superele- vation/cross-slope break and the roll angle of the vehicle as with the partial traversal moderate departure. Researchers also determined that the maximum roll angle decreased as the speed of the vehicle increased. This was primarily attributed to the use of larger design radii for the horizontal curves as design speeds increased, which is intended to provide additional stability to faster vehicles. As with the partial traversal moderate departure simulations, Table 8 shows that the tractor/single-van-trailer truck recovered from the departure maneuver and returned to the travel lane successfully for all simulation scenarios conducted with a full traversal moderate departure vehicle trajectory. 3.2.2 Tractor/Tanker-Trailer Truck Simulations Results from simulations for the tractor/single-van-trailer truck were used to select the simulation scenarios performed for the fully-loaded tractor/tanker-trailer truck. Table 9 shows the recovery results for the fully-loaded tractor/tanker-trailer truck with a partial traversal moderate departure vehicle trajectory. The tractor/tanker-trailer truck recovered from the departure maneuver and returned to the travel lane successfully for all simulation scenarios conducted with a partial traversal moderate departure vehicle trajectory. The maximum positive roll angles for the tractor/ tanker-trailer truck with a partial traversal moderate departure were larger than for the tractor/ single-van-trailer truck. The maximum roll angle for the selected scenarios was 4.1 degrees for 4 percent superelevation and 8 percent cross-slope break at 50 mph. Trends in the data showed that the maximum roll angle increased as the cross-slope break increased and the maximum roll angle decreased as the superelevation increased. The maximum roll angles were still quite benign, so identifying a trend between the speed of the vehicle and the maximum roll angle was difficult. Subsequently, the full traversal moderate departure vehicle trajectory was investigated for the tractor/tanker-trailer truck. Table 10 presents the recovery results for the tractor/tanker-trailer truck with a full traversal moderate departure vehicle trajectory. Again, the tractor/tanker-trailer truck recovered from the departure maneuver and returned to the travel lane successfully for all simulation scenarios conducted with a full traversal moderate departure vehicle trajectory. The maximum positive roll angles for the tractor/tanker-trailer truck with a full traversal moderate departure were larger than for the partial traversal moderate departure. The maximum roll angle was 9.8 degrees for 4 percent superelevation and 10 percent cross-slope break at 30 mph. Trends present in previous simulations were also present in these scenarios. Also the maximum roll angles decreased with increasing speed. 3.2.3 Tractor/Double-Van-Trailer Truck Simulations Investigation of the dynamic behaviors of the tractor/double-van-trailer truck began by selecting simulation scenarios to determine if the stability behavior of the tractor/double-van-trailer truck was similar to that of the tractor/single-van-trailer truck. As expected, the full traversal moderate departure scenarios yielded higher roll angles for the tractor/single-van-trailer truck and the tractor/tanker-trailer truck than the partial traversal moderate departure scenarios. Therefore,

2 - 32 Table 9. Tractor/Tanker-Trailer Truck with Partial Traversal Moderate Departure (Full Tanker Trailer) CSB 0% 4% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 S u p e r e l e v a t i o n 4% Y 3.0 Y 2.8 Y 4.1 Y 3.9 6% Y -2.0 Y -2.4 Y 2.9 Y 2.8 8% Shaded Cell – Indicates the design combination is not applicable per the AASHTO Green Book Empty Cell – Indicates the design combination was not simulated Y – Indicates the vehicle recovered from the departure maneuver and subsequent return to the travel lane Value in the cell indicates the maximum negative or positive roll angle experienced by the vehicle. A positive roll angle is a roll toward the passenger side of the vehicle, and a negative roll angle is a roll toward the driver side of the vehicle. Table 10. Tractor/Tanker-Trailer Truck with Full Traversal Moderate Departure (Full Tanker Trailer) CSB 0% 4% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 S u p e r e l e v a t i o n 4% Y 5.6 Y 5.3 Y 8.0 Y 7.3 Y 7.1 Y 9.8 6% Y 4.5 Y 4.4 Y 6.6 Y 6.3 Y 6.2 Y 5.5 8% Shaded Cell – Indicates the design combination is not applicable per the AASHTO Green Book Empty Cell – Indicates the design combination was not simulated Y – Indicates the vehicle recovered from the departure maneuver and subsequent return to the travel lane Value in the cell indicates the maximum negative or positive roll angle experienced by the vehicle. A positive roll angle is a roll toward the passenger side of the vehicle.

2 - 33 only the full traversal moderate departure vehicle trajectory scenarios were considered in the simulation combinations. Table 11 shows the recovery results for the tractor/double-van-trailer truck for a full traversal moderate departure vehicle trajectory. The tractor/double-van-trailer truck recovered from the departure maneuver and returned to the travel lane successfully for all simulation scenarios conducted with a full traversal moderate departure vehicle trajectory. The tractor/double-van- trailer truck with a full traversal moderate departure experienced higher roll angles than the tractor/single-van-trailer truck. The largest positive roll angle for the tractor/double-van-trailer truck was 9.9 degrees for the 4 percent superelevation, 10 percent cross-slope break, 50 mph simulation case. The maximum roll angle increased as cross-slope break increased, and the maximum roll angle decreased as superelevation increased. 3.3 Interpretation of Vehicle Dynamics Simulation Modeling Results The primary results of the vehicle dynamics simulation modeling were as follows:  The maximum roll angles were larger for the full traversal moderate departures compared to the partial traversal moderate departures. This was expected based on the results of previous research (Glennon et al., 1981).  All of the moderate departure scenarios, both partial and full traversal, resulted in recovery of the vehicles (i.e., none of the vehicles rolled over during the moderate departure simulation scenarios).  Maximum roll angles increased as the cross-slope break increased.  Maximum roll angles decreased as the superelevation increased.  Maximum roll angles decreased as speed increased. This was attributed to the increase in the design radius, which improves stability of faster vehicles. When interpreting the simulation results and drawing conclusions as to the need to change design guidance for pavement/shoulder cross-slopes, it is important to keep in mind the limitations of the overall approach to the vehicle dynamics simulations. Through vehicle dynamics simulation, combinations of key geometric design and other critical elements were evaluated to assess their impact on vehicle stability. The vehicle, vehicle trajectory, and geometric design factors that were varied and assessed included:  Vehicle type  Vehicle speed  Vehicle trajectory  Cross-slope break rate  Superelevation rate  Extent of vehicle encroachment In particular, three types of trucks were selected and simulated for evaluation of pavement/shoulder cross-slope breaks: a tractor/single-van-trailer truck, a fully-loaded tractor/tanker-trailer truck, and a tractor/double-van-trailer truck. For all three vehicle types, the target weight of the vehicle was approximately 80,000 lb. For the tractor/tanker-trailer truck, only the fully-loaded trailer condition was selected for evaluation of pavement/shoulder cross- slope breaks as it was assumed that the dynamic effects of liquid sloshing should be minimal for

2 - 34 Table 11. Tractor/Double-Van-Trailer Truck with Full Traversal Moderate Departure CSB 0% 4% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 S u p e r e l e v a t i o n 4% Y 7.6 Y 7.6 Y 9.4 Y 9.9 Y 9.7 6% Y 4.9 Y 4.9 Y 6.8 Y 6.8 8% Y 4.0 Y 4.0 Shaded Cell – Indicates the design combination is not applicable per the AASHTO Green Book Empty Cell – Indicates the design combination was not simulated Y – Indicates the vehicle recovered from the departure maneuver and subsequent return to the travel lane Value in the cell indicates the maximum negative or positive roll angle experienced by the vehicle. A positive roll angle is a roll toward the passenger side of the vehicle.

