BW8055 Us 000

Evaluating the Effectiveness of Electronic Stability Systems in
Reducing Truck Rollovers
Kelly M. Donoughe
Thesis submitted to the faculty of the Virginia

Polytechnic Institute and State University in partial
fulfillment of the requirements for the degree of
Master of Science
In
Civil Engineering
Hesham A. Rakha, Chair

Kathleen L. Hancock
Sangjun Park
December 3, 2010
Blacksburg, VA
Keywords: hardware-in-the-loop, electronic stability control,

truck rollover, TruckSim
© 2010 by Kelly Donoughe

Evaluating the Effectiveness of Electronic Stability Systems in
Reducing Truck Rollovers
Kelly M. Donoughe
Abstract
The objective of this research is to develop a customized hardware-in-the-loop system

that is used to test Electronic Stability Program (ESP) systems to prevent heavy truck rollovers
when navigating horizontal roadway curves. While most of the published literature on electronic
stability control focuses on the effectiveness of stability systems in passenger cars, very few
researchers have considered its application as it pertains to commercial vehicles. Detailed crash
data that have been extracted from the crashes that are represented in the Large Truck Crash
Causation Study database have been used to draw conclusions regarding the main cause of the
crashes and the geometry of the road upon which the crashes occurred. Those crash scenarios
were run through a hardware-in-the-loop system that communicates between the TruckSim
software, a vehicle dynamics based simulation program, and a real-time tractor-trailer braking
rig. The simulations were first run without the ESP enabled to determine the critical speed
which will cause the truck to roll, then the same simulation runs were executed with the Bendix
stability system enabled to determine the difference in speeds in which a rollover is inevitable
with and without the technology. A third speed that represents the lowest speed in which the
stability system activates was also determined. As requested by the National Highway Traffic
Safety Administration (NHTSA), this study also serves as a comparison between the Bendix
system and the Meritor WABCO system which has already been tested by the University of
Michigan Transportation Research Institute.
Acknowledgements
I would like to extend a sincere thank you to my advisor and committee chairman, Dr. Hesham
Rakha. His support and guidance throughout this process have been unmatched throughout my
academic experience. I would also like to thank my other committee members, Dr. Kitty
Hancock and Dr. Sangjun Park, for sharing their expertise in the field which has added to the
overall quality of this thesis.
I would like to extend huge thanks my family and friends for their support. To my mom and dad
who have supported me whole-heartedly in all of my adventures from the day I was born, I could
not have made it through this program without you two. To my sister, Jennifer, who has
provided continuous encouragement throughout this process, thank you. And to all of my
colleagues from the Transportation Infrastructure Systems Engineering program, thanks for your
friendship and good luck in all you do.
Finally, I would like to acknowledge and thank God for giving me this dream and providing me
with the skill sets to accomplish it and the family and friends to help me through it.
iii
Attributions
In addition to all of the guidance and editing help from my committee, I would like to
acknowledge the attributions of the other researchers that have helped me with various portions
of this research.
I am very appreciative of the help from Alrik Svenson from National Highway Traffic Safety
Administration (NHTSA) and Paul Grygier, Gary Heydinger, and Kamel Salaani from the
Vehicle Research and Test Center (VRTC) for providing the field data results that were used to
validate our hardware-in-the-loop setup. Their patience, support, and encouragement was highly
valued throughout the duration of this project.
I am grateful for the help of everyone that had a part in building and debugging the rollover
simulator. Many thanks go to Jared Bryson, Chris Gili, Wayne Swanson, and Stacy Payne all
from the Virginia Tech Transportation Institute. Additional thanks go to Wayne Lewis from
Blue Ridge Testing and Brendan Chan from Bendix.
iv
Table of Contents
Acknowledgements ........................................................................................................................ iii
Attributions .................................................................................................................................... iv
List of Figures ............................................................................................................................... vii
List of Tables ................................................................................................................................. ix
Chapter 1: Introduction ............................................................................................................... 1
1.1 Overview of Truck Rollovers and Rollover Mitigation ................................................... 1
1.2 Thesis Contributions ........................................................................................................ 2
1.3 Thesis Layout ................................................................................................................... 2
Chapter 2: Literature Review...................................................................................................... 3
2.1 Progression of Safety Technology Testing ...................................................................... 3
2.2 Basic Stability Control Mechanisms ................................................................................ 4
2.3 Overview of Rollover Characteristics .............................................................................. 4
2.4 Previous Roll Stability Control Research......................................................................... 6
2.5 Crash Databases ............................................................................................................... 7
2.6 Literature Review of Benefit Estimation ......................................................................... 8
Chapter 3: Study Design ........................................................................................................... 13
3.1 Detailed Summary of the UMTRI Study ....................................................................... 13
3.1.1 Scope of the Problem .............................................................................................. 13
3.1.2 Identification of Target Crash Types ...................................................................... 13
3.1.3 Linking the Crash Data with the HiL System ......................................................... 15
3.2 Evaluating Rollovers with the HiL Simulator ................................................................ 16
3.2.1 Truck Modeling ...................................................................................................... 16
3.2.2 Driver Model ........................................................................................................... 17
3.2.3 Testing Procedures .................................................................................................. 17
Chapter 4: Development of a Hardware-in-the-Loop Testbed for Evaluating Truck Safety
Systems 18
Abstract ......................................................................................................................................... 19
Introduction ................................................................................................................................... 19
Previous research ...................................................................................................................... 19
Hardware-in-the-Loop System...................................................................................................... 20
v
Hardware ................................................................................................................................... 21
Simulation Software.................................................................................................................. 22
HiL Computing Platform .......................................................................................................... 24
HiL Client ............................................................................................................................. 25
HiL Server ............................................................................................................................. 25
HiL Simulator ....................................................................................................................... 27
System Validation ......................................................................................................................... 28
Summary Findings and Conclusions ............................................................................................ 32
Acknowledgements ....................................................................................................................... 32
References ..................................................................................................................................... 33
Chapter 5: Further Validation of HiL System .......................................................................... 38
Chapter 6: Simulation Results .................................................................................................. 44
Chapter 7: Conclusions ............................................................................................................. 48
7.1 Summary ........................................................................................................................ 48
7.2 Future Research Activities ............................................................................................. 49
Thesis References ......................................................................................................................... 50
Appendix A: Formulas for the Calculation of the Curve Coordinates in Matlab ......................... 52
Appendix B: Plots of HiL Brake Pressures for Validation ........................................................... 54
Appendix C: Copyright Documentation ....................................................................................... 59
vi
List of Figures
Figure 1: MUTCD signs W1-13 (top) and W13-1P (bottom) ........................................................ 1
Figure 2: Free body diagram of the back of a semi-trailer during a steady turn............................. 5
Figure 3: Truck Configuration Diagram [12] ............................................................................... 17
Figure 4: Schematic diagram of the Hardware-in-the-Loop system. ............................................ 21
Figure 5: Front view of full braking system incorporated in the HiL system. The Bendix
Electronic Stability Program is outlined in the white box. The gray computer on the left is the
server and the off white computer on the right is the real-time unit. Below the computers are the
brake chambers for both the tractor and the trailer. ...................................................................... 22
Figure 6: Diagram to show the three main components of TruckSim: Truck parameters, a
modeling system (mathematical modeling pictured here), and Animation and/or plots of the
dynamic characteristics of the truck. ............................................................................................ 23
Figure 7: Screenshot of the LabVIEW graphical user interface. .................................................. 26
Figure 8: 30.5 mph (49.1 kph) – Plots of velocity and associated error for the rollover scenario 34
Figure 9: 30.5 mph (49.1 kph) – Plots of roll angle and associated error for the no wheel lift
scenario. ........................................................................................................................................ 35
Figure 10: 34.5 mph (55.5 kph) – Plots of velocity and associated error for the rollover scenario.
....................................................................................................................................................... 36
Figure 11: 34.5 mph (55.5 kph) – Plots of roll angle and associated error for the rollover
scenario. ........................................................................................................................................ 37
Figure 12: Plots of brake pressure over time for a truck with a center of gravity located at 2.3
meters that surpasses six degrees of roll when traveling around a curve too fast. ....................... 39
Figure 13: NHTSA and VRTC’s field test data with no wheel lift at 58 kph............................... 40
Figure 14: NHTSA and VRTC’s field test data with wheel lift at 60 kph.................................... 41
Figure 15: Simulation test data with no wheel lift at 42 kph. ....................................................... 42
Figure 16: Simulation test data with wheel lift at 44 kph. ............................................................ 43
Figure 17: Plot of the lowest critical speeds that caused each of the following events for the truck
model with a center of gravity located at 2.0 meters: initial ESP activation, at least six degrees of
roll, and complete physical rollover.............................................................................................. 45
Figure 18: Plot of the lowest critical speeds that caused each of the following events for the truck
model with a center of gravity located at 2.3 meters: initial ESP activation, at least six degrees of
roll, and complete physical rollover.............................................................................................. 46
meters that does not experience any wheel lift when traveling around a curve too fast............... 54
vii
meters that experiences single wheel lift while traveling around a curve too fast. ...................... 55
meters that surpasses six degrees of roll when traveling around a curve too fast. (Previously
shown as Figure 12) ...................................................................................................................... 56
meters that experiences multiple wheel lift when traveling around a curve too fast. ................... 57
meters that rolls over when traveling around a curve too fast. ..................................................... 58
viii
List of Tables
Table 1: Total crashes, deaths, and injuries that could be prevented if RSC and/or ESC were
implemented[12] ............................................................................................................................. 7
Table 2: Ferguson’s article entitled "The Effectiveness of Electronic Stability Control in
Reducing Real-World Crashes: A Literature Review"[21] .......................................................... 10
Table 3: Simulation results with and without the electronic stability system where LOC is a loss
of control event and RD is a road departure event [22] ................................................................ 10
Table 4: Number of Crashes Prevented based on Houser’s research [8] ...................................... 12
Table 5: Hardware computing requirements................................................................................. 24
Table 6: Comparison of VRTC and VTTI values for validation purposes. .................................. 29
Table 7: Table describing the absolute error in (a) velocity (km/h) and (b) roll angle (degrees)
between the real-time and non-real-time simulation results ......................................................... 30
Table 8: Table describing the relative error in (a) velocity (%) and (b) roll angle (%) between the
real-time and non-real-time results ............................................................................................... 31
Table 9: Critical speeds (kph) for the truck model with a center of gravity at 2.0 meters above
ground level .................................................................................................................................. 45
Table 10: Critical speeds (kph) for the truck model with a center of gravity at 2.3 meters above
ground level .................................................................................................................................. 46
Table 11: Critical speeds (kph) for when the truck in UMTRI’s study achieved greater than six
degrees of roll with Meritor-WABCO system [12] ...................................................................... 47
Table 12: Comparison of UMTRI’s findings from their study on the Meritor-WABCO system
(speeds listed in kph) .................................................................................................................... 47
ix
Chapter 1: Introduction
According to the United States Bureau of Transportation Statistics (BTS), the truck
transportation industry accounted for $127.6 billion dollars in 2007, comprising 0.9% of the total
U.S. Gross Domestic Product (GDP), and nearly one-third of the GDP attributed to
transportation activities [1]. As such, commercial motor vehicles, including straight trucks and
combination trucks, are an integral part of freight transportation within the United States.
Straight trucks are primarily used for short range operations such as pickup and delivery
services, and they typically serve within a 50-100 mile radius of their base location.
Combination vehicles, in contrast, are typically used for regional and long-distance applications.
While only 30% of commercial motor vehicles are long-range combination vehicles, they
account for 65% of the commercial vehicle miles traveled. In addition to the high mileage
accumulated by these long-range vehicles, they exhibit the highest lifespan crash-cost per
vehicle, identifying them as a prime target for transportation safety research. A particularly
dangerous type of accident that affects combination trucks is a rollover event, often occurring
when a driver misjudges the vehicle speed when entering complex curves associated with
highway interchanges. The purpose of this research is to investigate the circumstances leading to
rollover events, and the potential effect that stability control could have on reducing annual crash
rates [2].
1.1 Overview of Truck Rollovers and Rollover Mitigation

