Impacts of Road-Trains on the Geometric

Design of Highways
Natacha E. Thomas, M.ASCE 1; and Francisco J. Martinez-Perez 2
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

Abstract: For enhanced efficiency and sustainability, transportation must rely on new technologies. Researchers and engineers work nowa-
days on many valuable and green transportation ideas. One such idea creates vehicle platoons on freeways to achieve multiple benefits:
reductions in gas consumption, vehicle emissions, and traffic congestion through enhanced aerodynamics and road space utilization, as
well as improvements in safety and driver comfort through fail-safe mechanisms and moderation of collision damages given similar vehicle
speed. The technological enhancements are built into the vehicles. The existing freeway infrastructures do not necessitate modification. Still,
the interactions between the human factors, or lack thereof, and the new technologies may directly impact the traditional guidelines for
freeway design. Over time, the dedicated freeways of the future, whether rehabilitated or built anew, could satisfy guidelines derived spe-
cifically for road-trains. Further, freeways built to current guidelines may accommodate travel speeds higher than those originally anticipated
per design. It remains to understand (1) why existing freeway lanes can accommodate road-trains, (2) whether design guidelines for road-
trains are significantly different from current guidelines, and (3) whether road-train operation results in significant a posteriori design speed
increases. This article seeks to answer these very questions. The investigation of a continuum of transitory to end state scenarios concluded
that road-trains promote both shorter required and longer available sight distances, thereby promoting the adequacy of existing freeways. It is
shown that road-trains reduce minimum length requirements and increase a posteriori design speeds for all curves. Changes are less drastic for
horizontal than vertical curves. DOI: 10.1061/(ASCE)TE.1943-5436.0000751. © 2014 American Society of Civil Engineers.

Introduction The dedicated freeways of the future, whether rehabilitated or

built anew, could satisfy guidelines derived specifically for road-
Efficiency and sustainability are design guidance principles by trains. This article specifically addresses (1) the understanding of
which today’s engineers must abide. Efficiency maximizes the the adequacy of existing freeways for road-train operation, (2) the
ratio of product output, such as job access, mobility, or safety, to direct impacts of this operation on guidelines for geometric design
resource input, such as roadway construction, maintenance, and op- of freeways, and (3) the maximum speeds achievable, or a pos-
erational costs. Sustainability permits endurance, making the least teriori design speeds, given freeways built to AASHTO’s (2011)
impact on the environment and future generations. Innovations to- minimum guidelines. The article envisions separate whole free-
ward efficient and sustainable infrastructures are highly beneficial ways dedicated to road-trains; thus deriving revised guidelines
when resources are scarce, but their expansions can prove contro- for the geometric design of such freeways. Conceivably, the revi-
versial if not detrimental, as in transportation networks. sions could apply to individual dedicated travel lanes solely. Any
A newer transportation innovation is road-trains, automated or mixed operation on freeways of human-driven and autonomous
semi-automated vehicles equipped with specialized sensors and vehicles would render moot the design impacts derived.
wireless communication systems that travel on freeways in small Revisions derived mostly concern the horizontal and vertical
convoys or “platoons” on regular highway lanes. The communica- alignments as the cross-section impacts are easily anticipated.
tion systems afford the continuous exchange of information neces- The precise systems used for lateral guidance by autonomous
sary to coordinate vehicles and harmonize flow, reducing speed road-trains may decrease lane width by 0.6 to 0.9 m below the
fluctuations and traffic shock waves while maximizing the highway standard design width of 3.7 m (Shladover 2008). The sections that
capacity and the passengers’ safety, while also minimizing gas con- follow present the relevant literature, the methodology, the find-
sumption and exhaust pollution (Ashley 1998). The technological ings, the conclusions and the future studies. The study creates
enhancements are thus built into the vehicles and the existing free- varied road-train scenarios, both experimental or what-if, and
ways remain adequate. analyzes them.
Still, the human factors, or lack thereof, and the new opera-
tional mode may combine to directly impact freeway design.
Literature Review
1
Associate Professor, Dept. of Civil and Environmental Engineering,
Univ. of Rhode Island, Bliss Hall, 1 Lippitt Rd., Room 203, Kingston, For brevity, the literature reviews only the technological aspects of
RI 02881 (corresponding author). E-mail: [email protected] experimental road-trains likely to impact the geometric design of
2
Research-Assistant, Dept. of Civil and Environmental Engineering, freeways. It further points to the AASHTO (2011) guidelines likely
MS Program, Univ. of Rhode Island, 1 Lippitt Rd., Room 203, Kingston, to change under road-train operation.
RI 02881. E-mail: [email protected]
The experiments identified encompass the Californian Partners
Note. This manuscript was submitted on May 1, 2014; approved on
September 10, 2014; published online on October 24, 2014. Discussion for Advanced Transportation and Technology (PATH) (Alvarez and
period open until March 24, 2015; separate discussions must be submitted Horowitz 1999b), the European Safe Road-trains for the Environ-
for individual papers. This paper is part of the Journal of Transportation ment (SARTRE) (Robinson et al. 2010; Davila and Nombela 2010;
Engineering, © ASCE, ISSN 0733-947X/04014087(9)/$25.00. Bergenhem et al. 2010), PROMOTE CHAUFFEUR I and II

© ASCE 04014087-1 J. Transp. Eng.