2 - 35 the fully-loaded trailer condition. Also, because existing vehicle dynamics simulation models do not have the capability to simulate the dynamic effects of liquid sloshing in a tanker trailer, the dynamic effects of a tractor/tanker-trailer truck were not truly simulated in this research, but rather the characteristics of the tractor/single-van-trailer truck were modified and simulated to represent situations analogous to a tractor/tanker-trailer truck. Therefore, the simulation results for the fully-loaded tractor/tanker-trailer truck should be interpreted within the limitations of the simulation model. Additional combinations of tractor/tanker-trailer trucks were represented and simulated in this research to learn more about the possible differences in stability between a fully-loaded tanker trailer and a partially-loaded tanker trailer. However, results from simulation scenarios for the partially-loaded tractor/tanker-trailer trucks were not considered when making decisions regarding design criteria for pavement/shoulder cross-slope breaks because the validity and accuracy of the simulation results could not be verified. Two vehicle trajectories were selected for evaluation of pavement/shoulder cross-slope breaks. For the partial traversal moderate departure vehicle trajectory, the vehicle gradually drifts from the middle of the travel lane along a path tangent to the roadway curvature and only the passenger-side tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. For the full traversal moderate departure vehicle trajectory, the vehicle gradually drifts from the middle of the travel lane along a path tangent to the roadway curvature and all tires of the vehicle traverse the cross-slope break and encroach onto the shoulder before the vehicle is steered back to the travel lane. This vehicle trajectory served as the basis for AASHTO’s current design policy on pavement/shoulder cross- slope break. A third vehicle trajectory (i.e., full traversal severe departure vehicle trajectory representing a collision avoidance maneuver) was used for testing the vehicle dynamics simulation model to verify that realistic results were being obtained and to assess the primary combinations of simulation scenarios to be tested in this research, but the results from the full traversal severe departure simulations were not directly considered when making decisions regarding design criteria for pavement/shoulder cross-slope breaks as this maneuver was considered too extreme to serve as a basis for determining design policy. Based upon the simulation results for the tractor/single-van-trailer truck, the fully-loaded tractor/tanker-trailer truck, and the tractor/double-van-trailer truck for the partial and full traversal moderate departure vehicle trajectories, there is no evidence to suggest the need to reduce the threshold value of 8 percent as the maximum recommended cross-slope break. Of the three truck types evaluated, none rolled over in the simulation scenarios for the partial traversal moderate departure vehicle trajectory, nor the full traversal moderate departure vehicle trajectory, even when the cross-slope break was as high as 10 percent.

2 - 36 SECTION 4. Crash-Based Safety Analysis This section presents the crash-based safety analysis intended to investigate the extent to which large cross-slope breaks may contribute to crashes near superelevated horizontal curves. The usefulness of this investigation was limited because there are no existing crash databases available that provide data pertaining to the presence, absence, or magnitude of pavement/ shoulder cross-slope breaks at crash sites. However, some information about the potential role of pavement/shoulder cross-slope breaks in crashes can be obtained from interpretation of crash analysis results. The objective of the crash-based safety analysis was to investigate the effect of pavement/ shoulder cross-slope break on crash frequency and severity. The location of the break in cross- slope was also of interest, as more vehicles will likely encounter a cross-slope break that occurs at the boundary between the traveled way and shoulder (i.e., at the edgeline) than at a cross-slope break located within the shoulder. A crash-based analysis is relevant because there may be little need to recommend a change in design policy for cross-slope breaks if the change is not expected to significantly reduce crash frequency and/or severity. The first step in the crash-based safety analysis was to examine FARS data to investigate the potential role of cross-slope breaks in crash frequency and severity and to assess where fatal truck rollover crashes occur. The second step in the analysis was to investigate whether cross- slope break rate is a significant predictor of crashes based on a sample of horizontal curves at which the pavement/shoulder cross-slope break was measured in the field. 4.1 Examination of FARS Crash Data The three most recent years of FARS crash data (2011 – 2013) were examined to investigate the potential role of cross-slope breaks in crash frequency and severity and to assess where fatal truck rollover crashes occur. This investigation is potentially relevant to assessing the need to update AASHTO’s current design policy for pavement/shoulder cross-slope breaks on horizontal curves. For the three-year period from 2011 to 2013, Table 12 below shows:  the total number of fatal crashes  the total number of truck-related fatal crashes  the number of truck-related fatal crashes classified as untripped rollover or overturn crashes  the number of untripped fatal single-vehicle truck rollover or overturn crashes  the number of untripped fatal single-vehicle tanker truck rollover or overturn crashes An “untripped” rollover crash occurs due to cornering forces that cause the vehicle to become unstable, while a “tripped” rollover crash occurs when a vehicle is sliding sideways and the tires strike an object such as a curb that causes the vehicle to trip and roll. Therefore, crashes related to cross-slope break would be expected to be classified as untripped and would typically involve only a single vehicle. Tanker truck crashes are considered specifically in this review of the FARS

2 - 37 data because the NTSB study that recommended a review of the AASHTO policy on pavement/shoulder cross-slope break was an outcome of the investigation of a fatal tanker truck rollover crash. The data in Table 12 show that:  0.24 percent of all fatal crashes and 2.1 percent of fatal truck crashes are classified as untripped truck-related rollover (or overturning) crashes  0.18 percent of all fatal crashes and 1.6 percent of fatal truck crashes are classified as untripped single-vehicle truck rollover (or overturn) crashes  0.04 percent of all fatal crashes and 0.36 percent of fatal truck crashes are classified as untripped single-vehicle tanker truck rollover (or overturn) crashes  A large majority of untripped rollover truck crashes are single-vehicle crashes  While tractor/tanker-trailer trucks make up only about 5 percent of the truck population (Harwood et al., 2003), they account for nearly a quarter of untripped fatal single-vehicle truck rollover crashes Table 12. Selected FARS Statistics of Fatal Truck and Tanker Truck Rollover Crashes (NHTSA, 2015) Year Number of Fatal Crashes Number of Fatal Truck Crashes Number of Untripped Fatal Truck Rollover Crashes Number of Untripped Fatal Single-Vehicle Truck Rollover Crashes Number of Untripped Fatal Single-Vehicle Tanker Truck Rollover Crashes 2011 29,867 3,401 61 49 12 2012 30,800 3,512 61 43 10 2013 30,057 3,571 94 72 16 Total 90,724 10,484 216 164 38 FARS data do not include information about the pavement/shoulder cross-slope break at crash locations; however, information about whether the crash occurred on a curve or tangent is provided. Larger pavement/shoulder cross-slope breaks are common on curved roadway sections but relatively rare on tangent roadway sections. Table 13 further breaks down the data on untripped fatal single-vehicle truck and untripped single-vehicle tanker rollover crashes (last two columns in Table 12) by whether the crash occurred on a tangent or curved section of roadway. This breakdown allows a comparison of untripped single-vehicle rollover crashes as a proportion of all fatal truck crashes between curved and tangent sections. The last row in Table 13 shows that while only approximately 17 percent of fatal truck crashes occur on curves, 56 percent of untripped single-vehicle truck rollover crashes (and 63 percent of those specifically involving tanker trucks) occur on curves. Looking at this another way, Table 13 shows that of all fatal truck crashes, those on curves are approximately six times more likely to be untripped single- vehicle rollovers than those on tangents (5.3 percent compared to 0.8 percent). When looking specifically at tanker trucks, that likelihood is even greater (1.4 percent compared to 0.2 percent). While the data clearly show that untripped single-vehicle rollover truck crashes are much more likely to occur on curves than on tangents, they do not provide clear information on the role of cross-slope break in these crashes. It is possible that many other factors influence the likelihood of a rollover crash on a curve, including curve radius, superelevation, speed, pavement condition, and others. However, the increased likelihood of rollover crashes on curved roadway sections,