Though the trucking industry represents 4% of registered vehicles and 7% of the total
vehicle miles traveled, the professional truck drivers are disproportionally involved in for 8% of
fatal crashes and 4% of personal damage only crashes [3]. According to the Trucks Involved in
Fatal Accidents database for the years 1999 to 2005, 13% of the fatal
crashes involving heavy trucks were caused by rollovers. Research done
by Wang and Council determined that there are approximately 4,400 to
5,000 truck rollovers on interchange ramps every year, with an associated
cost of $405 to $460 million annually [4].
Most of the research regarding rollover prevention and mitigation

began in the early 1990’s. The most frequently used method of mitigation
involves signs that warn truck drivers as they enter a potentially
dangerous curve. The optional sign, W1-13 should be accompanied by
sign W13-1P in order to provide an advisory speed limit as seen in Error!
Reference source not found.. Although these signs are present at most
of the locations that have a high rollover frequency, it is thought that
Figure 1: MUTCD drivers become desensitized to these signs which then reduces the
signs W1-13 (top) and effectiveness of the warning [5]. Rollover warning devices can provide an
W13-1P (bottom)
additional or alternate technology to reduce rollovers around curves. The
1
rollover warning device technology utilizes sensors implanted in the pavement to measure the
truck’s speed, weight, load, height, and vehicle configuration as it traverses a section of the road
[5]. If a truck is suspected of experiencing a potential rollover situation, a sign flashes or lights
up to alert the offending vehicle, encouraging them to lower their speed to prevent a rollover.
One study found that this type of dynamic sign produced a reduction in the number of rollovers
in the Washington D.C. area over a three-year period beginning in 1993 [6]. Another alternate
technology for rollover prevention, that of a rollover training device, which resides within the
vehicle itself. This system monitors the rollover threshold of the truck and displays a diagram
that shows how close the driver is getting to the condition where the truck may rollover [7].
Lastly, the most recently developed technologies for rollover prevention are known as electronic
stability systems. These systems are multipurpose by monitoring the truck’s motion, and to
actively engaging brakes to prevent rolling over or jack-knifing. While a number of researchers
have studied the benefits of stability control system applications in passenger cars and sports
utility vehicles, their application to commercial motor vehicles has been largely overlooked until
recently. By implementing stability control technology throughout the entire US fleet of trucks,
a significant reduction in commercial vehicle accidents can be achieved, thus saving time, and
money, and most importantly, lives.
1.2 Thesis Contributions

In order to assess the potential for electronic stability systems to decrease the number of
truck rollover crashes and increase safety, the primary objectives of this thesis are twofold:
- Build a hardware-in-the-loop testbed in order to evaluate electronic stability systems

in application with tractor-trailer
- Quantitatively evaluate the effectiveness of an electronic stability system to reduce
speed and prevent rollovers.
1.3 Thesis Layout

This thesis includes:
- a literature review of relevant research;

- summary of the study’s design;
- a detailed description of the hardware-in-the-loop setup;
- a description of the TruckSim program, the truck used in the simulations, and the
method used to create the simulation scenarios;
- the procedures used to validate the setup;
- a summary of the findings from the simulation runs;
- conclusions of the research and possibilities for the expansion of this research topic.
2
Chapter 2: Literature Review
This literature review provides an overview of the history of safety technology testing in
vehicles; a basic description of how stability systems analyze driving behavior and how they
determine when to engage; a brief overview of rollover characteristics; a summary of previous
roll stability control research and testing; an synopsis of the crash databases that were utilized;
and a review of the various methods that have been used to calculate the effectiveness of safety
technology.
2.1 Progression of Safety Technology Testing

In the early days of safety technology testing, vehicle designs and new technologies were
tested on test tracks with controlled driving and environmental conditions. While test tracks
provided insight into the vehicle’s dynamics, the tests require professional drivers and specially-
equipped trucks which lead to expensive and potentially dangerous situations.
Similar to test track testing, field operational tests (FOT) use fleets of vehicles to collect
driving data; however, FOTs collect the data as normal drivers are going along with their
everyday business. New systems can be installed on specific vehicles so researchers can monitor
the vehicle’s performance to evaluate the effectiveness of the technology [8]. This method has
the benefit of capturing real-life driving situations; however, it does not evaluate the limits of the
technology and it requires significant time and resources to complete such a study. Even within
a long study period, the technology might not be pushed to the extreme and therefore cannot be
evaluated in its entirety.
The next step in the evolution of dynamic vehicle testing involves modeling vehicles in a
simulation software environment. These models are built to match the measured characteristics
from test track experiments [9]. Safety technology models are then established and
superimposed onto the vehicle models to evaluate their effectiveness. Effective modeling of the
safety technologies is often difficult because the crash detection algorithms are typically
proprietary information. Consequently, this method is useful for initial testing procedures,
however, it is unable to provide detailed information on the idiosyncrasies of the technology
since it is based on a model of idealistic performance.
The current state of the art in the development of testing new technologies is to
physically integrate the technology with a simulation software by establishing a hardware-in-the-
loop (HiL) system. These systems utilize vehicle characteristics provided by the simulation
software and apply them to a physical mechanical setup of the remaining vehicle characteristics.
This enables the system to be tested in a safe environment without the need for a professional
driver or expensive adaptations to the trucks. In 1999, Lee and Suh utilized a HiL system to test
an antilock brake system and a traction control system [10]. Since then, a few other systems
were built to develop and test systems for passenger vehicles [11] and tractor-trailers [12].
3
2.2 Basic Stability Control Mechanisms
In general, electronic stability systems monitor a vehicle’s wheel speed, steering wheel
angle, lateral acceleration, and yaw rate. The system then compares the driver’s steering input to
the actual motion of the vehicle. If the sensors detect any movements that the driver does not
intend, such as one tire spinning faster than the rest, the stability control system will intervene to
reduce the risk of loss of control and rollover. The early versions of the active roll stability
control systems for tractor-trailers only reduced the torque on the engine, commonly referred to
as the “Jake Brake”; however, newer technology increases the effectiveness by also using the
mechanical brakes to slow the vehicle in a quicker manner. Applying the brakes to individual
tires creates an opposite moment that will work against the rollover forces to keep the vehicle on
its intended path.
Since the stability systems respond to eminent loss of control events, it is pertinent to
discuss the defining characteristics of loss of control. There are two main components that are
important for the system to monitor: lateral tire slip and yaw rate. Judging the driver’s control of
the vehicle solely by the lateral slip measurement would be insufficient because vehicles utilize
this lateral motion to create the directional change that makes a vehicle turn. Yaw rate, another
parameter, measures of the vehicle’s angular velocity about its vertical axis [13]. Each brand of
technology uses different thresholds to determine what classifies as a loss of control event, but
the systems all respond in similar ways. If a driver begins to lose control of his/her vehicle, the
system will register the lateral tire slip and yaw rate as being over the limit and will take control
by braking individual wheels [13]. The vehicle will brake the outer front wheel when the driver
experiences an oversteering event and will brake the inner rear wheel to counteract for
understeering.
2.3 Overview of Rollover Characteristics

Many researchers have noted that it is very difficult for a driver to accurately distinguish
when they are about to rollover versus when they are safely traversing a curve or other maneuver
[7, 14]. Even a small increase in roll stability can reduce the number of rollover crashes [14].
4
Figure 2:: Free body diagram of the back of a semi
semi-trailer
trailer during a steady turn
In a single-plane
plane rollover scenari
scenario, as depicted in Figure 2,, the destabilizing and
stabilizing rolling moments must be equal in order for the truck to remain upright. The
destabilizing moments are caused by the lateral acceleration at the center of gravity and the t
lateral offset of the center of gravity from the center of the truck’s track. This equilibrium is
represented in the following equation:
Equation 1

∗ ∗
∗
∗ ∆
2
Where, W is the weight of the truck
truck,
h is the height of the truck

truck,
ay is the lateral acceleration

acceleration,
Fi is the vertical force on the tires

tires,
T is the track width of the truck, and
Δyy is the lateral motion of the cent

center
er of gravity with respect to the center of the truck’s
trackwidth.
5
The premise of roll stability systems is to counteract the rollover tendency by creating a
stabilizing moment that leans the vehicle into the turn to help the center of sprung mass to stay
on the inside of the vehicle centerline [14]. When a vehicle rolls over, the load goes from being
equally distributed among the two sides of the truck to being solely supported on one side.
A tractor-semitrailer with a torsionally rigid frame, as described later in this report, has
nine degrees of freedom. The tractor portion of the truck accounts for five degrees of freedom:
yaw, side-slip, sprung mass, roll angle, steer angle roll angle, and drive axle roll angle. The
trailer accounts for two: roll angle of the sprung section and roll angle of the trailer axle group.
The last two degrees of freedom are from the articulation angle between the tractor and the
trailer[14].
2.4 Previous Roll Stability Control Research

In 2006, the Federal Motor Carrier Safety Administration published a report on the test
results from a Roll Advisor and Controller (RA&C) stability system. The purpose of this
research was to determine the percent of rollovers that could be prevented if the system were to
be nationally implemented. The research used recorded data from a previous field operational
test (FOT) sponsored by the U.S. Department of Transportation. The FOT, titled the “Intelligent
Vehicle Initiative,” tested the two components of the RA&C system that assists in preventing
rollover crashes: the roll stability advisor and the roll stability control. While none of the test
trucks rolled during the year-long data collection which spanned over 4 million miles, the data
involved 137 “critical conflicts” where the lateral acceleration was measured to be approaching
the static rollover threshold [8].
The RA&C system not only reduced engine torque, but it also engaged the engine
retarder and applied pressure to the drive axle and trailer brakes. The roll stability controller
activated when the trailer began to roll more than five degrees thus preventing the trailer from
further increasing its roll angle. A trailer that was not equipped with the system could
experience a very dangerous roll angle of up to fifteen degrees in the same situation. The
research proved that the roll stability control system always intervened at speeds lower than what
would cause an unequipped vehicle to roll [8]. Using the Vehicle Dynamics Analysis, Non-
Linear (VDANL) software Version 6.0 to calculate the prevention ratio, the researchers
determined that the roll stability controller yielded a 53% reduction of rollovers from vehicles
that were traveling at excessive speeds in a curve and a 69% reduction for those using the
combined advisor and controller system for vehicles traveling at excessive speeds in a curve. It
was noted that these intelligent vehicle stability systems could also assist in preventing run-off-
road crashes, which were not included in the original prevention calculations, because they are
typically caused by the same type of maneuvers that cause rollovers, such as going too fast
through a curve; however, those were not incorporated within the scope of the project.
In 2009, the University of Michigan’s Transportation Research Institute (UMTRI)

performed a study on the efficacy of Meritor WABCO’s electronic stability control systems
6
(ESC) and roll stability control (RSC) systems for heavy trucks and tractor-semitrailers. UMTRI
used an HiL system to simulate potential rollover events instead of collecting field data. The
HiL data was augmented by information from detailed crash reports, input from an expert panel,
field operation test results, and fleet crash data [12]. A brief overview of UMTRI’s methods is
presented in this section; more detailed information will be presented later.
UMTRI’s team of experts identified crashes that might have been prevented if the
stability systems were installed in the trucks. The primary databases used were the nationally
representative sampling of truck crashes that are recorded in the Large Truck Crash Causation
Study (LTCCS), Trucks Involved in Fatal Accidents (TIFA), and the General Estimates System
(GES). Due to the high quality of the reports and its attention to detail, the LTCCS database was
the main source for determining which cases would be used as simulation input, evaluated by the
expert panel, or both. The other databases were used to determine the average number of annual
truck crashes in order to establish the overall safety benefit that could be attributed to the
stability system. UMTRI concluded that a total of approximately 3,500 crashes could be
prevented by the Meritor WABCO roll stability control system and approximately 4,600 crashes
could be prevented based on the implementation of the electronic stability system.
Table 1: Total crashes, deaths, and injuries that could

be prevented if RSC and/or ESC were implemented[12]
Annual Prevented by:

Crash Rate RSC ESC
Crashes 11,224 3,489 4,659
Deaths 255 106 126
Injuries 14,233 4,384 5,909
2.5 Crash Databases

The Large Truck Crash Causation Study (LTCCS), produced by the National Highway
Traffic Safety Administration in combination with the Federal Motor Carrier Safety
Administration, is a subset of the General Estimates System (GES) that aims to collect data with
the intent of being used for countermeasure development. It contains 963 nationally
representative crashes that were collected from April 2001 to December 2003 and that involved
heavy trucks with a gross vehicle weight rating (GVWR) greater than 10,000 pounds. Of its 963
cases, it includes data involving 1,123 large trucks, 959 passenger vehicles, 249 fatalities and
1,654 injuries [15, 16]. It includes all crash types and injury severities ranging from fatal to non-
incapacitating but evident injury (B-injury). Since the main goal of the LTCCS is to collect data
that would be useful in developing crash prevention systems and countermeasures, it typically
contains a detailed narrative of each crash, a scene diagram, photographs of the scene from
multiple angles, vehicle information, and driver information. This database is available to the
public at the following web address: http://www-nass.nhtsa.dot.gov/LTCCS_PUB/SEARCH
FORM.ASPX.
7
The Center for National Truck and Bus Statistics at the University of Michigan produces
a database called Trucks Involved in Fatal Accidents (TIFA). The TIFA database contains
detailed information on all trucks with a GVWR greater than 10,000 lbs (classified as medium
and heavy trucks) that were involved in a fatal crash within the United States each year. The
General Estimates System (GES) is composed of a nationally representative sample of all
police-reported crashes per year. The GES database can distinguish between trucks and cars
and also contains important details that are necessary to identify relevant rollovers and loss of
control (LOC) events. Between these two databases, an estimate of the number of trucks that
could benefit from the implementation of stability systems can be determined while considering
and drawing conclusions about various environmental and vehicle factors that are documented
in these databases.
2.6 Literature Review of Benefit Estimation

The induced exposure and direct exposure method are the two main methodological ways
used to find the effectiveness of safety technology in passenger vehicles. The induced exposure
method requires crash data to be separated into two groups: case crash types which are crashes
that are expected to be avoided by the presence of a safety system and control crash types which
are not expected to be affected. This requires the researcher to make assumptions about the
technology’s capability to reduce crashes. In the collection of research that demonstrates the
effectiveness of electronic stability control (ESC) in passenger vehicles, some researchers
assumed that only single vehicle crashes would be affected while others took many more details
into consideration. Some examples of control crashes would be rear-end crashes [17, 18] and
multi-vehicle crashes[19, 20]. Loss of control is rarely a cause of a multi-vehicle crash yet some
studies still counted some multi-vehicle crashes in their crash data if the crash was caused by a
loss of control event. Since there are many multi-vehicle crashes, the few loss of control crashes
that were included in this data set constitute a very small percentage of crashes; therefore, their
presence in the analysis is left up to each individual researcher. Once the crashes were placed
into the case or control group, the researcher completed an odds ratio which determines the
percent change by dividing the ratio of case crashes to control crashes with ESC over the
corresponding ratio of the crashes without ESC.
Equation 2
# !/# #$ !

# % !/# #$ % !
The odds ratio compares whether the probability of two events is the same. In the studies
involving ESC, the odds ratio compares whether or not the probability of a crash occurring with
an electronic stability system is the same as the probability of a crash occurring with a system.
8
When a larger database of crashes can be obtained, the direct exposure method may
prove to be a more accurate analysis since it does not require any assumptions about which crash
types will and will not be affected. The direct exposure method requires information regarding
the number of registered vehicles with and without the safety system. The first step is to
determine the number of expected crashes.
Equation 3
&' ! !%#

() "##
+%''" )$
∗ ,# %# +%'' -$
Use the calculated number of expected crashes as the denominator of the risk ratio calculation.
Equation 4
∑() "+%''" )$ %#

.
&' %#
For both methods, indirect exposure and direct exposure, find the percent change to determine
the effectiveness of the safety technology.
Equation 5
0# !#, 100 ∗ 1

In Equation 5, the ratio can be either the risk ratio or the odds ratio. The problem with these
types of statistical analyses is that there are typically multiple brands of the same type of safety
technology and they are all programmed to react at different thresholds. Singling out each
specific brand of safety system would yield such small sample sizes that it would be difficult to
find statistically significant results[21].
Regardless of which statistical analysis method the researchers used and disregarding that
the systems were not exactly the same, all of the studies came to similar conclusions that the
electronic stability systems for passenger vehicles nearly always have positive effects.
Specifically, the category of severe and fatal crashes had the highest percent reduction. When
comparing cars and sports utility vehicles (SUVs), SUVs saw the greatest reduction of crashes
due to their high center of gravity which makes them more susceptible to loss of control and
rollovers.
9
Table 2: Ferguson’s article entitled "The Effectiveness of Electronic Stability
Control in Reducing Real-World Crashes: A Literature Review"[21]
Vehicle Type Crash Type Reductions

Cars Single-vehicle 30-50%
SUV Single-vehicle 50-70%
All passenger vehicles Fatal Single-vehicle 70-90%
All passenger vehicles Multi-vehicle little to none
By applying the data from Table 2, a total of approximately 10,000 crashes could be prevented
every year. This data represents average driving and weather conditions; however, the systems
are believed to have increased benefits when roads are not dry.
Researchers at the University of Iowa analyzed the effectiveness of electronic stability

control by utilizing the National Advanced Driving Simulator (NADS) [22]. The test subjects
were taken through five scenarios on wet pavement. Three of the scenarios involved cars that
suddenly pulled out or approached the test vehicle from the left lane, right lane, and head-on.
Another one was an unforeseen wind gust that broadsided the vehicle and the last one involved
navigating a difficult decreasing radius curve. An algorithm was developed to detect loss of
control by using a combination of mathematical and subjective definitions, by recording
excessive slip rates and by documenting the vehicle’s orientation when it comes to rest. The
NADS study made sure to clarify that there is a difference between road departure and loss of
control. This study defined road departure as when a driver is still able to maintain directional
control of the vehicle even though the vehicle has departed from the pavement. Loss of control
is defined as when the vehicle leaves the roadway and comes to a final resting position in which
the vehicle is not facing in the direction it was traveling in originally. Road departure was
considered less severe than loss of control because the driver is able to maintain some control
over his vehicle in order to avoid other obstacles which a true loss of control event would not
have been able to avoid. This simulation helps to distinguish whether the stability system can
provide the benefit of maintaining directional control even when a vehicle leaves the roadway
[22].
Table 3: Simulation results with and without the electronic stability system where LOC is a loss of control
event and RD is a road departure event [22]
LOC, system ON LOC, system OFF RD, system ON RD, system OFF
Scenario (N=160) (N=40) (N=160) (N=40)
Frequency Percent Frequency Percent Frequency Percent Frequency Percent
Dec. radius curve 0 0 3 7.5 35 21.9 13 32.5
Wind gust 5 3.1 32 80 24 15 7 17.5
Left incursion 0 0 15 37.5 0 0 4 10
Obs. avoidance 0 0 13 32.5 3 1.9 0 0
Right incursion 1 0.6 13 32.5 13 8.1 7 17.5
10
Table 3 shows that the decreasing-radius curve was less effective in demonstrating the positive
effect of ESC on LOC, primarily because when ESC was enabled, vehicles skidded beyond the
road and usually came to a complete stop parallel to the road. This was considered as road
departure since the vehicle was still stable.
The estimated benefits for the implementation of RSC in tractor-trailers were derived
from the General Estimates System (GES) and the Fatality Analysis Reporting System (FARS)
from a five year period starting in 1995. The statistical results from the USDOT FOT were
applied to the overall large truck crash statistics to find the potential benefits of the improved
system. Using the 137 critical events from the FOT, Houser estimated the probability of each
scenario becoming a true rollover crash through a series of calculations. The research used a
form of the indirect exposure statistical method which gives “the overall probability of a crash
without the RSC, both given that a driving conflict has occurred.” Calculating the prevention
ratio was the first task. When the prevention ratio is less than one, it means that the technology
is effective at preventing crashes as long as a critical event has occurred.
Equation 6
03 !|
0
035 !|
Where, PR1 is the prevention ratio of the probabilities of a crash occurring with RSC to a crash
occurring without RSC given that a specific scenario occurs. Pwo is calculated based on the
conditional probability with regards to the corporate rate of crashes for going too fast around a
curve based on GES data divided by the rate at which scenarios occur. The two terms of this
equation are calculated by Equation 7.
Equation 7
Δ)9,>
03 6!7 ,9 : 1
Φ< C
σ ∗ vA,B
Where, ,9 indicates conflict number j,

03 6!7 ,9 : is the probability of a crash resulting from conflict j,
Δ)9,> is the increase in speed of conflict j that results in a rollover,
vA,B is the speed during the FOT of conflict j at the peak lateral acceleration,
Φ is the Gaussian cumulative distribution, and
σ is a scaling factor which was estimated to be 0.0010 (Battelle, 2003)
The scaling factor was calculated by ensuring that σ2x)9 2 = 0.000114 as calculated by the
035 !| in Houser’s research[8]. In other words, the squared scaling factor times the
unperturbed velocity squared must equal the probability of a rollover without the system. This
scales the variance of the perturbed velocity so that the FOT observations match the historical
11
crash data. The prevention ratios were applied to national crash statistics to yield the following
table:
Table 4: Number of Crashes Prevented based on Houser’s research [8]
Rollover (Fast Turn)

Type of Truck Number of Crashes Avoided
Number of Trucks in Crashes/Year RSA (33%) RSC (53%) RA&C (69%)
HazMat Tankers 4 1.4 2.3 3
All Tankers 46 15 24 32
Tractor Trailers 471 155 250 325
Large Trucks 787 260 417 543
The research team recognizes that this study was not an exhaustive study including all possible
rollover situations and that it only represents a small sample of possible rollover situations.
12
Chapter 3: Study Design
This study is meant to serve as a comparison to the report that was previously published
by the University of Michigan Transportation Research Institute (UMTRI)[12]. The
identification of crash types, detailed analysis of crash cases, and selection of curve radii to be
used have all been established in the UMTRI report and will be summarized in section 3.1.
To compare the efficacy of the Bendix Electronic Stability Program to the Meritor-
WABCO system, it will be tested through a hardware-in-the-loop system using the pneumatic
braking system from an actual 5-axle tractor-trailer and inputs from TruckSim, a real-time
simulation program. For this report, a rollover will be defined as when a truck trailer’s roll angle
reaches or exceeds six degrees. Physical rollover values will also be presented; however, these
results hold less merit because TruckSim’s mathematical model calculations become more
unstable as the truck approaches physical rollover. Additionally, in the field data where the
trucks are equipped with outriggers to prevent complete rollovers, the trucks can not physically
exceed six degrees of roll. So for validation purposes, a rollover will be considered at six
degrees of roll. This research focuses on quantitatively identifying the difference in speed
required to roll a truck without the ESP and to roll a truck with the ESP enabled through the
novel approach of hardware-in-the-loop simulation since more traditional methods require
historical crash data with and without the system which is not currently available.
3.1 Detailed Summary of the UMTRI Study

As mentioned in the Literature review, UMTRI performed a similar set of research tests
to test the effectiveness of the Meritor-WABCO stability system for large trucks. This section
will provide a summary of the methods that UMTRI used to create the rollover scenarios and the
results that they found from their simulations.
3.1.1 Scope of the Problem