J. Transp. Eng.
 (Braun et al. 1999), and the German KONVOI (Deutschle 2010). Table 1. Comparison of PATH and SARTRE Experimental Road-Train
The PATH project, by the Institute of Transportation Studies at Projects
the University of California, Berkeley, in collaboration with the Characteristics PATH SARTRE
California Department of Transportation, represented the precursor
Driving system Autonomous, Semi-autonomous,
study. In-vehicle magnetometers sensed magnets buried at intervals driverless truck-led by
in the roadbed, thus providing an autonomous means for vehicles professional driver
to monitor and adjust their locations and velocities within a pla- Controlling devices Magnets in Vehicle camera,
toon. The response to changes in the motions of vehicles ahead roadbed and radar and laser sensors
occurred instantaneously as compared with human drivers, given vehicle
that the vehicles received 50 times per second radio communication magnetometers
updates of speeds and accelerations (Alvarez and Horowitz 1999b). Fuel consumption 20% Under study
A successful PATH demonstration showcased a driverless decrease
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

Connected vehicles 8 3–5

8-vehicle platoon experiment traveling at 105 km=h at fixed dis-
number
tance headways of 6.5 m in August 1997 near San Diego. Exten- Vehicle gap 6.5 m 5–15 m
sive studies of platoon operation during the late 1990’s estimated Date 1997 2012
an effective throughput of about 4,200 vehicles per hour per lane Location California Barcelona, Spain
(vphpl) versus a throughput range of 2,000 to 2,500 vphpl under Maximum speed 105 km=h 85 km=h
normal operation (Alvarez and Horowitz 1999a). Capacity in-
crease varied with platoon size and vehicle spacing within platoon.
At any travel speed, smaller vehicle spacings produce a greater Methodology
yield but challenge the platoon control system more. A balance
between operation enhancement and design control challenge must The scrutiny of existing experimental designs and modes of oper-
be established. PATH anticipated up to 20% reduction in gas con- ation of road-trains helps apprehend their conceptual pattern and
sumption due solely to the aerodynamics of road-trains (Zabat et al. anticipate what-if operational scenarios. This pattern relates to the
1995) with no consideration of congestion reduction. adequacy of current freeways to meet the operational needs of road-
Shifting costs to the private sector best motivates system adopt- trains. The sum total of existing and what-if scenarios enables the
ability in the United States. PATH’s buried magnets within the derivation of the sought-after revised guidelines and a posteriori
roadbed placed infrastructure costs on the public sector. More re- speeds.
cent road-trains, including the European SARTRE, require no Each existing road-train design dictates criteria for five distinct
roadbed modifications and as such seem more adoptable. design variables that impact vertical and horizontal freeway align-
The SARTRE project is a three-year program funded by the ments: perception reaction time (PRT), deceleration rate (a), driver
European Commission under the Framework 7 program including eye height (h1 ), obstacle height (h2 ), and headlight mounting height
Ricardo, U.K., Idiada and Robotiker Tecnalia of Spain, Institut fuer (h3 ). Ongoing experiments as described, California’s PATH,
Kraftfahrwesen Aachen (IKA) of Germany, SP Technical Europe’s SARTRE and KONVOI, and France’s CHAUFFEUR I
Research Institute of Sweden, and Volvo Car Corporation and and II decrease or eliminate PRT, either through use of expert driv-
Volvo Technology of Sweden. SARTRE aims to develop safe ers or autonomous vehicles. In addition, SARTRE, CHAUFFEUR I
and environmental road-trains on unmodified public highways and II, and KONVOI increase the driver eye and the headlight
given full interaction with other vehicles (Robinson et al. 2010; mounting heights given a truck-led (T-led) platoon. PATH enhan-
Davila and Nombela 2010; Bergenhem et al. 2010). ces driver eye height through a back of the rearview mirror location
A successful demonstration showcased a five-vehicle road-train, for camera or radar systems. All pilots increase the lead vehicle
in May 2012 near Barcelona, Spain, as led and controlled by a truck deceleration rate given expert drivers or autonomous vehicle sys-
with a professional or expert driver, accelerating to a speed of tems. No changes in horizontal sight line offset (HSO) are antici-
85 km=h, at separations of 5–15 m, and stopping together as a pated for road-trains because this criterion does not relate to
whole. Performance proved higher than for a stream of even highly operation.
skilled human drivers. SARTRE is currently undertaking a new Decreases in PRT along with increases in deceleration rate
environmental phase in order to determine the total percent reduc- engender the decrease in required sight distance on curves. On
tion in fuel consumption achieved. the other hand, increases in h1 , h2 , or h3 engender increases in
Further projects include the European KONVOI (Deutschle available sight distance on vertical curves. Modes of operation that
2010) and the French CHAUFFEUR I and II (Braun et al. promote either change, a decrease in required sight distance or an
1999), supported by Daimler Benz and IVECO, which link specifi- increase in available sight distance, enhance the adequacy of
cally commercial trucks by wireless radio to form truck convoys, existing freeways. All pilots thus enhance this adequacy, thereby
leaving the roadbed unmodified. The lead truck is operated by an limiting investments in freeway infrastructure retrofits. What-if
expert driver, as for SARTRE. Table 1 presents a summary of the scenarios must be as accommodating of existing freeways.
design features of the PATH and the SARTRE projects. The other
two projects are omitted because they resemble SARTRE notwith-
standing their convoys consisting solely of trucks. Chapter 3 Road-Train Scenarios
of AASHTO (2011) provides the user with guidelines on the der- This article assumes a number of operational scenarios to derive the
ivations of available and required sight distances along curves impacts of road-trains on the geometric design of freeways. Pilots
[stopping sight distance (SSD) and available sight distance (S), constitute the basis for two of the scenarios, the SARTRE-like and
respectively]. Elsewhere, sight distance is unrestricted. A main PATH-like scenarios. However, two other scenarios were derived
design principle of highway alignments, whether horizontal or ver- based on the developmental pattern of experimental projects.
tical, ascertains that available sight distance must be greater than Road-trains vary in the nature of their lead vehicle, whether
that required. If not, vehicle, driver, and passenger safety may passenger vehicle (P) or truck (T), and their vision and driving
be compromised. systems, whether human (H) operated or machine (M) operated.