2 - 38 where larger cross-slope breaks are more likely to occur, suggests that the role that cross-slope break plays in these crashes should be further investigated. Table 13. FARS Data: Distribution of Untripped Fatal Single-Vehicle Truck and Tanker Truck Rollover Crashes by Alignment (NHTSA, 2015) Year Number of Fatal Truck Crashes that Occurred on: Number (%) of Untripped Fatal Single- Vehicle Truck Rollover Crashes that Occurred on: Number (%) of Untripped Fatal Single-Vehicle Tanker Truck Rollover Crashes that Occurred on: Curved Alignment Tangent Alignment Curved Alignment Tangent Alignment Curved Alignment Tangent Alignment 2011 568 2,797 21 (3.7%) 28 (1.0%) 6 (1.1%) 6 (0.2%) 2012 586 2,866 25 (4.3%) 17 (0.6%) 7 (1.2%) 3 (0.1%) 2013 569 2,948 46 (8.1%) 26 (0.9%) 11 (1.9%) 5 (0.2%) Total 1,723 8,611 92 (5.3%) 71 (0.8%) 24 (1.4%) 14 (0.2%) Percentages are based on the number of fatal truck crashes presented in the corresponding first two columns of the table. 4.2 Analysis of Crashes on Horizontal Curves Because existing crash databases do not include quantitative information on the presence, absence, or magnitude of pavement/shoulder cross-slope breaks, a safety analysis was conducted for a small sample of horizontal curves at which the pavement/shoulder cross-slope break was measured in the field. Roadway design and crash data were collected for a sample of horizontal curves covering a range of superelevations and cross-slope breaks collected in four states—Iowa, Kansas, Texas, and Washington. The safety analysis was designed to evaluate the effect of pavement/shoulder cross-slope break on crash frequency and severity. 4.2.1 Site Selection and Field Data Collection Several state highway agencies were contacted regarding the availability of horizontal curve data and information on superelevation, shoulder cross-slopes, and pavement/shoulder cross-slope breaks and possible means of gathering such information from highway plans, electronic roadway inventories, or statewide/regional lidar datasets (e.g., datasets generated from mobile laser scanning systems for analysis of curves, slope, superelevation, and pavement distress) for site selection purposes. However, very few highway agencies include horizontal curve data in their inventory files, and none of them had information on the magnitude of shoulder cross- slopes or pavement/shoulder cross-slope breaks to assist with the site selection process. After discussing options with various states on means to identify horizontal curves with a range of superelevations and cross-slope breaks for potential inclusion in this analysis, it was decided to consider sites from Iowa, Kansas, Texas, and Washington for analysis based on the availability and willingness of DOT staff to assist with this project and on whether horizontal curve data are available in electronic files. It was further determined that the analysis should focus on horizontal curves on rural two-lane highways and freeways with speed limits of 55 mph or higher. This decision was based on a NHTSA report (Deutermann, 2000) characterizing fatal rollover crashes. Deutermann examined the characteristics of passenger vehicles and their drivers involved in fatal rollover crashes and reported that most of the crashes (90 percent) occurred on undivided two-way roads or divided roads with no barriers. Also, a majority of these rollover crashes occurred on rural roads with speed limits of 55 mph or higher.

2 - 39 A combination of two approaches was used to identify horizontal curve sites for consideration in the analysis. The first approach was to identify potential sites based on their site characteristics, using the following criteria:  Routes should contain horizontal curves that likely have a wide range of values for superelevation.  Routes should not have been recently overlaid or resurfaced.  Sites should be on mainline roadways of higher functional class (i.e., 50 to 70 mph design speed) and should be expected to carry truck traffic.  Horizontal curves should have a minimum paved shoulder width of 4 ft.  Cross-slope break between the shoulder and the roadway should be located at or near the edgeline.  Roadside should be relatively forgiving with few (if any) fixed objects present. The second approach was to first identify horizontal curves on rural two-lane highways and freeways where truck rollover crashes had occurred, and then to further down-select from the locations where these crashes occurred based upon the site characteristic criteria listed above. In some instances, Google Earth and Google Street View were used to review the sites prior to field data collection. With assistance from DOT staff in Iowa and Kansas, several sites in those states were identified for potential inclusion in the analysis. Roadway inventory and crash databases for Texas and Washington were also used by the research team to identify sites for potential inclusion in the analysis. Once sites were identified, the research team collected the following site characteristics in the field (or obtained from electronic databases):  Roadway type (rural two-lane highway or freeway)  Horizontal curve radius  Horizontal curve length  Maximum superelevation  Average lane width  Shoulder cross-slope  Average shoulder width  Location of pavement/shoulder cross-slope break relative to the edgeline  Grade  Direction of travel  Direction of curve (left or right)  Presence of rumble strips  Speed limit  Presence of curve warning and/or advisory speed signs  Pavement type  Shoulder type Cross-slope break was then calculated as the algebraic difference between the maximum superelevation of the travel lane and the shoulder cross-slope.

2 - 40 4.2.2 Descriptive Statistics—Site Characteristics and Crash Counts A total of 64 curves on rural two-lane highways and 44 curves on rural freeways were ultimately used in the analysis. Table 14 shows the range of site characteristics of the horizontal curves, separately for each state and roadway type. Total curve length in the third column represents the sum of the lengths of all the curves of a given roadway type in a given state. Columns for curve length, curve radius, grade, superelevation, cross-slope break, lane width, shoulder width, location of break, speed limit, advisory speed, and annual average daily traffic (AADT) show the range (minimum to maximum) for the respective data element. The next-to-last two columns show the number of curves with a curve warning sign and the range of advisory speeds when denoted at the sites, respectively. The distributions of the site characteristics shown in Table 14 were evaluated to check whether their values were approximately evenly distributed across state-roadway type combinations. Unfortunately, this was not the case. Some site characteristics showed very little variability within a state-roadway type combination (e.g., all but one curve on Texas freeways had a curve radius of 5,730 ft; percent grade showed very little variability in all state-roadway type combinations as shown in Table 14). Other variables were confounded with each other; for example, freeway AADTs and state could clearly be ordered as follows: lowest AADTs in Washington, next lowest AADTs in Iowa, medium AADTs in Texas, and highest AADTs in Kansas. The variable of most interest in this analysis is cross-slope break. Initially, the location of the break was also considered but was subsequently dropped. The joint distribution of these two variables is shown in Figure 16, separately for each roadway type-state combination. Only curves on Washington roadways had negative cross-slope breaks. Curves on freeways in Iowa and Kansas and on two-lane roadways in Texas showed very little variability in cross-slope break. Similarly, location of break (measured as distance in feet from the edgeline) hardly varied for curves for either roadway type in Iowa or freeways in Kansas. Based upon the joint distribution of cross-slope break and location of break, the location of the cross-slope break was excluded from the analysis. The following site characteristics were deemed of greatest importance and considered as independent variables in the statistical modeling effort:  AADT  Cross-slope break  Degree of curvature, calculated as ହ଻ଷ଴ோ௔ௗ௜௨௦ (where the curve radius is expressed in feet)  Superelevation (percent) Although originally calculated on a continuous scale, cross-slope break was categorized as follows due to the distribution of that variable over the range of values:  None: ≤ 0 percent  Small: > 0 to ≤ 4 percent  Large: > 4 percent