The UMTRI report states the percentages of trucks that could be affected by the
implementation of stability systems and discusses the technology’s ability to interact with
antilock braking systems. UMTRI determined that 4.5% of the total number of crashes were
due to loss of control and 1.2% of those resulted in a subsequent rollover. These types of crashes
are likely to be reduced by implementing electronic stability control and rollover stability control
systems. An additional 2.2% of crashes were tripped rollover crashes without a loss of control
event that would not likely be affected.
3.1.2 Identification of Target Crash Types

Since this report intends to compare the results of UMTRI’s study of the Meritor-
WABCO electronic stability control and roll stability control systems, this report will
summarize the methods and procedure that UMTRI followed in determining which crash types
to include in this research endeavor.
13
Crash Databases
The two rollover definitions that were included in the database search algorithm were
untripped rollovers and rollovers as the first harmful event. Tripped rollovers and subsequent
event rollovers were not included because they are unlikely to be affected by the technology.
The loss of control cases that were included are loss of control caused by control/traction loss or
completing a collision avoidance maneuver, poor road conditions that lead to a first harmful
event, and loss of control due to a first event jackknife. Crash causes not included were loss of
control due to pre-crash instability and run-off-the-road crashes involving a single vehicle due to
an avoidance maneuver. The pre-crash instability cases were excluded because after further
evaluation, these crashes were incorrectly classified.
UMTRI used a combination of crash databases to determine which crashes should be

utilized in these studies. Between the GES and the TIFA databases, an estimate of the number
of trucks that could benefit from the implementation of stability systems can be determined
while considering and drawing conclusions about various environmental and vehicle factors that
are documented in these databases. The LTCCS database was used to test the crash selection
algorithms to ensure that it was correctly identifying valid crashes since it provides the most
detailed crash information. Selection algorithms were applied to the LTCCS database and
evaluated to determine their applicability to the study by reviewing the crash narrative, the
scene diagrams, and scene photos. If the results included crashes that were not relevant to
stability control, the algorithm was revised and evaluated again until the search resulted in
relevant crashes.
There were two specific crash databases that were not utilized by the University of
Michigan: the NASS Crashworthiness Data System file and FMCSA’s Motor Carrier
Management Information System Crash file. The Crashworthiness Data System file was not
included because it mainly focused on passenger vehicle crashes and only included information
on trucks if the truck was involved in a crash with a passenger vehicle. The Motor Carrier
Management Information System Crash file provides information on trucks and buses, but it is
primarily used as a census of crashes that contain injuries or fatalities of certain levels of
severity, rather than a detailed account of those crashes.
Selection Algorithms for Relevant Crash Types

The three databases that are being utilized, TIFA, GES, and LTCCS, contain similar
information that makes it possible to identify the same crash in all three databases. When
looking for crash types that will be affected by stability control systems, a rollover is simple to
detect; however, the cause of the rollover is not usually evident. The crash files may specify
excessive speed in a curve as the cause but after further review, an expert panel may identify the
tire skids as typical for rollovers that are not related to high speeds around a curve. To weed out
irrelevant tripped rollovers, the researchers searched for crashes where the rollover was the first
harmful event in the crash. They were also interested in crashes where there was excessive yaw
or insufficient yaw prior to the crash which could have been mitigated by a stability system.
14
The researchers in the UMTRI study developed an algorithm to identify probable crashes
in the GES database and then used the same algorithm in the LTCCS database. The researchers
then read the crash description given in LTCCS to see how adequately the algorithm performed.
The algorithm captures untripped rollovers and loss of control events. Loss of control
events are more difficult to identify because they usually occur before a crash and sometimes it
can be improperly documented or not documented at all. The GES database reports the vehicle’s
role in the crash including a specific code if the vehicle lost control. UMTRI included single
vehicle crashes that were coded crashes caused by control/traction loss, left or right roadway
departures, avoidance maneuvers, and road departures. Multiple vehicle crashes were also
included if they were coded as control/traction loss that also had a collision with another vehicle
traveling in the same or opposite direction. The crashes in GES that were coded as LOC due to
excessive speed were typically not included because the expert panel evaluation revealed that the
skid marks that were left on the scene were typical of a simple rollover and that a LOC event did
not occur separate from the rollover. Relevant jackknife crashes were included if they were
identified by the GES, LTCCS, and TIFA databases as first harmful event jackknifes. First event
jackknifes directly correspond to yaw instability and LOC.
As the UMTRI report states, their algorithm selects the crashes that will most likely
benefit from the implementation of a stability system. They do not intend to imply that those are
the exclusive crashes that will benefit. There may be some more crashes hidden within the codes
that could still benefit from the technology.
3.1.3 Linking the Crash Data with the HiL System

After removing the loss of control crashes that were analyzed separately by an expert
panel of researchers, the remaining 74 relevant rollover cases are divided up into curve radii bins
to simplify the estimations. For the first set of tests, the cases are placed into one of four bins:
1. Radius of <100m with a high center of gravity

2. Radius of <100m with a medium center of gravity
3. Radius of >100m with a high center of gravity
4. Radius of >100m with a medium center of gravity
Cases with a radius of less than 100m will be run as a radius of 68m which is the average radius
of all the radii in that bin and cases with a radius greater than 100m will be simulated with a
radius of 227m which is the average of the radii greater than 100m. Similarly, a high center of
gravity is assumed to be located at 2.3m height and a medium center of gravity is located at 2.0m
height.
In addition to the two curve radii that UMTRI used in their research, a curve radius of
150 meters will also be included in this report in order to verify the conclusions about the
behavior of the Bendix electronic stability program. The curves presented in this approach were
designed using the American Association of State Highway and Transportation
Officials (AASHTO) Greenbook to as a guide to create spiral curves that gradually increase from
15
an infinite radius curve into the intended constant radius curve. We assumed a rate of centripetal
acceleration of 1.3 ft/s3 and a design coefficient of friction of 0.4. The equations and method
used to determine the centerline coordinates of the curves are presented in Appendix A.
Evaluating Effectiveness
For each curve radius mentioned above, a critical velocity was determined which
represented the highest speed a truck could enter a curve and not experience a rollover. Entry
speeds that were greater than the critical speed lead to the rollover occurring early in the curve.
From that observation, it can be noted that if a truck rolls over at a point farther in the curve than
where the truck traveling at the critical speed rolls, then the driver’s behavior or response must
have caused the rollover. UMTRI found that for radii between 100-200m, any truck that could
successfully make it through the first 100m of the curve should be able to complete the maneuver
as long as the driver does not do anything to perturb the truck [12].
3.2 Evaluating Rollovers with the HiL Simulator

In our study and the UMTRI study, the rollover scenarios that were developed in section
3.1 of this report were modeled in a simulation software called TruckSim. These scenarios were
then run through the hardware-in-the-loop setup to determine the speed at which critical events
occurred.
3.2.1 Truck Modeling

The truck that is modeled in TruckSim for this research is based on a Volvo tractor
pulling a Fruehauf box trailer, with dimensions as shown in Figure 3. This truck was modeled
due to the field test data from NHTSA and Vehicle Research and Test Center (VRTC) that was
available to validate this model. The validation of this model is explained in the Transportation
Research Board paper in Chapter 4 and further validation procedures are explained in section
Chapter 5.
There are two box-shaped loads that are placed on the trailer to modify the truck’s center
of gravity. The front load weighs 10,464.4 kg and the rear load weighs 8,650 kg. These loads
are selected and located above the trailer axles to create realistic values for the roll inertia and
radius of gyration. In order to increase or decrease the center of gravity of the truck the loads are
simply raised or lowered without changing their weights. The braking capacity of the truck
model is consistent with S-Cam brakes that have a maximum torque of 7,500 N-m at a maximum
brake pressure of 0.8 MPa.
16
Figure 3: Truck Configuration Diagram [12]
3.2.2 Driver Model

For this report, the driver is set to steer the truck along the curve of the road in order to
keep the truck on the intended path. TruckSim’s steering algorithm optimizes the steering input
and avoids any sudden or unnecessary corrections.
3.2.3 Testing Procedures

To run the simulations, the truck is set to accelerate from a stopped position until it
reaches the desired speed for the simulation. This occurs on a straight-line section of the
designed roadway before the truck reaches the beginning of the curve. During this acceleration
period, small sinusoidal steering maneuvers are completed so that the stability control program is
able to identify and estimate the trailer load based on its own algorithm. Essentially, once the
truck reaches the desired speed, it is held constant at that speed until the truck reaches the point
in the road where the curve begins. At this point, the truck’s throttle is dropped down to zero for
the remainder of the maneuver.
In order to find the critical speeds at which the truck’s roll angle surpasses six degrees of
roll or until a physical rollover occurs in the simulation, the simulations are run at increasing
speed in 0.5 kph intervals until the critical speed is identified. This procedure was followed for
all of the validation runs and for the results presented in this report.
17
Chapter 4:
Development of a Hardware-in-the-Loop Testbed for Evaluating
Truck Safety Systems
Paper Accepted for Presentation at TRB 2011 and being considered for Publication in TRR
Kelly Donoughe
Charles E. Via, Jr. Department of Civil and Environmental Engineering
Virginia Tech Transportation Institute
3500 Transportation Research Plaza
Blacksburg, VA 24061
[email protected]
Hesham Rakha (Corresponding Author)