© ASCE 04014087-2 J. Transp. Eng.

J. Transp. Eng.
 Road-train design characteristics impact directly on the control deployment state between SARTRE and PATH and the latter,
variables for freeway design. (Human drivers of road-trains are the design end-goal. It further selects the SARTRE-like and Path-
all considered expert drivers.) like scenarios, THH and PMM, to generate a total of 5 study sce-
PATH’s M-vision systems adapt to the back of the rearview mir- narios that include the do-nothing. Three scenarios are discounted,
ror. Assuming this trend keeps for all road-trains given machine PHH, PHM and TMM, to limit study scope. [PHH would only dif-
vision systems, P-led or T-led road-trains would engender different fer from AASHTO’s scenario (2011) per its expert drivers. Further,
driver’s eye height, h1 . Hence, both vision systems and lead vehicle two highly unlikely scenarios, PMH and TMH, were discounted.]
type impact h1 . M-vision could assist a transitional and H-driven The sections that follow introduce and describe all scenarios
SARTRE-like scenario by providing alerts to the driver on hazard- analyzed.
ous conditions at distances ahead and within timeframes beyond
human capabilities. A platoon driver would then be afforded shorter Do-Nothing Scenario (S.0)
PRTs than otherwise. It follows that driving system and lead ve- This scenario constitutes the null scenario and entails no road-train
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