2 - 41 Table 14. Range of Site Characteristics of Horizontal Curves Included in Crash-Based Safety Analysis State Number of Curves Total Curve Length (ft) Curve Length Range (ft) Curve Radius Range (ft) Grade Range (%) Super- elevation Range (%) Cross-Slope Break Range (%) Average Lane Width Range (ft) Average Shoulder Width Range (ft) Location of Break Range (ft) Speed Limit (mph) Number of Sites with Curve Sign Advisory Speed (mph) Range of AADT (vpd) Rural Freeways (44 Sites) IA 7 12,672 1,056 – 2,112 2,473 – 6,966 -1.4 – 0.5 1.8 – 3.4 3.6 – 5.1 10.5 – 12 10 – 11 2 – 2.5 65 0 NA 18,330 – 25,000 KS 3 5,280 1,584 – 2,112 1,823 – 3,750 0 – 0.5 3.5 – 4.5 6.6 – 7.0 12 10 2 – 2.5 70 0 NA 55,106 – 55,376 TX 21 33,454 700 – 2,830 5,730 – 7,639 -1.1 – 1.7 0.6 – 3.5 0 – 9.8 11.5 – 12.5 5 – 13 0 – 12 75 0 NA 26,254 – 63,004 WA 13 17,453 535 – 2,323 1,500 – 2,800 -2.0 – 2.4 3.4 – 8.4 -0.3 – 1.5 11.3 – 12.4 3.8 – 10.8 0 – 10.2 70 0 NA 7,053 – 10,774 Rural Two-Lane Roads (64 Sites) IA 8 20,064 2,112 – 3,168 1,244 – 4,770 -1.2 – 1.4 2.2 – 4.9 3.7 – 8.2 10 – 11.5 8.5 – 16a 1 – 3 55 0 NA 1,152 – 3,546 KS 0 TX 33 25,562 354 –1,520 318 – 2,865 -4.9 – 3.9 1.1 – 8.8 0 – 4.5 10 – 11.5 1 – 6 0 – 6 55 – 65 27 30 – 50 294 – 2,084 WA 23 23,692 301–3,477 409 – 2,700 -2.7 – 3 0.3 – 7 -1.8 – 4.5 10.9 – 17.6 3 – 20.73 0 – 8.4 55 – 65 5 30 – 55 667 – 4,674 a Includes gravel shoulder width.

Fig Crash da  I  K  T  W Crash co were talli single-ve crash typ relationsh feasible— highways a sufficie modeling ure 16. Cro ta for the fol owa (2004 – ansas (200 exas (2003 ashington unts for the ed accordin hicle truck r e, roadway t ip between only four o experience nt number o . ss-Slope Br lowing year 2014) 0 – 2013) – 2009) (2007 – 201 64 curves on g to the four ollover cras ype, and sta cross-slope f the 44 cur d any single f single-veh eak vs. Loc s were analy 1) rural two-l crash types hes. Table 1 te. This tabl break and s ves on rural -vehicle tru icle and sin 2 - 42 ation of Br zed: ane highway : total, singl 5 shows the e clearly sh ingle-vehicl freeways an ck rollover c gle-vehicle r eak by Roa s and the 4 e-vehicle, s distribution ows that a sa e truck rollo d one of the rashes in th ollover cras dway Type 4 curves on ingle-vehicl of these cra fety analys ver crashes 64 curves o e five-year p hes were av and State rural freewa e rollover, a sh counts b is to quantif was not n rural two eriod. How ailable for ys nd y y a -lane ever,

2 - 43 Table 15. Crash Counts by Crash Type, Roadway Type, and State Crash Type Roadway Type (Number of Sites) Crash Count (Range) State IA KS TX WA Total Freeway (44) 0 0 0 2 1 1-5 2 0 11 11 6-10 2 0 4 1 11-15 1 0 1 0 16-20 1 0 0 0 21+ 1 3 3 0 Rural Two-Lane Highway (64) 0 0 N o Si te s 20 0 1-2 2 10 9 3-4 3 3 11 5-6 1 0 2 7+ 2 0 1 Single Vehicle Freeway (44) 0 0 0 3 1 1-5 3 0 14 11 6-10 2 0 3 1 11-15 1 0 0 0 16-20 0 0 1 0 21+ 1 3 0 0 Rural Two-Lane Highway (64) 0 1 N o Si te s 20 0 1-2 2 10 11 3-4 2 3 10 5-6 1 0 1 7+ 2 0 1 Single-Vehicle Rollover Freeway (44) 0 5 0 8 4 1-5 2 2 12 9 6-10 0 1 0 0 11-15 0 0 1 0 16-20 0 0 0 0 21+ 0 0 0 0 Rural Two-Lane Highway (64) 0 6 N o Si te s 29 12 1-2 2 4 10 3-4 0 0 0 5-6 0 0 0 7+ 0 0 1 Single-Vehicle Truck Rollover Freeway (44) 0 7 1 19 13 1-5 0 2 2 0 6-10 0 0 0 0 11-15 0 0 0 0 16-20 0 0 0 0 21+ 0 0 0 0 Rural Two-Lane Highway (64) 0 8 N o Si te s 33 22 1-2 0 0 0 3-4 0 0 0 5-6 0 0 1 7+ 0 0 0

2 - 44 4.2.3 Analysis Approach The safety effect of cross-slope break on curves was estimated separately for rural two-lane highways and rural freeways and the following remaining three crash types:  Total crashes  Single-vehicle crashes  Single-vehicle rollover crashes The six models were each estimated using a generalized linear model approach assuming a negative binomial (NB) distribution of crash counts using the combined crash data and selected roadway geometrics. In all models, state was used as a random effect to account for correlation between curves within states. All analyses were performed using PROC GENMOD of SAS/STAT software, Version 9.4 of the SAS ® System. AADT was initially considered for inclusion as a fixed factor, and the coefficient of log (AADT) would have been estimated. However, as mentioned earlier, AADT ranges were almost non- overlapping from state to state. Therefore, estimating the shape of the relationship between crash counts and AADT over the entire AADT range became problematic. It was thus decided to model crashes per exposure expressed in number of crashes per million vehicle miles traveled (MVMT) per year, rather than crashes per mile per year, to account for varying traffic volumes and curve segment length across sites. The final three remaining factors modeled were:  Cross-slope break (at three levels: none/small/large)  Degree of curvature, calculated as ହ଻ଷ଴ୖୟୢ୧୳ୱ  Superelevation (percent) For each roadway type, the final regression model is of the following general form: ்ܰ௢௧,ௌ௏,௢௥ ௌ௏ோ ൌ ܯܸܯܶ ൈ ݁ݔ݌ሾܽ ൅ ܾ ൈ ܫ஼ௌሺ௦௠௔௟௟ሻ ൅ ܿ ൈ ܫ஼ௌሺ௟௔௥௚௘ሻ ൅ ݀ ൈ ܦܥ ൅ ݁ ൈ ܵሿ (3) where: NTot, SV, or SVR = Number of total, single-vehicle, or single-vehicle rollover crashes per year on the curve ICS(small) = Indicator variable for small cross-slope break; 1 if small; 0 otherwise ICS(large) = Indicator variable for large cross-slope break; 1 if large; 0 otherwise DC = Degree of curvature = ହ଻ଷ଴ ௙௧஼௨௥௩௘ ோ௔ௗ௜௨௦ ሺ௙௧ሻ S = Superelevation (percent) LC = Curve length (mi) AADT = average annual daily traffic volume on the curve (veh/day) MVMT = exposure in MVMT on the curve per year (LC×AADT×365×10-6) a, b, c, d, e = regression coefficients to be estimated The base condition represented in Equation (3) is a curve with a zero cross-slope break; that is, both indicator variables, ICS(small) and ICS(large), are equal to 0.