Charles E. Via, Jr. Department of Civil and Environmental Engineering
Phone: (540) 231-1505
Fax: (540) 231-1555
[email protected]
Wayne Swanson
[email protected]
Sangjun Park
[email protected]
Jared Bryson
[email protected]
Total word count: 4,112 (text) + 3,000 (12 tables & figures) = 7,112 words
Paper 11-1376 at the 90th Transportation Research Board Annual Meeting
18
Abstract
The research presented in the paper develops a hardware-in-the-loop system that can be used to
test in-truck safety technology. The paper describes the various hardware and software
components that are required to develop the system, how the various components are integrated,
and some initial validation tests of the system. Specifically, the paper describes how the
hardware-in-the-loop system interacts with the TruckSim simulator to produce realistic results.
The models used for the validation of the system are described, followed by a demonstration of
the system validity. The absolute and relative errors between the TruckSim simulation results
and the hardware-in-the-loop simulation are demonstrated by comparing the temporal variation
in velocity and the truck roll angle. The paper demonstrates the benefits of utilizing an accurate
hardware-in-the-loop system for the testing of future in-vehicle safety systems.
Introduction
The trucking industry statistically represents 4% of registered vehicles and 7% of the total
vehicle miles traveled; however, when it comes to accident fatalities, trucks are overrepresented
as evidenced by their involvement in 8% of fatal crashes in comparison to their involvement in
only 4% of property damage only crashes.” Additionally, the vehicle involvement rate of trucks
in injury crashes has been reduced from 45% to 33% over the past ten years from 1998 to 2008
while the corresponding involvement rate for fatal crashes has only been reduced by 0.5% over
the same ten year span [3]. It is clear that while the rate of lower severity crashes have been
reduced, the high severity crashes need to be further researched in order to improve the overall
crash ratings. Building a hardware-in-the-loop (HiL) system that is capable of testing new truck
crash avoidance technologies in a safe and realistic manner is critical in order to test and verify
the effectiveness of these technologies. Having a direct comparison of various systems will
assist in selecting the optimum system.
Previous research
Before the invention of HiL systems, vehicle designs and new technologies were tested on
controlled test tracks. While test tracks provided insight into the vehicle’s dynamics, the tests
required professional drivers and specially equipped trucks which lead to expensive and
potentially dangerous situations.
Similar to test track testing, field operational tests (FOT) use fleets of vehicles to collect real
driving data. New systems can be installed on specific trucks so researchers can monitor the
truck’s performance to evaluate the effectiveness of the technology [24]. This method has the
benefit of capturing real-life driving situations; however, it does not evaluate the limits of the
technology and it requires significant time and resources to complete such a study. Even within
a long study period, the technology might not be pushed to the extreme and therefore cannot be
evaluated in its entirety. For example, a U.S. Department of Transportation study on testing the
effectiveness of a system called the Roll Advisor and Controller (RA&C) collected field data for
19
one year and although the experiment collected over 4 million miles of data, there were only 137
“critical conflicts” where the lateral acceleration was measured to be approaching the static
rollover threshold, but no physical rollovers occurred [24].
The next step in the evolution of dynamic vehicle testing involves modeling vehicles in a
simulation software environment. These models are built to match the measured characteristics
from test track experiments [9]. Safety technology models are then established and
superimposed onto the vehicle models to evaluate their effectiveness. Effective modeling of the
safety technologies is often difficult because the crash detection algorithms are typically
proprietary information. Consequently, this method is useful for initial testing procedures,
however, it is unable to provide detailed information on the idiosyncrasies of the technology
since it is based on a model of idealistic performance.
The next step in the development of testing new technologies is to physically integrate the
technology with the simulation software by establishing a HiL system. These systems utilize
vehicle characteristics provided by the simulation software and apply them to a physical
mechanical setup of the remaining vehicle characteristics. This enables the system to be tested in
a safe environment without the need for a professional driver or expensive adaptations to the
trucks. Lee and Suh utilized a HiL system to test an antilock brake system and a traction control
system [10]. A few other systems were built to develop and test systems for passenger vehicles
[11] and tractor trailers [23]. The University of Michigan Transportation Research Institute
(UMTRI) performed a study for the National Highway Traffic Administration (NHTSA) to
evaluate the performance and the potential safety benefits of electronic stability control systems
(ESC) and roll stability control systems (RSC) for heavy trucks and tractor-semitrailers. Due to
the unavailability of detailed crash data that would compare crashes that involved trucks with
and without the system, an analysis was performed by using a combination of hardware-in-the-
loop simulations, input from an expert panel, field operation test results, and fleet crash data
[23]. The system that was built at the Virginia Tech Transportation Institute for the research
presented in this paper is able to model vehicle handling maneuvers using input information
including but not limited to: steering angles, brake application, load characteristics, truck
characteristics, and road characteristics.
Hardware-in-the-Loop System
The HiL simulator is based on a commercially available simulation engine, TruckSim, and a
pneumatic tractor trailer braking system. Software developed in National Instruments’
LabVIEW 8.6.1 environment provides an interface to the TruckSim simulation engine Dynamic
Link Library. Figure 4 shows a simplified schematic diagram of the HiL system to further
clarify how the software and hardware interact.
20
Figure 4: Schematic diagram of the Hardware
Hardware-in-the-Loop system.
Hardware
The braking rig that was used in th
this
is study is based on a stock 2006 Volvo VNL pulling a 2007
Utility MX6000 van trailer per NHTSA’s recommendation. The essential components of the
braking system were assembled in a 35”d x 38”w x 78”h frame in order to reduce the overall
space that the rig occupies. The pneumatic tubing and electrical wires were sized to the
equivalent lengths used in an actual truck and coiled up in order to preserve the airflow
characteristics of the system.
The truck is fitted with an anti-lock

lock brake system (ABS), tract
traction
ion control, and an electronic
stability system. According to Federal Motor Vehicle Safety Standard 121, NHTSA has required
since 1997 that tractors have antilock control on at least one rear axle and that for at least one of
the axles, the wheels on that axle must be individually controlled by a separate modulator. A
semitrailer must have at least one axle with antilock control, and full trailers must have it on at
least one front and one rear axle. In accordance with the NHTSA mandate, each brake chamber chamb
on the tractor is controlled by its own modulator and the trailer brakes are augmented with a
trailer ABS [25].
In addition to the required hardware, the VTTI rig is equipped with a Bendix Electronic Stability
Program (ESP). The ESP is the first of many technologies that can be incorporated into the rig.
This system utilizes the samee sensors as the ABS that evaluates wheel speeds, steering angle,
lateral acceleration, and yaw rate. The Bendix ESP communicates with the truck ECU in order
to inform the system when to engage and intervenes when there is excessive lateral acceleration
orr a large yaw rate in which the system’s algorithm identifies as a potential loss of control event.
When this happens, the system engages individual brakes in order to reduce the risk of a rollover.
21
For the current project, the Bendix ESP will
be initially disabled in order to determine
the critical speed which will guarantee a
rollover if none of the other inputs change.
Once the critical speed is determined, the
Bendix system is enabled and evaluated
based on its ability to prevent the rollover.
Simulation Software
TruckSim is a software program that can be
used to model, analyze, and simulate the
dynamic behavior of large vehicles such as
tractor trailers, buses, military vehicles, and
articulated vehicles. TruckSim contains a
series of data input screens as shown in
Figure 6 which contain both graphical and
numerical data for the characteristics of the
truck, roadway, and driver. The discussion
in this paper will focus on the elements of
TruckSim that were utilized in the study. It
does not intend to imply that the aspects
presented below represent the limitations of
the software.
Figure 5: Front view of full braking system
incorporated in the HiL system. The Bendix Electronic
Stability Program is outlined in the white box. The gray
computer on the left is the server and the off white
computer on the right is the real-time unit. Below the
computers are the brake chambers for both the tractor
and the trailer.
22
Figure 6:: Diagram to show the three main components of TruckSim: Truck parameters, a modeling system
(mathematicall modeling pictured here), and Animation and/or plots of the dynamic characteristics of the
truck.
In TruckSim, the user is able to define the characteristics of the vehicle as well as the geometric
properties of the road. The user has the option of spe
specifying
cifying vehicle characteristics such that,
when the various characteristics are combined together, they can simulate almost any type of
large vehicle. These characteristics include, but are not limited to, the following properties for
the tractor and the trailer:
• Sprung mass- weight; roll, pitch, and yaw inertia

• Aerodynamics - lateral, longitudinal, and vertical forces; roll, pitch, and yaw moments
• Tires - (assigned by axle) mixed type; dual tires and their spacing from each other; rolling
radius, rolling resistance, width, spring rate, and max allowable force
• Steering wheel torque - a plot of total kingpin moment with respect to steering-wheel
steering
torque
• Power train - engine model; torque converter model; transition model; transfer case; front
and rear differential
rential (for four wheel drive)
There is also a section to specify characteristics by axle which include suspension type, brakes,
steering system and the distance from the front axle. Anything that is not specified in the
program can either be modeled with a specifically created parameter file or run through a
hardware-in-the-loop
loop setup while utilizing the real
real-time setting.
23
Additionally, a load can be added to the trailer and assigned values for roll, pitch, and yaw
inertia. These loads are theoretical loads that are meant to represent an actual load. It is
important to note that two or more hypothetical loads can overlap in order to simulate a different
load. This is acceptable as long as the combined loads have the same static and dynamic
properties as the actual load they are meant to embody.
The roadway has its own set of programmable characteristics. The following are options that
may be adjusted depending on what type of scenario the user wants to run.
• Centerline geometry - can be input as an X-Y table of horizontal coordinates or can be

built with the segment builder
• Centerline and off-center elevation
• Friction
Once the parameters are set, the user is ready to run the mathematical model through TruckSim.
The run can be set to start and end at a user specified time or distance. After the mathematical
model has finished running, the user is able to view a video of the truck during the simulation in
the surface animator or view graphs of forces or properties of engine components in the post
processer section of the screen. The plotter reports the results at previously specified output time
steps. From these plots, determining the minimum and maximum values is simple. By default,
only one simulation run is plotted and simulated, however, the user can overlay other completed
runs in order to easily compare values.
HiL Computing Platform

The HiL computing platform consists of two central computers, the HiL Simulator real-time
system and the HiL Server management PC. End user access to the system is through a HiL
Client PC. Each of these systems is briefly described in the following sections. More detail can
be found in the final report.
Table 5: Hardware computing requirements
Mechanical License file located at: C:\Flexlm\k107302.lic

TruckSim RT Simulation 8.00 Windows XP requires USB hardware key authentication
Corporation (license renewed annually)
National License file located at: C:\Program Files\National
LabVIEW Instruments 8.60 Windows XP Instruments\Shared\License
Corporation Manager\Licenses\LabVIEW|RealTime_PKG_080
National Operating system for Real Time PC (never
LabVIEW RT Instruments 8.60 ETS RT expires)
Corporation
24
HiL Client
The HiL Client is a PC running Microsoft Windows that has both TruckSim and the HiL Client
software installed. The researcher uses the TruckSim software to generate simulation case runs,
and the HiL Client software to submit and receive these runs to and from the HiL Server. The
client computer has a network connection to the HiL Server computer via the local area network.
Figure 7 shows a screenshot of the LabVIEW graphic user interface that was designed to run and
monitor the simulations.
HiL Server
The HiL Server is a PC running Microsoft Windows that is located with the simulator hardware.
The computer has two network interfaces. The first network interface connects to the local area
network. This network interface has a static IP address so the HiL Client software can identify
(via a local configuration file) the HiL Server. This interface is used to connect to the HiL
Clients. The second network interface connects the HiL Server via Ethernet crossover cable
directly to the HiL Simulator.
The HiL Server computer runs two main applications that are associated with this project. FTP
server using Microsoft IIS services is configured. The computer also has the HiL Server
software running. The HiL Server software monitors the HiL Simulator software. Depending on
the status of the HiL Simulator, the HiL Server will perform different tasks. If the HiL Simulator
is idle, then the HiL Server will check its queue for jobs to be processed. If the queue contains a
job, the software then FTPs the files to the HiL Simulator and requests that the simulation start.
25
Figure 7:: Screenshot of the LabVIEW graphical user interface.
Once the simulation is in progress, the HiL Server Software displays the current progress
progr of the
simulation. If the real-time
time system is taking longer than it should to complete each iteration of
the simulation, the “Finished Late” light will be illuminated to inform the researcher that the
machine is no longer running in real
real-time. While a simulation is running, the researcher can
monitor the real-time
time brake pressures for each axle, treadle pressures, air tank pressures, steering
angle, yaw rate, lateral acceleration, torque, engine revolutions per minute, and speed of the
vehicle. If an error message appears in the bottom portion of the window, it can usually be
cleared by pressing the “DTC Clear” button which clears all stored errors and restarts the Bendix
stability system. The treadle control and electronic stability system can be tog
toggled
gled on or off
26
before or during a simulation by pressing the toggle buttons. The HiL system’s toggle button
makes it easy to run each simulation with the stability system disabled and later with it enabled
to determine the speed at which the stability system is successful in preventing a critical event or
rollover. When the simulate button is pressed the real-time simulation will start. The box in the
bottom left-hand side of the window allows the researcher to manually apply brake pressures on
the tractor or trailer brakes. After the simulation is complete, the HiL Server software FTPs the
results from the HiL Simulator to the HiL Server.
HiL Simulator
The HiL Simulator computer is a National Instruments PXI-8108 chassis with an embedded
controller running a real time operating system. The chassis includes multiple PXI modules:
•PXI-6251 Multi-Function Digital Acquisition (DAQ) Module
•PXI-6602 Timer/Counter Module
•PXI-6713 Analog Output Module
•PXI-8464/2 Two Port CAN Interface Module
The DAQ module reads system pressures throughout the simulator hardware. The simulator uses
sixteen pressure transducers, occupying all analog input channels on the DAQ card. The two
individual analog outputs on the DAQ card are used to control the primary and secondary brake
treadle positions. The analog output card provides an analog output to control the trailer airbag
pressure, as well as digital outputs to control the ignition line and the enable/disable line of the
stability control feature of the Bendix brake ECU. Both CAN channels provide messages to the
Bendix brake ECU: the first CAN channel interfaces to a proprietary CAN network, while the
second CAN channel is configured to emulate an engine ECU using the J1939 protocol.
The PXI controller has an executable that launches at startup. The executable has multiple
parallel timed loops. The executable immediately begins processing messages in the two
communication loops. One timed loop provides proprietary CAN communications. A second
time loop provides J1939 communications. The timer/counter timed loop immediately begins
stimulating the digital waveforms for the wheel speed sensors with a period related to the wheel
speed. The main timed loop waits until a simulation is requested. Once the request is made, the
main timed loop resets the ignition line, initializes the TruckSim simulation, steps through the
simulation at the specified time step and finally terminates the TruckSim simulation. During
each iteration the software reads the pressure transducers and provides the results to the
TruckSim simulation DLL as an input parameter. Upon terminating the TruckSim simulation,
the output files are created. Once the files are created, the main loop sets the status to simulation
complete and waits until the next simulation is requested.
27
The LabVIEW system communicates via the CAN 2.0 bus and runs a 100ms real-time loop
which can be modified to run at higher or lower frequencies. The main concern regarding the
hardware and software interactions are ensuring that all of the correct and necessary variables are
being transferred to the stability systems. Since the stability systems typically contain
proprietary algorithms to detect potential rollovers or loss of control situations, the companies
are usually reluctant to release the information that is needed to ensure proper evaluation of the
equipment.
System Validation
The VTTI TruckSim real-time simulation results were validated using vehicle test data provided
by NHTSA. The vehicles were tested at the Vehicle Research and Test Center (VRTC) in East
Liberty, Ohio and consisted of a Volvo tractor and a Fruehauf van trailer. A TruckSim model
that had already been validated with experimental data was provided [9, 26]. VTTI and VRTC
both used one truck with four different loading configurations. The different loading
configurations had trailer center of gravities located at 2.0 m, 2.2 m, 2.3 m, and 2.73 m. Each
loading configuration was placed on the truck and set to attempt a steering maneuver of 199
degrees at a rate of 175 degrees per second once it reached a specified speed. Each configuration
was simulated to determine the speed at which the following events occurred:
1. Highest speed with no wheel lifts