hicle types alone do not uniquely determine vision system height; implementation and thus no changes in the current AASHTO
as the SARTRE and SARTRE-like scenarios would result in com- (2011) design guidelines. It serves as gauge for the performance
pletely different h1 values. The variable h3 only varies with lead enhancements by other road-train scenarios.
vehicle, h2 with both lead and following vehicles, whereas driving SARTRE-Like Road-Train Scenario (S.1 or THH)
and vision systems impact jointly on PRT. Deceleration, a, varies This scenario retains all SARTRE features that impact geometric
mainly with lead vehicle type and possibly driving system. design including a T-led platoon with a professional driver. As
Lead vehicle, vision, and driving systems encompass the road- such, PRT is expected to decrease below, while a and h1 increase
train design characteristics that impact the geometric design criteria above, S.0 levels.
and thus, the guidelines for their facilities. The remainder of the
discussion will refer to a particular road train scenario using a SARTRE-Like Scenario with Obstacle Warning System
three-letter acronym, XYZ, where X describes the lead vehicle type, (S.2 or THM)
P or T; Y describes the lead vehicle vision system, H or M; and Z This scenario constitutes a transitory what-if scenario between
describes the lead vehicle driving system, H or M. SARTRE and PATH, S.2. It incorporates PATH-like obstacle-
Road-train designs to analyze result from a choice of two lead warning systems into SARTRE-like T-led platoons with profes-
vehicles and two driving and vision systems. Thus, 8 ordered com- sional drivers. Obstacle-warning systems convert complex
binations need be considered, PHH, PHM, PMH, PMM, THH, situations requiring high PRTs into simple ones requiring much
THM, TMH, and TMM, given combination characters representing shorter PRTs, whereas the interpretation of a consistent and simple
X, Y, and Z, respectively, as described above. Each combination message becomes a routine and expected simple task. They detect
dictates the design criteria of PRT, a, h1 , h2 , and h3 within certain through imaging apparatus on rearview mirror backs at vision
margins. height, h1 , enhanced above those of S.0 and S.1 scenarios.
However, a global optimum design stretches the limits of these
criteria; for instance by locating driver’s eye height at infinity, thus Path-Like Road-Train Scenario (S.3 or PMM)
removing all sight-distance restrictions on freeway geometric de- This scenario retains all PATH’s features including P-led platoons
sign for daytime. Such a road-train could be interpreted in real life made autonomous by navigation systems, radar, and vision sensors.
as providing satellite-assisted and/or GPS-enabled vision to lead The human PRT to changes in driving conditions dwarfs that
vehicles in platoons. Coupled with nighttime assisted vision, sight anticipated for automatic driving, car-following, and lane-change
distance restrictions could be removed entirely. systems where data updates can occur at 50 times per second
PHH is SARTRE-like, H-driven with an H-vision system, but (PATH) and reactions in a negligible amount of time (Larburu et al.
P-led. A passenger car as lead vehicle offers no real advantages 2010). Hence, PRTs for this scenario are much decreased from their
over a truck (except maybe in acceleration up slopes). Safety is S.0 levels.
anticipated as a major driving force for road-train deployment, PATH-Like with Remote Obstacle Detection
and a truck as lead much enhances safety. Similarly, safety strongly (S.4 or Global Optimum)
deters from PHM, which is tantamount to PHH with M-vision as To reap optimal design benefits, the authors envisioned an end-state
enhancement. PMH is PATH-like except with human vision, what-if scenario, S4, which incorporates remotely-sensed obstacle-
presents no real advantage over PATH and is highly improbable warning systems with nighttime vision into PATH-like P-led pla-
given its human use simply as scout or vigil, instilling undue delays toons of autonomous vehicles. Available sight distance would in
to M-driven systems thereby negating their efficiencies. PATH, pre- theory be unlimited, hence would exceed that required under
selected for analysis, exemplifies PMM. any design option. Such systems would remove sight distance
SARTRE, also preselected for analysis, exemplifies THH. THM as a design constraint and engender lengths of curves based on
is SARTRE-like but with automated vision, presents some advan- the absolute minima. Determination of the technological feasibility
tages over SARTRE as likely to decrease PRT, and is therefore an- of such a system is, however, beyond the scope of this study.
alyzed. TMH presents similar disadvantages to PMH. TMM is a
PATH-like scenario that is, however, T-led. This scenario presents
some advantage above and beyond PATH given higher vision sys- Design Criteria for All the Scenarios
tem and headlight mounting. However, it presents none over the The vehicle types, driving systems, and vision systems anticipated
global optimum with an infinite driver’s eye location and the com- by the various scenarios enable the derivation of their design cri-
plete removal of sight distance limitations and is not selected for teria. To prevent damages to the regular passenger vehicles con-
analysis. tained in road-trains, whether leading or following, the obstacle
In summary, of the 8 precited combinations, the study selects 2 height, h2 , remains at its AASHTO (2011) value, 0.6 m, for all
what-if scenarios, THM and the global optimum, through an inter- scenarios. Given P-led platoons with M-vision systems placed
pretation of deployment probabilities and the advantageous gains at rearview mirror height, S.3 fixes h1 at 1.20 m. S.4 fixes h1 at
in design achieved. The earlier scenario constitutes a transitory infinity given satellite vision. To account for T-led rather than

© ASCE 04014087-3 J. Transp. Eng.

J. Transp. Eng.
 P-led platoons, S.1 and S.2 fix h1 and h3 at 2.50 m and 1.0 m, for PRT, driver eye height, and headlight mounting height,
respectively. The literature assigns 95th percentile values of respectively.
2.72 m, and 2.33 m to truck and truck driver eye heights, respec- For example, given scenario S.1.1, A ¼ 4%; V ¼ 100 km=h,
tively (Fambro et al. 1997). A back of the rearview mirror location and the associated SSD ¼ 145 m, a minimum length of crest curve,
places the stereovision imaging sensor halfway between driver eye L ¼ 80.1 m, is calculated using the relevant AASHTO (2011) equa-
height and truck height; thus approximately 2.50 m above the high- tion and the criteria for S.1.1 listed in Table 2. (A and V represent the
way surface. Truck headlight heights vary from 0.92 to 1.35 m, algebraic difference in grade and the speed, respectively.) Also,
with 0.97 and 1.08 m being the values exceeded at the 95 and given scenario S.1.1, A ¼ 4%, V ¼ 100 km=h, and the associated
90% levels, respectively (Fambro et al. 1997). Hence, the choice SSD ¼ 145 m, a minimum length of sag curve, L ¼ 118.9 m, is
of 1.0 m related. Both S.3 and S.4 maintain h3 at its AASHTO calculated. Both these values must be compared to the absolute min-
(2011) value of 0.6 m. imum curve length provided by AASHTO (2011) for aesthetics,
Systems with expert drivers benefit from enhanced PRTs. 0.6V or 60 m. This absolute minimum is exceeded in both the cases.
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

