Está en la página 1de 5

JOURNAL OF TRANSPORTATION

SYSTEMS ENGINEERING AND INFORMATION TECHNOLOGY


Volume 10, Issue 6, December 2010
Online English edition of the Chinese language journal
Cite this article as: J Transpn Sys Eng & IT, 2010, 10(6), 7681.

RESEARCH PAPER

Safety Evaluation for Freeway Exit Ramp Based on


Speed Consistency
GUO Tangyi1, DENG Wei1,*, LU John2
1. School of Transportation, Southeast University, Nanjing 210096, China
2. Department of Civil and Environmental Engineering, University of South Florida, Tampa, Florida 33620, USA

Abstract: Speed differential has direct relationship with road safety. Conventional speed consistency measures draw sample data
from independent population that follows a normal distribution. These methods may fall into ecological fallacy and overestimate the
safety level of road elements. After deducing the relationship between the difference in operating speed V85 and85( V ), the 85th
percentile value of individual speed reduction, this paper recommends 85( V ) and speed reduction rate 85( V )R is used to

evaluate ramp safety. Point speeds of individual vehicle at freeway diverge area, upper ramp, and lower ramp are collected by radar
guns. Safety performance of 37 exit ramps are appraised using three different measures, namely, V85 , 85( V ), and 85( V )R.

The results show that 85( V ) is 1.42 to 2.02 times of V85 , and that the evaluation result of 85( V )R is between the other two.

The measure 85( V )R considers not only individual speed but also the base speed on upstream element, which avoid the pitfalls of
ecological fallacy and overestimation possessed by conventional measures. It is a safer and reasonable measure that should be
adopted in practice.
Key Words: highway transportation; speed consistency; freeway exit ramp; difference in operating speed V85 ; the 85th percentile

value of individual speed reduction85 V ; speed reduction rate 85( V )R

Introduction

Freeway exit ramp is the only traffic connection between


freeway mainline and crossroad. To exit a freeway, drivers
may need to make lane changes to the deceleration lane, and
then slow down to enter their target ramp. While driving on a
ramp, to maintain safety and comfort, drivers adjust their
speed according to the alignment (radius, curvature grade,
super elevation, etc.) and sight condition. In the whole exiting
process, speed is always changing, not constant, making the
exit ramp a bottleneck of safety on freeway. Previous
research[1] showed that speed differential, rather than the
absolute value of speed contributes more to traffic crashes.
The bigger the difference, the greater the risk is. In other
words, the better the consistency, the safer the road is. It is
well accepted that safety performance of exit ramp can be
determined qualitatively by design consistency measures.
The factors influencing speed on a ramp include placement
distance of advance guide sign, length of deceleration lane,

ramp radius, deflection angle of ramp, length of ramp, grade,


pavement condition, etc. Lamm et al. built a model to describe
the relationship between the operating speed V85 and the
curvature change rate. In another model they used the radius
of curve as explanatory variable[2]. Later in 1990, they
modified the second model and emphasized the impact of
radius. Islam et al. drew similar conclusion that radius of
curve is the most significant parameter influencing operating
speed on curves[3]. It was also noticed that there were
remarkable differences among the operating speed on the
same horizontal curve at the point of curve (PC), point of
tangent (PT) and at the middle of curve (MC). The operating
speed at PC, PT and MC was predicted[4] by Islam. Krammes
et al. developed a speed profile model to evaluate horizontal
alignment consistency on two-lane rural highways. The model
estimated operating speed at several points along the
horizontal curve using degree of curve, length of curve, and
deflection of angle as variables[5]. Two similar models were
developed to predict operating speed by McFadden et al.[6] It

Received date: Mar 31, 2010; Revised date: Aug 9, 2010; Accepted date: Sep 6, 2010
*Corresponding author. E-mail: dengwei@seu.edu.cn
Copyright 2010, China Association for Science and Technology. Electronic version published by Elsevier Limited. All rights reserved.
DOI: 10.1016/S1570-6672(09)60074-X