2 - 45 If either degree of curvature or superelevation were not significant at the 10-percent level in the model including all three factors, they were removed from the model using backward stepwise elimination (i.e., the least significant of the two factors was excluded and the model rerun with the remaining factors, etc.). However, cross-slope break was kept in all models to make inference about its effect on safety. 4.2.4 Results of Analysis of Crashes on Horizontal Curves The final modeling results for the three crash types are shown in Table 16, separately for rural two-lane highways and rural freeways. These results show the following, based on the data from all states:  Degree of curvature has a significant effect on all three crash types on rural freeways only.  Superelevation has a significant effect on total crashes only and only on rural freeways.  Cross-slope break was not significant in any of the models. In other words, based on the available data, there is insufficient evidence to show that cross-slope break has a significant effect on total crashes, single-vehicle, or single-vehicle rollover crashes. Table 16. ANOVA Results—Statistical Significance Levels Associated with Factors in Negative Binomial Regression Models Crash Type Roadway Type (Number of Sites) Number of Sites in Cross-Slope Break Categories: None/Small/Large Factor (Levels) P-Valuea Statistically Significant? Total Rural Freeway (44) 24/6/14 Cross-slope break (none/small/large) 0.15 No Degree of curvature 0.09 At 10% level Superelevation 0.07 At 10% level Rural Two-Lane Highway (64) 43/12/9 Cross-slope break (none/small/large) 0.27 No Degree of curvature >0.10 No Superelevation >0.10 No Single Vehicle Rural Freeway (44) 24/6/14 Cross-slope break (none/small/large) 0.41 No Degree of curvature 0.09 At 10% level Superelevation >0.10 No Rural Two-Lane Highway (64) 43/12/9 Cross-slope break (none/small/large) 0.74 No Degree of curvature >0.10 No Superelevation >0.10 No Single- Vehicle Rollover Rural Freeway (44) 24/6/14 Cross-slope break (none/small/large) 0.17 No Degree of curvature 0.07 At 10% level Superelevation >0.10 No Rural Two-Lane Highway (64) 43/12/9 Cross-slope break (none/small/large) 0.47 No Degree of curvature >0.10 No Superelevation >0.10 No Single- Vehicle Truck Rollover Rural Freeway (44) No analysis performed Rural Two-Lane Highway (64) No analysis performed a A p-value >0.10 indicates that the factor was excluded from the final model; it is shown here for completeness only.

2 - 46 However, review of these results points to the fact that the sample sizes are too small and the datasets from the various states too disparate to draw meaningful conclusions. In particular, the effect of degree of curvature (equivalent to radius of curvature) is statistically significant only for rural freeways, but not for rural two-lane highways. By contrast, the AASHTO Highway Safety Manual (AASHTO, 2010, 2014) shows statistically significant relationships between crash frequency and radius of curvature for both freeways and rural two-lane highways. Since the sample sizes for the available datasets are not large enough to demonstrate the strong established relationship of horizontal curve radius to crashes, they are certainly not large enough to determine the existence of the more subtle relationship of pavement/shoulder cross-slope breaks to very specific crash types. 4.3 Interpretation of Crash-Based Safety Analysis The crash-based safety analysis was conducted in two parts. First, FARS data were examined for indications that pavement/shoulder cross-slope break may be a contributing factor in untripped truck rollover crashes. FARS data do not include information about the presence, absence, or magnitude of pavement/shoulder cross-slope breaks, but do provide an indication of whether a crash occurred on a tangent or curved section of the roadway. Because higher pavement/shoulder cross-slope breaks are rarely found on tangent sections but are commonly used on curved sections, assessing the location of truck rollover crashes can provide a rationale for a more detailed analysis. The FARS data showed that the proportion of fatal untripped single-vehicle truck rollover crashes to all fatal truck crashes is much higher on curves than on tangents, indicating a more detailed analysis of the effect of pavement/shoulder cross-slope break on crash frequency and severity was warranted. The second part of the crash-based safety analysis was to investigate, with a necessarily smaller but more detailed database, whether pavement/shoulder cross-slope break rate is a significant predictor of crashes on horizontal curves. This detailed analysis included 64 curves on rural two- lane highways and 44 curves on rural freeways in four states at which pavement/shoulder cross- slope breaks were measured and several years of crash data were collected. The study analyzed four crash types: total, single-vehicle, single-vehicle rollover, and single-vehicle truck rollover crashes. The primary crash type of interest was to be single-vehicle truck rollover crashes for consistency with the vehicle dynamics simulation modeling analysis that was performed and focused on single-vehicle truck rollover crashes. However, based upon the limited dataset that was developed, a safety analysis to quantify the relationship between cross-slope break and single-vehicle truck rollover crashes was not feasible. Therefore, the analysis focused on quantifying the safety effect of pavement/shoulder cross-slope break on curves for the following three crash types: total crashes, single-vehicle crashes, and single-vehicle rollover crashes. Based on the available data, the analysis results provide insufficient evidence to show that cross-slope break has a significant effect on total crashes, single-vehicle, or single-vehicle rollover crashes. Based on the three categories of pavement/shoulder cross-slope break that were considered in the analysis (none, small, and large), there is no evidence to suggest that horizontal curves with large cross-slope breaks ( > 4 percent) are expected to have more crashes than sites with cross-slopes breaks less than or equal to 4 percent. The results of the detailed crash-based safety analysis do not definitively answer the question of whether the magnitude of the pavement/shoulder cross-slope break affects crash frequency and

2 - 47 severity. While no statistically significant effect of pavement/shoulder cross-slope breaks on any of the three crash measures could be found, the available sample size was simply too small to draw meaningful conclusions.

2 - 48 SECTION 5. Conclusions and Future Research Needs The objective of this research was to assess AASHTO’s current design policy for pavement/ shoulder cross-slope breaks on the outside of horizontal curves to determine whether updates in design criteria are recommended. AASHTO’s current policy states that shoulder slopes that drain away from the paved surface on the outside of a superelevated horizontal curve should be designed to avoid too great a cross-slope break, calculated as the algebraic difference between the cross-slopes of the traveled way and the shoulder. AASHTO recommends that the cross- slope break be limited to a maximum of approximately 8 percent. Two separate work plans were executed to fulfill the objectives of this research: one work plan involved vehicle dynamics simulation modeling and the other involved a crash-based safety analysis. Through vehicle dynamics simulation, combinations of key geometric design and other critical elements were evaluated to assess their impact on vehicle stability, including:  Vehicle type  Vehicle speed  Vehicle trajectory  Cross-slope break rate  Superelevation rate  Extent of vehicle encroachment The primary measures of interest that were obtained as output from the vehicle dynamics simulation modeling were whether the vehicle recovered from the maneuver or rolled over and the maximum roll angles experienced by the vehicle. The simulation results for the tractor/ single-van-trailer truck, the fully-loaded tractor/tanker-trailer truck, and the tractor/double-van- trailer truck for the partial and full traversal moderate departure vehicle trajectories were the primary sources of information for drawing conclusions related to the need to change AASHTO’s current design policy for pavement/shoulder cross-slope breaks. For these respective simulation scenarios, none of the vehicles rolled over, even when the cross-slope break was as high as 10 percent. It was only in the simulation scenarios for the full traversal severe departure vehicle trajectories that vehicles were not always able to recover when encountering a pavement/shoulder cross-slope break of 8 percent or less. This scenario was an extreme maneuver studied primarily for purposes of testing the vehicle dynamics simulation model to demonstrate that truck rollovers occurred in some cases for which rollovers would be considered likely; trucks would be likely to rollover while making his extreme maneuver even if no pavement/shoulder cross-slope break were present. Thus, based upon the vehicle dynamics simulation modeling, it was concluded that there is no need to recommend a change to AASHTO’s current design policy for pavement/shoulder cross-slope break on the outside of horizontal curves. The crash-based safety analysis was intended to investigate the extent to which large cross-slope breaks may contribute to crashes near superelevated horizontal curves. FARS data show that untripped single-vehicle rollover truck crashes are much more likely to occur on curves than on tangents. The increased likelihood of rollover crashes on curved roadway sections, where larger cross-slope breaks are more likely to occur, suggests that a more detailed analysis of the effect of pavement/shoulder cross-slope break on crash frequency and severity was warranted.