Unfortunately, a literature review did not pinpoint a 90th percentile Note the difference with the minimum L ¼ 208.8 m and 178.4 m
PRT for professional truck drivers. Hence, S.1 and S.2 assume this for scenario S.0 or per AASHTO for crest and sag vertical curves,
percentile to range from 1.5 to 2.0 s and from 0.5 to 1.0 s, respec- respectively. Finally, given scenario S.1.1, HSO ¼ 10 m, V ¼
tively. Results for 90th percentile PRTs within or slightly outside 100 km=h, and the associated SSD ¼ 145 m, a minimum radius
these limits can be interpolated or extrapolated. The 90th percentile of horizontal curve, R ¼ 245.5 m, is calculated. However,
PRT of a professional driver is expected to lag that for alerted driv- AASHTO (2011) provides an absolute minimum radius of
ers under complex situation given S.0, 1.64 s. Scenarios S.3 and S.4 328 m at e ¼ 12%, which is not exceeded and must be respected.
quasi eliminate the human factors given fully autonomous vehicles Note the difference with the minimum R ¼ 403.3 m for scenario
and fix PRTs at 0 s. Note that the adopted PRT value of 2.5 s by S.0 or per AASHTO (2011).
AASHTO (2011) since 1954 exceeds the 90th percentile value for
all drivers. Other countries utilize design values as low as 2.0 s
under S.0 or normal operation (Fambro et al. 1997). Revisited (A Posteriori) Design Speeds
Systems with trained professional drivers or those autonomous A further interesting consideration is the increase in “design” speed
are expected to achieve a higher deceleration rate than for S.0. A achieved on roadways designed to current guidelines given road-
review of the literature could not reveal a 90th percentile value to train operation. Road-trains require shorter sight distances than are
utilize just for the professional driver population. Still, “Most driv- available on curves designed to AASHTO (2011). Latter curves
ers decelerate at a rate greater than 4.5 m=s2 when confronted with could thus accommodate increased operational speeds for these
the need to stop for an unexpected object in the roadway (AASHTO modes and can thus be construed retroactively as designed for
2011).” Thus, S.1, S.2, S.3, and S.4 assume a deceleration rate of higher speeds. (Higher free flow speeds, which can be related to
4.5 m=s2 , considered to be a conservative value for the 90th per- design speeds, may increase capacity for certain roadways further
centile as professional drivers would be expected to perform better mitigating congestion.)
than most. M-driven S.2 and S.4 could probably reach higher Available sight distances are calculated for crest, sag, and hori-
values up to the comfort limit for human passengers. zontal curves, respectively, given the AASHTO (2011) minimum
Table 2 summarizes the design criteria derived for all scenarios lengths of curves, L, recommended at different values of speed and
and enables their direct comparison to AASHTO’s typical values as the control criteria for each individual road-train scenario. Then, the
listed under S.0. It introduces sub-scenarios S.1.1 and S.1.2 of S.1 newly obtained S values help determine the maximum travel speeds
as well as S.2.1 and S.2.2 of S.2 as associated with the limits of the (identical here with design speeds for simplicity) for each scenario
value range of PRT for each scenario. The design criteria combine by equating available and required sight distances. The process is
with the AASHTO (2011) methodology to derive the guidelines for iterative for horizontal curves because side friction depends on
geometric design of freeways under road-train operation. Alter- speed. Care must be taken not to exceed the absolute maximum
nately, a methodology earlier derived by an author of the current speed, L=0.6, on vertical curves or that consistent with stability
article (Thomas et al. 1998) could be utilized for vertical curves. on horizontal curves.
Assuming S.2.1, for example, the AASHTO (2011) minimum
Revisited Guidelines for Curve Design required length of crest curve, L, of 208.1 m, given an initial V ¼
100 km=h and A ¼ 4%, leads to an available sight distance, S,
The methodology equates the available sight distances on vertical
equal to 240.3 m. Note that criteria h1 and h2 are taken as
and horizontal curves to the required stopping sight distances to
2.5 m and 0.6 m, respectively, as described in Table 2 for scenario
determine the minimum recommended lengths and the ensuing
S.2.1, in the AASHTO (2011) equation for available sight distance.
length reductions of curves associated with road-train operation.
Then, obtained available sight distance, S ¼ 240.3 m enters the
Care must be taken to substitute the design criteria associated
AASHTO (2011) equation for stopping sight distance, given
with the varied road-train scenarios, in Table 2, for the default
PRT ¼ 1.0 s and a ¼ 4.5 m=s2 as further described in Table 2
AASHTO (2011) criteria of 2.5 s, 3.4 m=s2 , 1.08 m, and 0.6 m
for scenario S.2.1, to solve for V ¼ 151.2 km=h, which does
not exceed the maximum value of 346.8 km=h and is thus ac-
Table 2. Design Criteria for All Study Scenarios cepted. [Note the enhancement in design speed as opposed to
S.1 S.2 AASHTO’s (2011) original value of 100 km=h.]
Assuming scenario S.2.1 still, the AASHTO minimum required
Scenarios S.0 S.1.1 S.1.2 S.2.1 S.2.2 S.3 S.4 sag curve length, L, of 178.4 m (2011), given an initial V ¼
PRT (s) 2.5 2.0 1.5 1.0 0.5 0 0 100 km=h and A ¼ 4%, leads to an available sight distance, S,
a (m=s2 ) 3.4 4.5 4.5 4.5 4.5 equal to 200.5 m. Note that criterion h3 is 1.0 m as described
h1 (m) 1.08 2.33 2.50 1.20 ∞ in Table 2 for scenario S.2.1. Then, obtained available sight dis-
h2 (m) 0.6 0.6 0.6 0.6 0.6 tance, S ¼ 200.5 m, enters the AASHTO equation for stopping
h3 (m) 0.6 1.0 1.0 0.6 0.6
sight distance (2011), given PRT ¼ 1.0 s and a ¼ 4.5 m=s2 as

© ASCE 04014087-4 J. Transp. Eng.