GUO Tangyi et al. / J Transpn Sys Eng & IT, 2010, 10(6), 7681

should be noticed that the variables used in the models


proposed by Krammes and McFadden have self-correlation,
which may produce relatively high R2. In other words, the
higher R2 of their models is not necessarily an indication of
better goodness-of-fit.
Researches on speed consistency and design consistency
can be classified into two areas; one is for single element, and
the other for successive elements. For a single element, the
consistency measures include the difference in real speeds
85( V ), difference in operating speeds V85 , and the
difference in operating speed and design speed V85 VD . For
successive elements, the measures are the difference in real
speeds V , difference in operating speeds V85 and the 85th
percentile speed reduction 85( V ). Based on these speed
differences, Leisch and Lamm provided the criteria for
consistency evaluation, which are summarized in the
following Table 1[6,7].

Aiming at weakening the pitfalls of aggregate measures,


McFadden et al. developed a measure named 85MSR, the85th
Maximum Speed Reduction[9]. This measure described the
maximum speed reduction between two successive elements.
MaFadden etc. and Misaghi etc. both proved the relationship
that 85MSR is about two times of traditional measure V85 .
They pointed out that using V85 would overestimate safety
performance. However, the maximum speed reduction of an
individual vehicle is hard to capture.
.

Table 1 Criterion of speed consistency evaluation


Element
type
Measure

Criterion

Single element

V85 ,

Successive
elements

V85 ,

Design
level

Safety
level

Good

Good

V85 VD

85MSR

10 km/h

10 km/h

1020km/h

1020 km/h

Fair

Fair

20 km/h

20 km/h

Poor

Poor

V85 , V85 VD are operating speed-based measures,


which assume that speeds are subject to a normal distribution.
The data used are aggregated. Using aggregated data rather
than individual data to analyze travel behavior may be
misleading. This is called ecologic fallacy which describes
the phenomenon that what seems true for a group may not be
true of the individual, as some information gets lost during the
aggregation process[8]. Fig. 1 illustrates an example problem
of ecological fallacy in transportation. Speed increases with
the increase of the radius on both element A and B, which
accords with our common sense. If we use the 85th percentile
speed to regress the relationship between speed and radius;
however, the fitting line goes down while the radius increases.
The 85th percentile speed on element A is larger than that of
element B. Nevertheless, the corresponding radius of element
A is smaller than that of element B. In other words, the 85th
percentile speed decreases with the increase of the radius,
which contradicts the reality in transportation.
The practical implications of ecological fallacy can be quite
serious. Decision-makers may draw misleading conclusion
that a smaller radius is required for element B. Then the
engineer or practitioners construct a curve with a small radius.
The consequence is that sharp speed reduction occurs on this
curve and thus a higher risk of crash. To avoid ecological
fallacy, it is admirable to adopt disaggregate measures if
possible.

Fig. 1 Example diagram of ecological fallacy

In this paper, individual speed reduction, and the 85th


percentile value were calculated based on the data collected.
The 85th percentile value is the speed reduction range that
85% vehicles do not exceed. The relationship between speed,
speed reduction and geometric design were modeled. Safety
performance for different successive elements at 37 sites were
evaluated based on the criterion of V85 , 85( V ), and
85( V )R.

Relationship between V85 and 85(V)


The measure V85 is the subtraction of operating speed on

two successive elements. This measure assumes independent


sampling. The speed samples collected at the elements are
drawn from independent population that follows a normal
distribution. Based on this assumption, the expression of
V85 is:
V85 = V1,|0.85| V2, |85| = ( 1 2 ) + Z|0.85| ( 1 2 )
(1)
where V1,|0.85|, V1,|0.85| = 85th percentile speed on element 1 and
2; 1, 2 = mean speed of element 1 and 2; Z|0.85|= the 85th
percentile value of the standardized normal distribution; 1,
2= standard speed deviation of element 1 and 2.
The steadily accepted measure 85( V ) is quite different
from V85 in that 85( V ) first does the subtraction based on
paired individual vehicle speed, and then calculates the 85th
percentile value from the subtraction set. The static 85th
percentile value is the speed differential that 85% vehicles do
not exceed. This measure allows dependency between speeds

GUO Tangyi et al. / J Transpn Sys Eng & IT, 2010, 10(6), 7681

of an individual vehicle over two successive elements.