2. Lowest speed with one wheel lift
3. Lowest speed with multiple wheel lifts
4. Lowest speed that creates at least six degrees of roll
5. Lowest speed that causes a physical rollover.
The speeds that VRTC determined for the aforementioned events were then compared against
the speeds that were produced by the real-time system. These parameters were determined and
found to be identical as shown in Table 6.
28
Table 6: Comparison of VRTC and VTTI values for validation purposes.
Volvo with CG at 2.3

Outcome VRTC VTTI
No Wheel Lift
29.5 mph (47.5 kph) 29.5 mph (47.5 kph)
No Rollover
Single Wheel Lift - LR Trailer
30.0 mph (48.3 kph) 30.0 mph (48.3 kph)
No Rollover
30.5 mph (49.1 kph)
Achieves > 6° Roll not determined
6.4081 degree
Multiple Wheel Lift
31.0 mph (49.9 kph) 31.0 mph (49.9 kph)
No Physical Rollover
Physical Rollover 31.5 mph (50.7 kph) 31.5 mph (50.7 kph)
Plots of the vertical wheel forces, yaw rate, lateral acceleration, and degrees of roll were
generated to show the similarities of using the internal mathematical model within TruckSim and
using the real-time system that will be used in the subsequent HiL tests. Examples of the
velocity and roll angle results from the 2.2 m center of gravity truck are presented in Figure 8
through Figure 11. The first two plots show the results for a scenario that was run at an initial
speed of 49.1 km/h where no rollover or wheel lift occurred. The last two plots show the results
of a rollover scenario that occurred at an entry speed of 55.5 km/h. In all of these figures, the
solid line represents the plot of the TruckSim non-real-time mathematical model (running the
stand-alone simulation software) and the dashed line represents the results when the same
simulation was run on the real-time HiL. Similarly, the red lines represent the tractor as Unit 1
and blue lines represent the trailer as Unit 2. In Figure 8 and Figure 10, subplot (a) shows the
speed of the tractor and trailer with respect to time whereas the roll angle of the tractor and trailer
with respect to time is shown in subplot (a) of Figure 9 and Figure 11. When the vehicle
physically rolls over, the roll angle approaches 90 degrees of roll, indicating that the truck is on
its side. Subplot (c) shows the associated error with respect to time. Subplot (b) and (d) in each
of the four figures show a histogram of the error associated with subplot (a) of the same figure.
The plots indicate that the error associated with the difference between the real-time and the non-
real-time results are of small magnitude and close to zero. Specifically, the error between the
simulation runs results from the slight offset of the real-time system from the non-real-time
system.
29
Table 7: Table describing the absolute error in (a) velocity (km/h) and (b) roll angle (degrees) between the
real-time and non-real-time simulation results
(a)
Absolute Error in Velocity (km/h)
Average Minimum Maximum
Unit 1 Unit 2 Unit 1 Unit 2 Unit 1 Unit 2
2.0 No wheel-lift 0.069 0.069 0 0 0.104 0.106
Center of Gravity Rollover 0.094 0.071 0 0 0.481 0.147
2.2 No wheel-lift 0.066 0.067 0 0 0.097 0.101
2.3 No wheel-lift 0.060 0.061 0 0 0.090 0.089
2.73 No wheel-lift 0.052 0.053 0 0 0.090 0.090
(b)
Absolute Error in Roll Angle (degree)
2.0 No wheel-lift 0.000 0.001 0 0 0.215 0.223
Center of Gravity Rollover -0.692 -0.694 0 0 8.029 7.904
2.2 No wheel-lift 0.000 0.001 0 0 0.257 0.266
2.3 No wheel-lift 0.000 0.001 0 0 0.275 0.285
2.73 No wheel-lift 0.002 0.003 0 0 0.304 0.315
The absolute error is shown to be very small in magnitude for the velocity and roll angle data, as
summarized in Table 7, Figure 8 and Figure 9. Both sets of data have a minimum error of zero
which indicates that the real-time and non-real-time simulation runs are exactly the same for part
of the simulation. This usually occurs at the beginning of the simulation and the error increases
with time. The cumulative nature of the error could spawn from the tendency for the real-time
system to stop the simulation a few hundredths of a second earlier than the non-real-time system.
The velocity errors are very consistent across the different center of gravity scenarios (less than 1
km/h). After six degrees of roll, the roll angle error becomes larger for the rollover situations.
This instability happens because the truck does not include torsional stiffness in its model. In
terms of relative error, the roll angle error of a rollover scenario is small when comparing three
degrees to ninety degrees of roll.
30
Table 8: Table describing the relative error in (a) velocity (%) and (b) roll angle (%) between the real-time
and non-real-time results
(a)
Relative Error in Velocity (%)
2.0 No wheel-lift 0.151 0.153 0 0 0.225 0.232
2.2 No wheel-lift 0.152 0.153 0 0 0.229 0.239
2.3 No wheel-lift 0.166 0.150 0 0 0.517 0.239
2.73 No wheel-lift 0.166 0.150 0 0 0.517 0.240
(b)
Relative Error in Roll Angle (%)
2.0 No wheel-lift -1.383 -0.805 0 0 62.874 73.450
2.2 No wheel-lift -1.363 -2.472 0 0 68.775 25507.9
2.3 No wheel-lift -1.069 -0.506 0 0 68.662 36.526
2.73 No wheel-lift -1.065 -0.501 0 0 70.117 32.565
The relative error compares the difference between the real-time and non-real-time data by
normalizing the error with respect to the non-real-time system estimate. While the velocity
errors are on average less than 0.2%, the relative error associated with the roll angle is much
larger, as summarized in Table 8. The average error for the simulation runs are near 1.0%, but
the maximum values are approach 60-70% error. While these errors appear large, it is important
to look further into the data to find the reason why. For example, the maximum relative error for
the trailer (unit 2) of a truck with the 2.2 center of gravity is 25,507% which occurs 0.159
seconds into the simulation. At this time, the truck is driving straight and the roll angle of the
non-real-time system is 0.0000000352 degrees of roll. The difference between the real-time and
non-real-time system is 0.000008985. Dividing the difference by the non-real-time value yields
a very large percent error for a difference that is nearly zero. Though the percent errors of the
roll angle may appear to be unreasonable, when looking at the histograms in subplots (b) and (d)
of Figure 8 and Figure 10 it is clear that the errors are very close to zero.
31
Summary Findings and Conclusions
The HiL system that was developed as part of this research effort is an invaluable tool for
accurately determining the effectiveness of new and improved truck technologies. Virtually any
technology can be installed in the loop and can be activated or deactivated by the touch of a
button. The HiL system allows vehicle technologies to be tested in extreme conditions that may
not be safe on a test track without the risk of incurring actual crashes.
The validation results demonstrate that the HiL produces simulation results that are consistent
with TruckSim, the standalone software. The absolute errors remain small, and the average
percent errors are less than 0.2% different from the velocity data and average close to 1%
different for the roll angle. These results demonstrate the validity of the testbed to accurately test
and evaluate alternative in-truck safety systems.
While the current HiL system is set up to test the Bendix Electronic Stability Program (ESP),
other various technologies can easily be integrated into the hardware rig. Safety benefits such as
reduction in the number of crashes and lives lost, resources, and time saved can be calculated
using the HiL thus reducing the costs of such evaluations.
Acknowledgements
The authors acknowledge the financial support of NHTSA and the Mid-Atlantic University
Transportation Center (MAUTC) in conducting the research. Gratitude is expressed to Wayne
Lewis from Blue Ridge Testing for his expertise in LabVIEW and Brendan Chan from Bendix
for his help configuring the HiL system. The authors appreciate the help of Alrik Svenson from
NHTSA and Paul Grygier and Kamel Salaani from the Vehicle Research and Test Center
(VRTC) for providing the comparison files for the model validation.
32
References
[1] NHTSA, "Traffic Safety Facts: Large Trucks," ed. Washington, D.C., 2008.
[2] A. Houser, et al., "A Simulation Approach to Estimate the Efficacy of Meritor
WABCO’s Improved Roll Stability Control," Publication FMCSA-MCRR-06-006.
Battelle, 2006. http://www.fmcsa.dot.gov/facts-research/research-
technology/report/rollstability-control.pdf. Accessed July 7, 2009.
[3] P. Mcnaull, et al., "Validation and Enhancement of a Heavy Truck Simulation Model
with an Electronic Stability Control Model," SAE International, 2010-04-12 2010.
[4] J.-C. Lee and M.-W. Suh, "Hardware-in-the Loop simulator for ABS/TCS," in
Proceedings of the 1999 IEEE International Conference on Control Applications (CCA)
and IEEE International Symposium on Computer Aided Control System Design
(CACSD), August 22, 1999 - August 27, 1999, Kohala Coast, HI, USA, 1999, pp. 652-
657.
[5] T. Toyohira, et al., "The Validity of EPS Control System Development using HILS," SAE
International, 2010-04-12 2010.
[6] J. Woodrooffe, et al., "Safety Benefits of Stability Control Systems for Tractor-
Semitrailers," University of Michigan Transportation Research Institute (UMTRI) DOT
HS 811 205, 2009.
[7] NHTSA, "Federal Motor Safety Standards; Air Brake Systems," D. o. Transportation,
Ed., ed, 2009.
[8] S. Chandrasekharan, et al., "Development of a Roll Stability Control Model for a Tractor
Trailer Vehicle," SAE International Journal of Passenger Cars-Mechanical Systems, vol.
2, p. 670, 2009.
33
Comparison of Real-time and Non-real-time Results for the 2.2m Center of Gravity
(a) (b)
(c) (d)
Figure 8: 30.5 mph (49.1 kph) – Plots of velocity and associated error for the rollover scenario
34
(a) (b)
(c) (d)
Figure 9: 30.5 mph (49.1 kph) – Plots of roll angle and associated error for the no wheel lift scenario.
35
(a) (b)
(c) (d)
Figure 10: 34.5 mph (55.5 kph) – Plots of velocity and associated error for the rollover scenario.
36
(a) (b)
(c) (d)
Figure 11: 34.5 mph (55.5 kph) – Plots of roll angle and associated error for the rollover scenario.
37
Chapter 5: Further Validation of HiL System
The HiL system went through several iterations of internal validation before it was
deemed officially validated through VRTC and NHTSA. The first method of validation was
described in the paper written for the Transportation Research Board and will be presented at the
2011 Transportation Research Board Annual Meeting. The second method of validation was
performed to verify that the stability program was reacting in the manner expected by the
manufacturer.
This was done by evaluating the brake pressures exerted by the stability program when a
truck traverses a curve at a speed which activates the system. In the maneuver that activates the
ESP, it is expected that the ESP will apply a higher modulated brake pressure to the side of the
truck that is on the outside of the turn. For example, in the case of the 199 degree steering
maneuver that was required by VRTC and NHTSA, the truck makes a turn to the left. In order to
provide the stabilizing moment as discussed in section 2.3, the ESP must apply brakes on the
right-hand side of the truck. Plotting the brake pressure over time confirms that this is
happening.
Figure 12 presents the plots of braking pressure over time for the 2.3 center of gravity
truck as it completes the 199 degree steering maneuver that was required for the VRTC
validation. In this specific scenario, the truck was traveling at 49.1 kph (30.5 mph) which is the
speed at which the truck first achieves greater than six degrees of roll. Subfigures (a) through (e)
correspond to axles one through five. The brake pressures are the highest on the first three axles
where they achieve nearly 0.8 MPa which is the maximum braking pressure. As expected, the
brake pressures are higher on the right side of the truck shown in red in Figure 12 compared to
the left side which is shown in blue. It is interesting that the last two axles apply nearly the same
pressure on both sides of the trailer. The rest of the plots of braking pressure for the various
validation simulations are presented in Appendix B.
38
(a) Axle 1 (b) Axle 2
(c) Axle 3 (d) Axle 4
(e) Axle 5
Figure 12: Plots of brake pressure over time for a truck with a center of gravity located at 2.3 meters that
surpasses six degrees of roll when traveling around a curve too fast.
39
Another validation was run to determine whether the ESP was working properly when
compared with field tests. The plots in Figure 13: NHTSA and VRTC’s field test data with no
wheel lift at 58 kph. Figure 13 and Figure 14 were created from the data that VRTC collected
from the field tests performed on a Volvo tractor with a Fruehauf box trailer in East Liberty,
Ohio. For these tests, the ESP-equipped truck accelerated from a stop to the desired speed then
turns the steering wheel to the right at a rate of 175 degrees per second until it gets to 199
degrees. It holds that steering angle for five seconds and then it straightens out at the same rate
in which it begun the maneuver. Figure 13 and Figure 14 show some of the variables that are
pertinent to this research. The field test results of longitudinal speed, roll angle, longitudinal
acceleration and lateral acceleration will be compared with the results from the simulator in order
to validate the system.
(a) (b)
(c) (d)
Figure 13: NHTSA and VRTC’s field test data with no wheel lift at 58 kph.
40
(a) (b)
(c) (d)
Figure 14: NHTSA and VRTC’s field test data with wheel lift at 60 kph.
In these figures, the red line represents the values recorded on the tractor and the blue lines
represent the values from the trailer. The ESP begins to intervene at a time of 1.050 seconds as
indicated by the black circles on each of the plots. There is a significant drop in speed once the
ESP intervenes as shown in the steep slope of subplot (a) and also shown in the large drop in
longitudinal acceleration which represents an aggressive deceleration. The ESP consistently
intervenes when the lateral acceleration of the tractor exceeds 0.40 G’s. In the runs from VRTC,
the ESP event lasts for just under five seconds which is almost equal to the amount of time that
the vehicle is turning.
41
(a) (b)
(c) (d)
Figure 15: Simulation test data with no wheel lift at 42 kph.
In order to draw meaningful comparisons from the simulator data to the field data the differences
between the two types of tests must be acknowledged. It should be noted that the velocities from
the field tests were lower than the speeds of the trucks that were run on the simulator. Another
difference is that the loads on the trailer were not configured in exactly the same manner. This
would have an effect on the roll angle and the lateral acceleration. The lateral acceleration and
the decrease in speed once ESP activates is affected by the truck’s braking capacity. The
simulation model used a braking model that represents S-Cam brakes whereas the Volvo tractor
and Fruehauf trailer was equipped with more powerful brakes.
42
(a) (b)
(c) (d)
Figure 16: Simulation test data with wheel lift at 44 kph.
Regardless of the differences in models, the ESP still behaves in a predictable manner. The ESP
activates at around 0.40 G’s of lateral acceleration. In Figure 16, the plot shows less than 0.40
G’s of lateral acceleration; however, that could be due to the smoothing function that was used to
eliminate the noise in the data from the real-time simulation. The overall shapes of each plot
looks similar with comparable peaks and plateaus.
43
Chapter 6: Simulation Results
This section contains results of the simulations that were run on trucks with center of
gravities at 2.0 and 2.3 meters above the ground-level that were traversing curves at high speeds
which would typically cause a truck to roll. These center of gravity locations and curve radii
were chosen based on the report published by UMTRI as mentioned previously [12]. Three
curve radii were modeled: one with a 68 meter radius, one with a 150 meter radius, and one with
a 227 meter radius. The first and the last radii were selected to provide a direct comparison to
the UMTRI report whereas the 150 meter curve was chosen to be able to draw more conclusive
results of the performance of the Bendix electronic stability control system. The results will first
be presented for all three radii, then the 68 meter radius curve and the 227 meter radius curve
results will be compared with UMTRI’s results.
The tables and plots on pages 45 and 46 show the results of our iterative simulations runs.
Under the test conditions given in this report, the truck with a low center of gravity (2.0 meters)
showed that the truck can enter a curve approximately 5.5 kph faster than the same truck without
ESP. For the truck with a higher center of gravity (2.3 meters), the ESP-equipped truck can enter
a curve at about 3.5 kph faster than if it was not equipped with ESP.
44
Table 9: Critical speeds (kph) for the truck model with a center of gravity at 2.0 meters above ground level
2.0 meter Center of Gravity Truck