J. Transp. Eng.
 further described in Table 2 for scenario S.2.1, to solve for Findings
V ¼ 136.9 km=h, which does not exceed the maximum value of
297.3 km=h and is thus accepted. Note the enhancement in design Given enhanced or lack of human factors, road-trains require
speed as opposed to AASHTO’s (2011) original value of shorter minimum curve lengths and radii than normal flows.
100 km=h. Figs. 1–3 present these minima for crest, sag, and horizontal
Further, the AASHTO (2011) minimum required horizontal curves, respectively, for all scenarios at different values of the
sight line offset, HSO, of 10.0 m, given an initial V ¼ algebraic difference in grade and speed. Further, roads designed
100 km=h, leads to an available sight distance, S, equal to to AAHTO’s (2011) guidelines accommodate faster travel, or
185.0 m, given criterion R ¼ 275 m. Then, obtained available sight can be perceived as a-posteriori designed to a higher speed, given
distance, S ¼ 185.0 m, enters the equation for required sight road-train operation. Figs. 4(a, b, c) present cross sections of
distance, given PRT ¼ 1.0 s and a ¼ 4.5 m=s2 as described in Figs. 1–3 at V ¼ 100 km=h. Fig. 4(c) presents the percent increase
Table 2 for scenario S.2.1, to solve for V ¼ 130.9 km=h. However, in design speed achieved a posteriori at initial design V ¼
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

this value does exceed the maximum travel speed associated to the 100 km=h, A ¼ 4% and HSO ¼ 10 m. The different graph
given radius, 111 km=h, which is thus adopted. Note the enhance- types aim to enable different visual comparisons with the line
ment in design speed as opposed to AASHTO’s (2011) value of type enabling general comparisons and the bar chart specific ones
100 km=h. at the set of common design criteria V ¼ 100 km=h, A ¼ 4% or
Similar approaches to the above may be followed to compute HSO ¼ 10 m.
the design speed at other combinations of algebraic differences Crest vertical curve design: Entering Fig. 4(a) with a value of
in grades and speeds. The next study section expands on findings. A ¼ 4% for S.1.1, a minimum length of crest curve, or L, equal to

Fig. 1. Design controls for crest vertical curves in meters for all scenarios: (a) S.0; (b) S.1.1; (c) S.1.2; (d) S.2.1; (e) S.2.2; (f) S.3

© ASCE 04014087-5 J. Transp. Eng.

J. Transp. Eng.
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Design controls for sag vertical curves in meters for all scenarios: (a) S.0; (b) S.1.1; (c) S.1.2; (d) S.2.1; (e) S.2.2; (f) S.3

80 m is obtained. Still from Fig. 4(a), the length of crest vertical of 328 m, given V ¼ 100 km=h and e ¼ 12%, which supersedes
curve recommended for scenarios S.1.1, S.1.2, S.2.1, S.2.2, S.3, that obtained, 246 m. Still using Fig. 4(c), given HSO ¼ 10 m, the
and S.4 equal 80, 64, 60, 60, 60, and 60 m, respectively. Note recommended minimum radii of horizontal curves to provide under
the drastic reduction in crest curve length achieved relative to scenarios S.1.1, S.1.2, S.2.1, S.2.2, S.3, and S.4 all equal the ab-
the customary AASHTO (2011) value of 208.0 m under S.0. These solute minimum curve radius, 328 m. For comparison purposes,
reductions vary from 130 to 150 m or from 60 to 70% of the pre- Fig. 4(c) also provides the AASHTO (2011) customary value of
stated base AASHTO (2011) value. 403 m, reflecting S.0. The anticipated reductions in curve radius
Sag vertical curve design: Entering Fig. 4(b), the lengths of sag achieved by the scenarios all equal 75 m, or 19% of the prestated
vertical curves recommended for scenarios S.1.1, S.1.2, S.2.1, base AASHTO (2011) value. The radial reductions multiplied by a
S.2.2, S.3, and S.4 equal 119, 106, 88, 78, 75, and 60 m, respec- factor of πΔ=180 convey the reductions in minimum recom-
tively. Note the reductions in sag curve length achieved relative to mended curve lengths. For angles swaying 40–120°, this factor
the customary AASHTO (2011) value of 180.0 m under S.0. These sways approximately 0.70–2.09. Thus, reductions of 13–40% in
reductions are not as drastic as for crest curves except maybe for the the minimum recommended lengths of curves are expected for
last four scenarios. They range from 60 to 120 m or 40–67% of the horizontal curves because of the high value of the absolute mini-
prestated base AASHTO (2011) value. mum radius on curve.
Horizontal curve design: Entering Fig. 4(c) at S.1.1 given V ¼ All the road-train scenarios contemplated quickly lead to the
100 km=h and HSO ¼ 10 m, obtains a minimum radius of 246 m. absolute minimum requirement for curve lengths and curve radii,
However, AASHTO (2011) provides an absolute minimum radius 60, 60, and 328 m, for crest curves, sag curves, and horizontal

© ASCE 04014087-6 J. Transp. Eng.