2 - 49 Results from the detailed analysis of crashes for 64 curves on rural two-lane highways and 44 curves on rural freeways in four states provide insufficient evidence to show that cross-slope break has a significant effect on total crashes, single-vehicle, or single-vehicle rollover crashes. The results from the detailed crash-based safety analysis indicate that the available sample size is too small to determine whether pavement/shoulder cross-slope breaks result in more frequent or more severe crashes. The results from the vehicle dynamics simulation modeling indicate that there is no need to recommend a change to AASHTO’s current design policy for pavement/shoulder cross-slope break on the outside of horizontal curves. There is some evidence to suggest that the recom- mended maximum cross-slope break could be increased to 10 percent. In particular, this may be possible for curves with higher superelevations, as vehicle dynamics simulation modeling showed that maximum roll angles decreased with increases in superelevation. However, this potential change to AASHTO’s design policy for pavement/shoulder cross-slope breaks on the outside of horizontal curves is not recommended as the current design policy with a maximum of 8 percent for the pavement/shoulder cross-slope break is the more conservative approach and because the scenario for tanker trucks with sloshing liquid could not be evaluated. While no changes to AASHTO’s current design policy for pavement/shoulder cross-slope break are recommended, use of mitigation measures such as edgeline or shoulder rumble strips that reduce the potential for full traversal departure onto the shoulder in high pavement/shoulder cross-slope break locations are recommended. Measures that reduce the likelihood of full traversal departures and limit encroachments onto the shoulder to partial traversal departures will reduce the maximum roll angles experienced by errant vehicles that encroach onto the shoulders. When installed, rumble strips should be placed on or close to the edgeline rather than further out onto the shoulder to reduce the likelihood of full traversal departures. Further research to more accurately incorporate tractor/tanker-trailer trucks in the vehicle dynamics simulation analysis is recommended once the capabilities of vehicle dynamics models become sophisticated enough to simulate the dynamic effects of liquid sloshing in a tanker trailer. It would also be desirable to substantially increase the size of the dataset used in the detailed crash-based safety analysis by including more sites from additional states with a wider range of cross-slope breaks and more crashes so the relationship between cross-slope break and single-vehicle truck rollover crashes can be more fully evaluated.

2 - 50 SECTION 6. References American Association State Highway and Transportation Officials (AASHTO). Highway Safety Manual, 2010. American Association State Highway and Transportation Officials (AASHTO). Highway Safety Manual Supplement, 2014. American Association of State Highway and Transportation Officials (AASHTO). Manual for Assessing Safety Hardware, First Edition. Washington, DC. 2009. American Association of State Highway and Transportation Officials (AASHTO). A Policy on Geometric Design of Highways and Streets. Washington, DC. 2011. American Association of State Highway and Transportation Officials (AASHTO), Roadside Design Guide, Washington, DC, 2011. Colorado Department of Transportation. Roadway Design Guide, 2011. Florida Department of Transportation. Green Book, 2011. Garcia, L.O., F.R. Wilson, and J.D. Innes. “Heavy Truck Dynamic Rollover: Effect of Load Distribution, Cargo Type and Road Design Characteristics.” In Transportation Research Record 1851. Transportation Research Board, Washington, DC. 2003. Glennon, J. C., T. R. Neuman, R. R. McHenry, and B. G. McHenry. HVOSM Studies of Cross- Slope Breaks on Horizontal Curves, Federal Highway Administration, Contract No. DOT-FH- 11-9575. Washington, DC. 1981. Glennon, J. C., B. G. McHenry, and T. R. Neuman. HVOSM Studies of Highway Cross-Slope Design, FHWA-RD-84-006, Federal Highway Administration, Contract No. DOT-FH-11-9575. Washington, DC. 1983. Glennon, J.C., and G.D. Weaver. “Highway Curve Design for Safe Vehicle Operations.” In Highway Research Record 371. Transportation Research Board. Washington, DC. 1972. Harwood, D.W., D.J. Torbic, K.R. Richard, W.D. Glauz, and L. Elefteriadou. Review of Truck Characteristics as Factors in Roadway Design. NCHRP Report 505. National Cooperative Highway Research Program. Washington, DC. 2003. Kentucky Transportation Cabinet. Geometric Design Guidelines. Kim, J.-H., S. Hayakawa, T. Suzuki, K. Hayashi, S. Okuma, N. Tsuchida. “Modeling of Driver's Collision Avoidance Maneuver Base on Controller Switching Model.” IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 35 (6), 1131-1143. IEEE Systems, Man, and Cybernetics Society. New York, NY. 2005. Louisiana Department of Transportation and Development. Road Design Manual, 2009.

2 - 51 Maeda, T., N. Irie, K. Hidaka, and H. Nishimura. “Performance of Driver-Vehicle System in Emergency Avoidance.” International Automotive Engineering Congress and Exposition. Society of Automotive Engineers. Detroit, MI. 1977. Mississippi Department of Transportation. Roadway Design Manual, 2001. National Transportation Safety Board (NTSB). Rollover of a Truck-Tractor and Cargo Tank Semitrailer Carrying Liquefied Petroleum Gas and Subsequent Fire, Indianapolis, Indiana, October 22, 2009. NTSB/HAR-11/01, Washington, DC, 2011. Nevada Department of Transportation. Roadway Design Manual, 2010. New Jersey Department of Transportation. Roadway Design Manual, 2011. Rice, R.S. and F. Dell’Amico. An Experimental Study of Automobile Driver Characteristics and Capabilities. Calspan Report No. ZS-5208-K-1. Calspan Corporation. Buffalo, NY. March 1974. Texas Department of Public Safety. Texas Commercial Motor Vehicle Driver Handbook. Austin, TX. 2010. Torbic, D.J., M.K. O’Laughlin, D.W. Harwood, K.M. Bauer, C.D. Bokenkroger, L.M. Lucas, J.R. Ronchetto, S. Brennan, E. Donnell, A. Brown, T. Varunjikar. Superelevation Criteria for Sharp Horizontal Curves on Steep Grades. NCHRP Report 774. National Cooperative Highway Research Program. Washington, DC. 2013.

APPE Simul Vehic This appe departure simulatio combinat departure in the tra all tires o vehicle is collision no pavem reasonab maneuve simulatio pavemen basis for departure of the de Table A- in TruckS tractor/si vehicle (A on the W determin The dyna because e effects of considere trailer tru represent TruckSim Four com to learn m a partiall character partially- While Tr some rea combinat NDIX A. ation Re le Trajec ndix presen vehicle traj n model to v ions of simu , the steerin vel lane (i.e f the vehicle steered bac avoidance m ent/shoulde le basis for h r simulated ns were not t/shoulder c determining and full tra sign criteria 1 presents th im based o ngle-van-tra ASHTO, 2 B-20D desig ed from Tab mic effects xisting veh liquid slosh d within the cks with dif situations a . binations o ore about t y-loaded tan istics than a loaded tank uckSim doe sonable app ions of tract sults fo tory ts simulatio ectory. This erify that re lation scena g inputs rep ., a collision traverse th k to the trav aneuver so r cross-slop ighway des as part of th directly con ross-slope b design poli versal mode for pavemen e primary c n the full tra iler truck an 011), and th n vehicle. T le 4-2 in MA of tractor/ta icle dynamic ing in a tan current stat fering heigh nalogous to f tractor/tank he possible ker trailer. A fully-loaded er trailer due s not allow f roximations or/tanker-tr r Full T n results for vehicle traj alistic resul rios to be te resent the si avoidance m e cross-slop el lane. Thi extreme tha e break is pr ign. This ve is research. sidered whe reaks as this cy. (Only si rate departu t/shoulder c haracteristic versal sever d tractor/tan e dimension he heights SH (AASH nker-trailer s models do ker trailer. G e-of-the-art ts of the cen fully- and p er-trailer tru differences i partially-l tanker trai to liquid sl or a fully dy were made ailer trucks m 2A-1 raversa scenarios b ectory was u ts were bein sted in this tuation whe aneuver) w e break and s vehicle traj t trucks are esent. There hicle traject The results f n making d maneuver w mulation res re vehicle tr ross-slope b s of the veh e departure ker-trailer t s for the tra for the cente TO, 2009). trucks were not have th iven that th of vehicle d ter of gravi artially-load cks were re n stability b oaded tanke ler, because oshing durin namic liqui with static l odeled in t l Severe ased on the sed for test g obtained a research. Fo re the driver hile travers encroach on ectory repr likely to rol fore, this sc ory represen rom the full ecisions reg as consider ults for the ajectories w reaks.) icles for wh vehicle traje ruck were b ctor/double rs of gravity not truly sim e capability e effect of s ynamics sim ty for the tra ed tractor/ta presented a etween a fu r trailer has the center o g a departu d load in a s oads. Chara his research Depar full traversa ing the vehi nd to asses r the full tra steers to av ing a horizo to the shoul esents an em lover in som enario does ted the mos traversal se arding desig ed too extre partial trave ere selected ich simulati ctory. Dime ased on the -van-trailer t of the trail ulated in th to simulate loshing liqu ulation, tra iler cargo w nker-trailer nd simulated lly-loaded ta different dy f gravity can re/recovery imulated ta cteristics of included: ture l severe cle dynamic s the primary versal sever oid an obsta ntal curve a der before t ergency e cases even not constitu t extreme vere departu n criteria fo me to serve rsal modera for evaluat ons were tes nsions for t WB-19 desi ruck were b er ballasts w is research the dynami id cannot be ctor/single-v ere simulate trucks in in this rese nker trailer namics shift in a maneuver. nker trailer, the four s e cle nd he if te a re r as a te ion ted he gn ased ere c an- d to arch and