Lowest Speed that Engages ESP Truck Achieves > 6 deg Roll Truck Rolls Over
68 m Without ESP --- 61.5 65
Curve Radius With ESP 49 66 70
150 m Without ESP --- 88 93
Curve Radius With ESP 70 90.5 99
227 m Without ESP --- 108.5 113.5
Curve Radius With ESP 86 109.8 119

119
120 114
109 110
99
100 93
91
88
86
80
70 70
Velocity (kph)
65 66
62
60
49
Lowest Speed that Engages ESP
40 Truck Achieves > 6 deg Roll

Truck Rolls Over
20
0
Without With Without With Without With
ESP ESP ESP ESP ESP ESP
68 m Curve 150 m 227 m

Radius Curve Radius Curve Radius
Figure 17: Plot of the lowest critical speeds that caused each of the following events for the truck model with a
center of gravity located at 2.0 meters: initial ESP activation, at least six degrees of roll, and complete
physical rollover.
45
Table 10: Critical speeds (kph) for the truck model with a center of gravity at 2.3 meters above ground level

Lowest Speed that Engages ESP Truck Achieves > 6 deg Roll Truck Rolls Over
68 m Without ESP --- 57 59
Curve Radius With ESP 48.5 61 63
150 m Without ESP --- 82 83.5
Curve Radius With ESP 68.5 83.5 87
227 m Without ESP --- 99.5 102
Curve Radius With ESP 84.5 102 105

120
105
102 102
100
100
87
85
82 84 84
80
Velocity (kph)
69
61 63
59
57
60
49
Lowest Speed that Engages ESP
40 Truck Achieves > 6 deg Roll

Truck Rolls Over
20
0
Without With Without With Without With
ESP ESP ESP ESP ESP ESP
68 m Curve 150 m 227 m

Radius Curve Radius Curve Radius
Figure 18: Plot of the lowest critical speeds that caused each of the following events for the truck model with
a center of gravity located at 2.3 meters: initial ESP activation, at least six degrees of roll, and complete
physical rollover.
UMTRI only presented their results for when their truck exceeded six degrees of roll, but the
Meritor-WABCO system had two stability systems that were being tested in addition to the base
case where the truck is only equipped with anti-lock brakes.
46
Table 11: Critical speeds (kph) for when the truck in UMTRI’s study achieved
greater than six degrees of roll with Meritor-WABCO system [12]
68 m Curve 227 m Curve

Critical Speeds 2.0m C of G 2.3m C of G 2.0m C of G 2.3m C of G
(kph) ABS 62.8 57.9 109.4 101.4
RSC 70.8 66 112.7 106.2
ESC 80.5 70.8 120.7 106.2
Their results are similar to ours in showing that the trucks with the lower center of gravity
locations are more affected by the stability systems than the higher center of gravity trucks;
however, the magnitude of the difference in critical speeds from the ABS-only case to the ESC
case is greater in the UMTRI model compared to what we found.
Table 12: Comparison of UMTRI’s findings from their study

on the Meritor-WABCO system (speeds listed in kph)
68 m Curve 227 m Curve

2.0m C of G 2.3m C of G 2.0m C of G 2.3m C of G
Absolute Relative Absolute Relative Absolute Relative Absolute Relative
Error Error Error Error Error Error Error Error
Critical Speeds
Difference in
ABS 1.3 2.1% 0.9 1.6% 0.9 0.8% 1.9 1.9%

(kph)
ESC 14.5 18.0% 9.8 13.8% 10.95 9.1% 4.2 4.0%
One main difference in the simulations that UMTRI performed is the model used for the
truck brakes. They used propriety braking information from the manufacturer which was more
powerful than the default braking model that we used from TruckSim. This default braking
model is not as aggressive as the brakes that are usually installed in trucks during the twenty-first
century. Additionally, UMTRI’s procedure required the truck to maintain the target speed until
pressure is applied to the brakes by the RSC or ESC system and the throttle position would go to
zero. In discussions with VRTC and NHTSA, this latter difference was determined to be
negligible since the change in speed from the short coast down experienced by our truck was less
than 0.5 kph over the first four seconds of the curve.
47
Chapter 7: Conclusions
7.1 Summary
This research included building a valid hardware-in-the-loop (HiL) system to model and
test the Bendix Electronic Stability Program (ESP) with real-time simulations. To build a valid
HiL system, there are several key steps to complete. First, determine which components are not
able to be modeled accurately through the simulation program. In this research, it was necessary
to have a physical pneumatic braking system in order to have a correct measure of the air-line lag
time, an ESP module with full sensor input and communication, and a partial vehicle electrical
system to determine how a truck performs when the electronic stability control is utilized. Next,
a computer program should be written to communicate the information generated by the
simulation software, TruckSim, to the hardware. It is very important to be meticulous during this
step because various components of the engine system require the messages to be sent in
different formats and frequencies. If the messages are not sent correctly, the system will not
yield accurate results. Once the signals to and from the modeling software were deemed to be
correct, the next task is to ensure that the stability system is working properly. Since the
manufactures of the stability systems typically keep their detection algorithms proprietary, it can
be difficult to resolve what signals needed to be sent to the ESP and in what format these signals
should be presented. After the system has been built and validated, it is important to run a
simple maneuver each day to verify that all of the hardware components are still reacting
correctly. This maneuver can be a simple straight-line braking maneuver that verifies the time it
takes for the truck to reach the desired speed, the time it takes to decelerate to a complete stop,
and the stopping distance of the truck. Additionally, wheel speeds and brake pressures should be
checked to ensure that each of the brakes is working properly.
In building the HiL system, we discovered the criticality of manufacturer cooperation.