J. Transp. Eng.
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Sight line offsets in meters on horizontal curves for all scenarios: (a) S.0; (b) S.1.1; (c) S.1.2; (d) S.2.1; (e) S.2.2; (f) S.3

© ASCE 04014087-7 J. Transp. Eng.

J. Transp. Eng.
 10000
16

Algebraic Difference in Grade, A (%)

S4
14

12

Radius, R, Centerline of Inside Lane (m)

10

8
1000
6

4 S.4

0 Min. Radius For

Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

0 100 200 300 400 500 600 Design Purposes

(a) Length of Crest Vertical Curve, L (m) 100

10
0 10 20
Horizontal Sight Line Offset, HSO, Centerline Inside
(c) Lane to Obstruction (m)

16
Design Speed Increase (%)
S.4
Algebraic Difference in Grade, A (%)

14 500

12
400
10
300
8

6 200

4 100

2
0
0 S.0 S.1.1 S.1.2 S.2.1 S.2.2 S.3 S.4
0 100 200 300 400 500 600
Length of Sag Vertical Curve, L (m) Crest Vertical Curve Sag Vertical Curve Horizontal Curve
(b) (d)

Fig. 4. Comparative performance of varied scenarios and curve types at V ¼ 100 km=h: (a) minimum recommended lengths of crest curves for all
scenarios given V ¼ 100 km=h; (b) minimum recommended lengths of sag curves for all scenarios given V ¼ 100 km=h; (c) minimum recom-
mended radii for horizontal curves for all scenarios given V ¼ 100 km=h; (d) design speed increase for all curves and scenarios given
V ¼ 100 km=h, A ¼ 4% or HSO ¼ 10 m

curves, respectively. This outstanding performance truly prevents horizontal curves. Noticeably, the decrease in radius is invariant
comparisons between the various scenarios. For sag curves with scenario. This is due to the high value of the absolute mini-
however, the absolute minimum length is only reached by the mum radius surpassing the radii constraints per sight distance for
PATH-like system with satellite vision. This system, as anticipated, all scenarios.
performs overall as well as or better than all other systems. Also, as
expected, PATH outperforms SARTRE and both their subscenarios
perform better at lower PRT values than at higher ones (where a Conclusions
comparison is enabled). However, the SARTRE-like scenario with
obstacle detection comes close to performing as well as PATH-like The study aimed to understand the adequacy of existing facilities
scenarios. for road-train operation, to determine the necessary changes in
Overall, the scenarios achieve up to 71% decrease in crest curve design guidelines for dedicated road-train facilities and the a pos-
length. S.4 achieves a design speed increase up to 420.1% on crest teriori increases in design speeds afforded road-trains on existing
vertical curves. Scenario S.2.2 proves to increase design speed at facilities. An exhaustive literature review scrutinized the experi-
59%; thus more than S.3 at 48%, due to its enhancement in driver mental modes of road-train operations. Experimental modes formed
eye height. As for sag vertical curves, S.4 achieved the greatest the basis for study scenarios in addition to AASHTO’s (2011) and
decrease in curve length, at 67%, and increase in design speed, what-if scenarios. In total, five different scenarios were investigated.
at 194%. The increase in design speed proves not as substantial The methodology presented the design criteria that apply to study
as for crest curves. Finally, very timid decreases in radius, at scenarios, such as reduced perception-reaction times, enhanced
19%, and increases in design speed, up to 46%, are reached for deceleration rates, as well as the increased driver’s eye and light

© ASCE 04014087-8 J. Transp. Eng.

J. Transp. Eng.
 beam heights, and the solution approaches. It also provided an in- speeds; thereby furthering the geometric design impacts of road-
sight into deriving varied transitory and end state road-trains. trains. Enhanced longitudinal frictions permit even higher deceler-
The new road-trains promote both shorter required and longer ations than those utilized in study.
available sight distances, thereby promoting the adequacy of Future studies will address sensitivity analysis of the results
existing freeways. Further, they decrease the minimum required achieved to variations in speed and algebraic difference in grades.
lengths of curves of all types beyond their AASHTO (2011) levels. They will also assess for real-life roads constructed to AASHTO
Crest and sag curves in particular suffer significant reductions in guidelines the reductions in construction cost, the increases in traf-
minimum length in the range of 62–71% and 34–67%, respectively. fic capacities, and the exact reductions in fuel consumption follow-
Only moderate reductions in minimum radius are achieved for ing given road-train operation.
horizontal curves, 19%. All reductions were computed at V ¼
100 km=h and A ¼ 4%. As expected, the smaller the perception-
reaction time, all other factors being unchanged within the same References
Downloaded from ascelibrary.org by University of Texas at San Antonio on 11/16/14. Copyright ASCE. For personal use only; all rights reserved.