2A-2 Table A-1. Primary Vehicle Input Parameters Used in TruckSim Vehicle Inputs Tractor/Single- Van-Trailer Truck Tractor/Tanker-Trailer Truck Tractor/Double- Van-Trailer Truck Full Tanker Lower Half Tanker Upper Half Tanker Right Half Tanker Tractor type Sleeper Cab Sleeper Cab Sleeper Cab Sleeper Cab Sleeper Cab Day Cab Total mass of vehicle1 (lb) 80,158 80,161 57,482 57,482 57,482 80,151 Total mass of tractor (lb) 18,629 18,629 18,629 18,629 18,629 12,698 Total mass of trailer(s) (lb) 15,999 15,999 15,999 15,999 15,999 17,262 (two trailers + dolly) Trailer ballast (laden) (lb) 45,358 45,358 22,679 22,679 22,679 48,610 (both) Height of center of gravity of trailer ballast (in) 73 81 66.3 95.1 81 73 Offset of center of gravity from middle of trailer (in) 0 0 0 0 19.7 0 Total length of vehicle (ft) 69 69 69 69 69 72 Total length of trailer to back axle (ft) 44.2 44.2 44.2 44.2 44.2 24.1 Distance from front axle to center of tandem of tractor (ft) 18.5 18.5 18.5 18.5 18.5 11.5 1 Weight of vehicle model verified by summing the tire forces underneath the vehicle driving along a flat road.  A tanker trailer that is fully loaded.  A tanker trailer with the lower half filled, to simulate a partially-loaded tractor/tanker- trailer truck under normal conditions. The CG of this payload was lower than the fully- loaded tanker trailer.  A tanker trailer with the upper half filled, to simulate a top-heavy or unstable tractor/tanker-trailer truck. The CG of this payload was higher than the fully-loaded tanker trailer.  A tanker trailer with the right half filled, to simulate a tractor/tanker-trailer truck during a roadway departure correction. The CG of this payload was shifted to the right side, or passenger side, of the trailer. While the validity and accuracy of the simulation scenarios for the fully-loaded tractor/tanker- trailer truck were considered appropriate for making decisions regarding design criteria for pavement/shoulder cross-slope breaks for the partial traversal moderate departure and full traversal moderate departure vehicle trajectories, the validity and accuracy of the simulation results for the partially-loaded tractor/tanker-trailer trucks could not be verified. For this reason and because test scenarios for partially-loaded tractor/tanker-trailer trucks were simulated only for the full traversal severe departure vehicle trajectory which was considered too extreme to serve as a basis for determining design policy, simulation results for the partially-loaded tractor/ tanker-trailer trucks were not directly considered when making decisions regarding design criteria for pavement/shoulder cross-slope breaks.

2A-3 Table A-2 provides simulation results for the tractor/single-van-trailer truck for the full traversal severe departure vehicle trajectory. Table A-3 through Table A-6 provide simulation results for the four combinations of tractor/tanker-trailer trucks modeled for the full traversal severe departure vehicle trajectory.

2A-4 Tablel A-2. Tractor/Single-Van-Trailer Truck: Full traversal severe departure Superelevation 4% Cross-Slope Break 0% 4% 6% 8% Speed (mph) 30 50 60 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Y Y N N Y N N Max Neg Roll Angle -1.34 -5.45 -7.32 -7.87 -9.34 -2.23 -2.57 Max Pos Roll Angle 4.44 6.73 9.05 16.23 25.82 8.01 10.86 Superelevation 6% Cross-Slope Break 0% 4% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Y Y Y Y N N N Y N N N Max Neg Roll Angle -2.48 -6.57 -8.41 -8.85 -8.94 -10.57 -11.15 -3.34 -3.70 Max Pos Roll Angle 3.18 5.48 7.72 6.80 14.65 23.53 17.83 6.44 8.96 Superelevation 8% Cross-Slope Break 0% 4% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) Y Y Y Y Y Y Y Y Y N N Y N N N Max Neg Roll Angle -3.64 -7.7 -9.52 -9.99 -10.09 -11.53 -12.33 -4.42 -9.84 -4.81 Max Pos Roll Angle 1.91 4.27 6.41 5.51 12.93 20.86 15.99 4.92 26.73 6.87

2A-5 Tablel A-3. Fully-Loaded Tractor/Tanker-Trailer Truck: Full traversal severe departure Superelevation 4% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y Y N Y N N N N N Max Neg Roll Angle -1.82 -8.68 -7.35 Max Pos Roll Angle 7.06 24.36 22.14 Superelevation 6% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y N Y Y N N N N N N N Max Neg Roll Angle -2.86 -10.29 -7.04 -8.30 Max Pos Roll Angle 5.56 21.82 32.62 16.74 Superelevation 8% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) Y Y N Y Y N N N N N N N Max Neg Roll Angle -3.96 -11.79 -9.54 -8.74 Max Pos Roll Angle 4.12 19.62 29.31 12.99

2A-6 Tablel A-4. Partially-Loaded Tractor/Tanker-Trailer Truck (Lower Half): Full traversal severe departure Superelevation 4% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -1.7 -2.894 -3.59 -1.653 -3.383 -4.261 -1.63 -3.67 -4.567 Max Pos Roll Angle 0.715 1.252 1.483 3.666 5.522 5.726 4.927 6.983 7.941 Superelevation 6% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -3.327 -4.03 -4.722 -5.035 -2.815 -4.534 -5.387 -5.774 -2.79 -4.822 -5.705 -6.072 Max Pos Roll Angle 0.092 0.331 0.214 2.376 4.361 4.575 4.385 3.462 5.809 6.619 5.93 Superelevation 8% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -4.797 -5.176 -5.867 -6.181 -3.974 -5.684 -6.521 -6.91 -3.949 -5.949 -6.847 -7.21 Max Pos Roll Angle 1.052 3.195 3.415 3.233 2.071 4.633 5.259 4.729

2A-7 Tablel A-5. Partially-Loaded Tractor/Tanker-Trailer Truck (Upper Half): Full traversal severe departure Superelevation 4% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y N N N N N N N N Max Neg Roll Angle -3.348 Max Pos Roll Angle 7.249 Superelevation 6% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y N N N N N N N N N N N Max Neg Roll Angle -4.045 Max Pos Roll Angle 5.211 Superelevation 8% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) Y N N N N N N N N N N N Max Neg Roll Angle -4.075 Max Pos Roll Angle 2.549

2A-8 Tablel A-6. Partially-Loaded Tractor/Tanker-Trailer Truck (Right Half): Full traversal severe departure Superelevation 4% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) N N N N N N N N N Max Neg Roll Angle Max Pos Roll Angle Superelevation 6% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) N N N N N N N N N N N N Max Neg Roll Angle Max Pos Roll Angle Superelevation 8% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) N N N N N N N N N N N N Max Neg Roll Angle Max Pos Roll Angle