When the manufacturer is not willing to communicate and assist in the wiring and programming
of the HiL, the task will become much more challenging. It is also helpful to communicate with
other organizations that have experience in building and debugging an HiL system. Debugging
the system proved to be an extremely iterative process. During our building and validation
stages, we would discover one problem that needed to be resolved and once that specific problem
was fixed we would find additional errors that were masked by the initial problem. This appears
to be a common occurrence among HiL systems. Once everything seems to be working
properly, it is important to run several different types of validation runs to put the truck through
different scenarios, then to verify that the truck performed as expected. After the system is
completely validated, the data can be collected for the intended research.
In the research presented in this thesis, simulations were run on the HiL with trucks that
have a center of gravity located at 2.0 meters and 2.3 meters above the ground, and along spiral
curves that have radii of 68 meters, 150 meters, and 227 meters, respectively. The electronic
stability system is found to activate at speeds lower than those required to result in a rollover for
48
non-ESP-equipped vehicles. For the scenarios and limitations presented, the ESP is found to
provide more protection for trucks with a low center of gravity location, relative to those with a
high center of gravity. Since the research presented here uses a braking model that is less
powerful than what is typically seen on large trucks, it is anticipated that a braking model that
uses greater torque capacities will result in a higher allowable curve entrance velocity when
comparing the rollover potential of ESP-equipped vehicles in comparison to their non-ESP-
equipped counterparts. With these adjustments, the results from this study are likely to be more
comparable the results found by the University of Michigan Transportation Institute’s study on
the Meritor-WABCO electronic stability system [12].
The analysis in this thesis is limited to five-axle tractor-trailers, and does not consider
other classes of heavy vehicles such as tankers, military vehicles, or buses. Additionally, this
research does not consider the analysis or prevention of either tripped rollovers, or those
rollovers caused by a sudden driver input.
7.2 Future Research Activities

With the development of the hardware-in-the-loop testbed, other truck configurations,
brake types, and geometric configurations may be included in future research to more fully
assess the effectiveness of implementing an electronic stability system. More specifically, future
research may include additional curve radii, environmental conditions and more variation in the
center of gravity and loading configurations. There exists a need for a database containing the
distribution of curve radii within the United States. With this database, the number of
preventable rollovers could be estimated based on the correlation between curve radii and
occurrence of rollovers. These findings would be of special interest to the owners of truck fleets
who are interested in determining how many accidents, and thus how much money, these
systems would save them. Finally, further research may determine system benefits with respect
to a more complete set of crash costs which include personal injury, physical vehicle damage,
clean-up time, traffic delay and more.
49
Thesis References
[1] (2010, November 8). RITA Beureau of Transportation Statistics. Available:

http://www.bts.gov/
[2] D. Murray, et al., "Analysis of Benefits and Costs of Roll Stability Control Systems for
the Trucking Industry," FMCSA2009.
[3] NHTSA, "Traffic Safety Facts: Large Trucks," ed. Washington, D.C., 2008.
[4] J. Wang and F. M. Council, "Estimating truck-rollover crashes on ramps by using a
multistate database," Transportation Research Record, pp. 29-35, 1999.
[5] D. Baker, et al., "Effectiveness of truck rollover warning systems," 2001, pp. 134-140.
[6] A. T. Bergan, et al., "Intelligent truck rollover advisory systems," in Intelligent
Transportation Systems, October 15, 1997 - October 15, 1997, Pittsburgh, PA, United
states, 1998, pp. 140-147.
[7] C. B. Winkler and R. D. Ervin, "On-board estimation of the rollover threshold of tractor
semitrailers," Pretoria, South Africa, 1999, pp. 540-551.
[8] A. Houser, et al., "A Simulation Approach to Estimate the Efficacy of Meritor WABCO's
Improved Roll Stability Control," 2006.
[9] P. Mcnaull, et al., "Validation and Enhancement of a Heavy Truck Simulation Model
with an Electronic Stability Control Model," SAE International, 2010-04-12 2010.
[10] J.-C. Lee and M.-W. Suh, "Hardware-in-the Loop simulator for ABS/TCS," in
Proceedings of the 1999 IEEE International Conference on Control Applications (CCA)
and IEEE International Symposium on Computer Aided Control System Design
(CACSD), August 22, 1999 - August 27, 1999, Kohala Coast, HI, USA, 1999, pp. 652-
657.
[11] T. Toyohira, et al., "The Validity of EPS Control System Development using HILS," SAE
International, 2010-04-12 2010.
Semitrailers," University of Michigan Transportation Research InstituteOctober 2009.
[13] P. A. MacLennan, et al., "Vehicle rollover risk and electronic stability control systems,"
Injury Prevention, vol. 14, p. 5, 20 January 2008 2008.
[14] D. J. M. Sampson and D. Cebon, "Achievable roll stability of heavy road vehicles," 2003,
pp. 269-287.
[15] J. Hedlund, et al., "Large Truck Crash Causation Study (LTCCS) Analysis Series: Using
LTCCS Data for Statistical Analyses of Crash Risk," FMCSA-RI-05-037. Washington,
50
DC: US Department of Transportation, Federal Motor Carrier Safety
Administration2006.
[16] J. Hedlund, et al., "Large Truck Crash Causation Study Analysis Brief," FMCSA-RAA-
07-017. Washington, DC: US Department of Transportation, Federal Motor Carrier
Safety Administration2007.
[17] G. Bahouth, "Real World Crash Evaluation of Vehicle Stability Control (VSC)
Technology," in 49th Annual Proceedings, Association for the Advancement of
Automotive Medicine, Boston, 2005, pp. 19-34.
[18] C. Tingvall, et al., "The Effectiveness of ESP (Electronic Stability Programme) in
Reducing Real Live Accidents," in Proceedings of the 18th International Technical
Conference on the Enhanced Safety of Vehicles, Washington, D.C., 2003.
[19] J. Dang, "Preliminary Results Analyzing the Effectiveness of Electronic Stability Control
(ESC) Systems," National Highway Traffic Safety Administration, Washington, DC
DOT-HS-809-790, 2004.
[20] P. Green and J. Woodrooffe, "The Effectiveness of Electronic Stability Control on Motor
Vehicle Crash Prevention," University of Michigan Transportation Research
Institute2006.
[21] S. A. Ferguson, "The Effectiveness of Electronic Stability Control in Reducing Real-
World Crashes: A Literature Review," Traffic Injury Prevention, vol. 8, p. 10, 2007.
[22] G. S. Watson, et al., "Design of simulator scenarios to study effectiveness of electronic
stability control systems," in Driver Behavior, Older Drivers, Simulation, User
Information Systems, and Visualization, 2001 Wisconsin Avenue NW, Green Building,
Washington, DC 20007, United States, 2006, pp. 79-86.
Semitrailers," University of Michigan Transportation Research Institute (UMTRI) DOT
HS 811 205, October 2009.
[24] A. Houser, et al., "A Simulation Approach to Estimate the Efficacy of Meritor
WABCO’s Improved Roll Stability Control," Publication FMCSA-MCRR-06-006.
Battelle, 2006. http://www.fmcsa.dot.gov/facts-research/research-
technology/report/rollstability-control.pdf. Accessed July 7, 2009.
[25] NHTSA, "Federal Motor Safety Standards; Air Brake Systems," D. o. Transportation,
Ed., ed, 2009.
[26] S. Chandrasekharan, et al., "Development of a Roll Stability Control Model for a Tractor
Trailer Vehicle," SAE International Journal of Passenger Cars-Mechanical Systems, vol.
2, p. 670, 2009.
51
Appendix A: Formulas for the Calculation of the Curve Coordinates
in Matlab
Equation 8
) D ∗ 15 ∗ (F + )
Where v is the intended velocity of the vehicle in the curve (mph)
(fs +e) is the coefficient of side friction plus the superelevation (this term was assumed to
be equal to 0.40)
Equation 9
3.15 ∗ )
H=
∗
Where L is the length of the curve (ft)
r is the radius of the curve (ft)
c is the rate of centripetal acceleration (assumed to be 1.3 ft/s3)
The angle of the spiral arc is given by:
Equation 10
H 180
K= ∗
2∗ M
Where K is given in degrees
Equation 11
HF
KF = N O ∗ K
H
Where KF is the angle of the spiral based on the small step increments of Ls
The x and y coordinates (in feet) for the given angles of KF are given by:
Equation 12
HF
&F = ∗ (100 − 0.0030462 ∗ KF )
100
52
Equation 13
HF
F = − ∗ 0.58178 ∗ KF − 0.000012659 ∗ KFT
100
From there, the last coordinate of the spiral curve becomes the first coordinate for the circular
part of the curve that is fitted according to the required radius.
53
Appendix B: Plots of HiL Brake Pressures for Validation
(e) Axle 5
does not experience any wheel lift when traveling around a curve too fast.
54
(e) Axle 5
experiences single wheel lift while traveling around a curve too fast.
55
(e) Axle 5
surpasses six degrees of roll when traveling around a curve too fast. (Previously shown as Figure 12)
56
(e) Axle 5
experiences multiple wheel lift when traveling around a curve too fast.
57
(e) Axle 5
rolls over when traveling around a curve too fast.
58
Appendix C: Copyright Documentation
59
60
Kelly M. Donoughe
12/12/2010
61

BW8055 Us 000

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

BW8055 Us 000

Uploaded by

Copyright:

Available Formats

Evaluating the Effectiveness of Electronic Stability Systems in

Reducing Truck Rollovers

Thesis submitted to the faculty of the Virginia

Hesham A. Rakha, Chair

Keywords: hardware-in-the-loop, electronic stability control,

© 2010 by Kelly Donoughe

The objective of this research is to develop a customized hardware-in-the-loop system

1.1 Overview of Truck Rollovers and Rollover Mitigation

Most of the research regarding rollover prevention and mitigation

1.2 Thesis Contributions

- Build a hardware-in-the-loop testbed in order to evaluate electronic stability systems

1.3 Thesis Layout

- a literature review of relevant research;

2.1 Progression of Safety Technology Testing

2.3 Overview of Rollover Characteristics

h is the height of the truck

ay is the lateral acceleration

Fi is the vertical force on the tires

T is the track width of the truck, and

Δyy is the lateral motion of the cent

2.4 Previous Roll Stability Control Research

In 2009, the University of Michigan’s Transportation Research Institute (UMTRI)

Table 1: Total crashes, deaths, and injuries that could

Annual Prevented by:

2.5 Crash Databases

2.6 Literature Review of Benefit Estimation

#      ! /#  #$    !

&' ! !%#

∑() "+%''" )$  %#

0# !#,  100 ∗ 1

Vehicle Type Crash Type Reductions

Researchers at the University of Iowa analyzed the effectiveness of electronic stability

Where, ,9 indicates conflict number j,

Table 4: Number of Crashes Prevented based on Houser’s research [8]

Rollover (Fast Turn)

3.1 Detailed Summary of the UMTRI Study

3.1.1 Scope of the Problem

3.1.2 Identification of Target Crash Types

UMTRI used a combination of crash databases to determine which crashes should be

Selection Algorithms for Relevant Crash Types

3.1.3 Linking the Crash Data with the HiL System

1. Radius of <100m with a high center of gravity

3.2 Evaluating Rollovers with the HiL Simulator

3.2.1 Truck Modeling

3.2.2 Driver Model

3.2.3 Testing Procedures

Hesham Rakha (Corresponding Author)

Paper 11-1376 at the 90th Transportation Research Board Annual Meeting

The truck is fitted with an anti-lock

• Sprung mass- weight; roll, pitch, and yaw inertia

• Centerline geometry - can be input as an X-Y table of horizontal coordinates or can be

HiL Computing Platform

Table 5: Hardware computing requirements

Mechanical License file located at: C:\Flexlm\k107302.lic

•PXI-6251 Multi-Function Digital Acquisition (DAQ) Module

•PXI-6602 Timer/Counter Module

•PXI-6713 Analog Output Module

•PXI-8464/2 Two Port CAN Interface Module

1. Highest speed with no wheel lifts

Volvo with CG at 2.3

(c) Axle 3 (d) Axle 4

2.0 meter Center of Gravity Truck

2.0 meter Center of Gravity Truck

# !/# #$ !

&' ! !%#

∑() "+%''" )$ %#

0# !#, 100 ∗ 1