general scenario, the shorter the minimum recommended vertical

curve length or horizontal curve radius. The same pattern holds AASHTO. (2011). A guide on geometric design of highways and streets,
for a posteriori increases in design speed. 6th Ed., Washington, DC.
In spite of its lower headlight mounting height, PATH, as pas- Alvarez, L., and Horowitz, R. (1999a). “Safe platooning in automated high-
senger vehicle-led and machine-driven, exceeds the performance way systems. Part I: Safety regions design.” Vehicle System Dynamics,
of SARTRE, as truck-led and human-driven, in curve length 32(1), 23–56.
reductions given SARTRE’s lengthy human PRTs. However, a Alvarez, L., and Horowitz, R. (1999b). “Safe platooning in automated
highway systems. Part II: Velocity tracking controller.” Vehicle System
SARTRE-like scenario equipped with obstacle detection through
Dynamics, 32(1), 57–84.
an assisted vision system generates reductions in crest curve length Ashley, S. (1998). Smart cars and automated highways, ASME, New York.
that exceed or are tantamount to those of the fully automated PATH. Bergenhem, C., Huang, Q., Benmimoun, A., and Robinson, T. (2010).
No single scenario has more potential to reduce curve length than “Challenges of platooning on public motorways.” Proc., 17th ITS
a PATH-like scenario that is remote vision assisted, whether pas- World Congress, ITS America, Washington, DC.
senger vehicle-led or truck-led, as sight distance restrictions get Braun, A., Schulze, M., Schwarz, J., Sonntag, J., and Ulken, U. (1999).
lifted, and the absolute minimum recommendations prevail for ver- “Report on safety analysis of system components and hazard analysis
tical and horizontal curves. The incredible geometric design perfor- of tow-bar.” Contract Number: TR1009, PROMOTE-CHAUFFEUR
mance of this system makes its feasibility worth investigating. Consortium.
The reductions in required sight distances and the enhancements Davila, A., and Nombela, M. (2010). “SARTRE: Safe Road-trains for the
in available sight distances induced by road-train operation on Environment.” Proc., 17th ITS World Congress, ITS America,
curves built strictly to minimum AASHTO (2011) guidelines may Washington, DC.
contribute to increased travel, or a posteriori design, speeds. The Deutschle, S. (2010). KONVOI—Development and examination of the ap-
plication of electronically coupled truck convoys on highways, Institut
methodology computed these anticipated increases in travel speeds
fuer Kraftfahrzeuge, RWTH Aachen Univ., Germany.
on all curve types given some basic assumptions. For horizontal
Fambro, D. K., Fitzpatrick, K., and Koppa, R. J. (1997). “Determination of
curves, for instance, HSO was kept constant at 10 m. For V ¼ stopping sight distances.” NCHRP, Rep. 400, Transportation Research
100 km=h and A ¼ 4%, increases in the range of 36–59% and Board, Washington, DC.
23–46% could be expected for travel speed on crest and sag curves, Larburu, M., Sanchez, J., and Rodriguez, D. J. (2010). “Safe road-trains for
respectively, for scenarios other than S.4. The latter scenario gen- the environment: Human factors’ aspects in dual mode transport sys-
erates increases in design speed of 420 and 198%, respectively. tems.” Proc., 17th ITS World Congress, ITS America, Washington, DC.
Only moderate increases in speed are achieved for horizontal Robinson, T., Chan, E., and Coelingh, E. (2010). “Operating platoons
curves, 11%. on public motorways: An introduction to the SARTRE platooning pro-
The undertaken study may serve to guide the selection of road- gramme.” Proc., 17th ITS World Congress, ITS America, Washington,
trains in sustainability studies where various advantageous or DC.
disadvantageous design aspects of their deployments must be Shladover, S. (2008). “Assessment of the applicability of cooperative
weighed. The current study constitutes a valuable guide in weigh- vehicle-highway automation systems (CVHAS) to bus transit and
intermodal freight: Case study feasibility analyses in the metropolitan
ing various scenarios with regard to freeway geometric design.
Chicago region.” Final Rep., Rep. No. CA04-0494, California Partners
Further, the study has highlighted the potential gains in design for Advanced Transit and Highways, Richmond, CA.
efficiency associated with technological enhancements in side Thomas, N., Hafeez, B., and Evans, A. (1998). “Revised control parameters
and longitudinal tire frictions on wet pavement. The high value for vertical curves.” J. Transp. Eng., 10.1061/(ASCE)0733-947X(1998)
of the absolute minimum radius of horizontal curves that promotes 124:4(326), 326–334.
stability truly prevents any outstanding radius reductions owing to Zabat, M. A., Stabile, N. S., and Browand, F. K. (1995). “Estimates of fuel
road-trains. Side friction enhancements would result in much lower savings from platooning.” Proc., ITS America Annual Meetings, ITS
minimum curve lengths and much higher a posteriori design America, Washington, DC, 1203–1208.

© ASCE 04014087-9 J. Transp. Eng.

J. Transp. Eng.