APPE Simul Trajec This appe of pavem conclusio shoulder  P t t t m  F t t s t w p c d Simulatio and a trac trajectori single-va B-2 prov moderate tractor/ta Table B- traversal tractor/do NDIX B. ation Re tories ndix provid ent/shoulde ns related to cross-slope. artial traver ravel lane al ires of the v he vehicle is ildest depa ull traversa ravel lane al raverse the c teered back he partial tra hen all tire roduced mo ross-slope b esign policy n results fo tor/double- es are provi n-trailer truc ides simulat departure v nker-trailer 4 provides s moderate de uble-van-tr sults fo es simulatio r cross-slope the need to The two ve sal moderat ong a path t ehicle traver steered bac rture scenar l moderate d ong a path t ross-slope b to the travel versal mod s (i.e., tires o re extreme r reak. This v on paveme r a tractor/si van-trailer tr ded in this a k for the pa ion results f ehicle trajec truck for the imulation re parture veh ailer truck fo r Mode n results for breaks and change AA hicle traject e departure: angent to th se the cross k to the trav io simulated eparture: Th angent to th reak and en lane. This v erate departu n both side esponses th ehicle trajec nt/shoulder ngle-van-tra uck for part ppendix. Ta rtial travers or the tracto tory. Table partial trav sults for the icle trajector r the full tra 2B-1 rate De the two veh which serv SHTO’s cu ories are des The vehicle e roadway c -slope break el lane. Thi as part of th e vehicle gr e roadway c croach onto ehicle traje re trajectory s of the vehi an when onl tory served cross-slope iler truck, a ial and full t ble B-1 prov al moderate r/single-van B-3 provide ersal moder fully-loade y. Table B- versal mod parture icle trajecto ed as the pri rrent design cribed as fo gradually d urvature and and encroa s vehicle tra is research. adually drif urvature and the shoulde ctory was so as Glenno cle) traverse y the passen as the basis break. fully-loaded raversal mo ides simula departure v -trailer truck s simulation ate departur d tractor/tan 5 provides s erate depart Vehicle ries selecte mary source policy for p llows: rifts from th only the pa ch onto the jectory repr ts from the all tires of r before the mewhat mo n et al. (198 the cross-s ger-side tir for AASHT tractor/tan derate depa tion results ehicle trajec for the full results for e vehicle tra ker-trailer tr imulation re ure vehicle t d for evalua s for drawin avement/ e middle of ssenger-sid shoulder bef esented the middle of th the vehicle vehicle is re severe th 1) reported t lope break i es traverse t O’s current ker-trailer tr rture vehicle for the tract tory, and Ta traversal the fully-loa jectory, and uck for the sults for the rajectory. tion g the e ore e an hat t he uck, or/ ble ded full

2B-2 Tractor/Single-Van-Trailer Truck Table B-1. Tractor/Single-Van-Trailer Truck: Partial traversal moderate departure Superelevation 4% Cross-Slope Break 0% 6% 8% 10% Speed (mph) 30 50 60 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -0.32 -0.97 -1.18 -0.32 -0.97 -1.19 -0.31 -0.97 -1.27 -0.29 -0.99 -1.36 Max Pos Roll Angle 0.70 0.03 1.95 2.45 2.25 2.47 3.43 3.27 3.00 4.46 4.34 Superelevation 6% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -1.5 -2.16 -2.38 -2.70 -1.50 -2.16 -2.43 -2.82 -1.15 -2.16 -2.52 -2.89 Max Pos Roll Angle 0.71 1.29 1.17 0.64 1.21 2.26 2.20 1.63 Superelevation 8% Cross-Slope Break 0% 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -2.67 -3.36 -3.59 -3.96 -2.67 -3.37 -3.71 -4.12 -2.65 -3.37 -3.80 -4.20 Max Pos Roll Angle 0.14 0.04 1.08 1.04 0.53

2B-3 Table B-2. Tractor/Single-Van-Trailer Truck: Full traversal moderate departure Superelevation 4% Cross-Slope Break 0% 6% 8% 10% Speed (mph) 30 50 60 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -0.47 -1.13 -1.34 -0.47 -1.36 -1.64 -0.49 -1.50 -1.78 -0.56 -1.63 -1.94 Max Pos Roll Angle 0.85 0.09 5.56 4.77 4.54 7.12 6.43 6.26 8.52 8.10 7.93 Superelevation 6% Cross-Slope Break 0% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -1.50 -2.16 -2.38 -2.70 -1.50 -2.16 -2.43 -2.82 -1.15 -2.16 -2.52 -2.89 -3.46 Max Pos Roll Angle 0.71 1.29 1.17 0.64 1.21 2.26 2.20 1.63 6.33 Superelevation 8% Cross-Slope Break 0% 6% 8% 10% Speed (mph) 30 50 60 70 30 50 60 70 30 50 60 70 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 214 758 1200 1810 Recover (Y/N) Y Y Y Y Y Y Y Y Y Y Y Y Y Y Max Neg Roll Angle -2.87 -3.59 -3.97 -5.04 -2.86 -3.98 -4.34 -5.45 -2.90 -4.13 -4.49 -5.63 -3.18 -5.83 Max Pos Roll Angle 3.08 2.67 2.50 2.70 4.49 4.35 4.23 4.43 6.08 6.09

2B-4 Tractor/Tanker-Trailer Truck Table B-3. Fully-Loaded Tractor/Tanker-Trailer Truck: Partial traversal moderate departure Superelevation 4% Cross-Slope Break 6% 8% Speed (mph) 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Max Neg Roll Angle -0.77 -1.17 -0.86 -1.28 Max Pos Roll Angle 3.02 2.81 4.06 3.90 Superelevation 6% Cross-Slope Break 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Max Neg Roll Angle -2.00 -2.44 -2.11 -2.56 Max Pos Roll Angle 1.88 1.76 2.91 2.85 Table B-4. Fully-Loaded Tractor/Tanker-Trailer Truck: Full traversal moderate departure Superelevation 4% Cross-Slope Break 6% 8% 10% Speed (mph) 30 50 60 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Y Y Max Neg Roll Angle -1.42 -1.75 -0.621 -1.61 -1.94 -0.74 Max Pos Roll Angle 5.55 5.32 7.97 7.31 7.13 9.76 Superelevation 6% Cross-Slope Break 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Y Y Max Neg Roll Angle -2.77 -3.16 -1.74 -2.98 -3.35 -3.54 Max Pos Roll Angle 4.52 4.38 6.61 6.28 6.16 5.49

2B-5 Tractor/Double-Van-Trailer Truck Table B-5. Tractor/Double-Van-Trailer Truck: Full traversal moderate departure Superelevation 4% Cross-Slope Break 8% 10% Speed (mph) 30 50 60 30 50 60 Curve Radius (ft) 250 926 1500 250 926 1500 Recover (Y/N) Y Y Y Y Y Max Neg Roll Angle -2.94 -3.28 -2.65 -3.39 -3.65 Max Pos Roll Angle 7.63 7.56 9.39 9.91 9.67 Superelevation 6% Cross-Slope Break 6% 8% Speed (mph) 30 50 60 70 30 50 60 70 Curve Radius (ft) 231 833 1330 2040 231 833 1330 2040 Recover (Y/N) Y Y Y Y Max Neg Roll Angle -4.04 -4.46 -4.38 -4.77 Max Pos Roll Angle 4.88 4.89 6.81 6.80 Superelevation 8% Cross-Slope Break 6% Speed (mph) 30 50 60 70 Curve Radius (ft) 214 758 1200 1810 Recover (Y/N) Y Y Max Neg Roll Angle 0 0 Max Pos Roll Angle 3.98 4.03