85( V ) can be expressed as:
85(V ) = (V1 V2 )|0.85|
= ( 1 2 ) + Z|0.85|

(2)

12 + 2 2 2 cov(V1 ,V2 )

where cov(V1, V2)= the covariance term that reflects the


correlation between V1 and V2. Other variables have the same
meaning as in Eq. (1).
Define as the correlation coefficient between V1 and V2.
possesses the following property:
| =

cov(V1 , V2 )

1 2

| 1 or 1 2 cov(V1 , V2 ) 1 2

(3)

If V1 and V2 are independent, then =0; if V1 and V2 are


positively (negatively) correlated, then =1 (-1). According to
Eq. (3),
2 1 2 2 cov(V1 , V2 ) 2 1 2
12 + 2 2 2 1 2 12 + 2 2 2cov(V1 ,V2 )
( 1 2 ) 12 + 2 2 2 cov(V1 ,V2 )
V8585( V )
The fundamental difference between V85 and 85( V ) is
that 85( V ) rather than V85 allows dependency between
V1 and V2 over two elements. In reality, the speed distribution
is not necessarily normally distributed. Speed varies from
tangent to curve due to different alignments, and varies from
one drivers habit to anothers due to their diverging personal
characteristics (e.g. age, gender, trip purpose, fatigue). Last
but not least, vehicle attributes (make, model, delivery
capacity) have noticeable effect on operating speed. Thus, the
85th percentile driver at the upper element is not necessarily
the same as the 85th percentile driver at the following element.
So a simple subtraction of the two operating speeds is
arbitrary. As alternatives, McFadden et al. (2000) suggested
85MSR and P. Misaghi et al. (2005) recommended 85( V ) to
evaluate the design consistency[10]. Previous researches mainly
focus on two-lane rural highways, rare has been found for
freeway exit ramps. In this paper, we introduced a new
measure, the individual speed reduction rate 85( V )R, to
appraise the safety performance of freeway exit ramp. This
measure considers not only the individual speed and aggregate
speed, but also the base speed on upstream element.

Data Collection

3.1 Exiting Behavior Analysis


The traffic stream approaching freeway exit becomes two,
one driving straight along the mainline and the other leaving
the mainline through the exit. The speed of straight flow
decreases slightly near the deceleration lane and then
increases to their desired driving speed after they pass the exit
taper. The speed of exiting flow begins to decrease when they
see the exit sign. Drivers may continue to slow down to adjust
to the change of curvature and super elevation on ramps. The
speed changing profile is illustrated in Fig. 2. The entire

exiting process can be divided into five stages.


In stage 1, drivers drive at their desired operating speed V0
on freeway.
At the end of stage 1 and the beginning of stage 2, exiting
drivers see the exit sign or the deceleration lane and then
begin to take lane-changing action to move onto the
deceleration lane. During this process, drivers may reduce
their speed to obtain a safe and comfortable driving workload.
In stage 3, the speed of exiting traffic decreases steadily
from V0 to V1, the speed at the beginning of ramp.
In stage 4, speed continues to decrease and reaches the
speed at the middle of the curve.
In the last stage, drivers adjust their speeds to coordinate
with speed V2 on crossing roads.
3.2 Investigation Plan
The study area is divided into four segments: freeway
section, deceleration lane, upper ramp, lower ramp. The
procedure for collecting individual point speed is designed as
follows:
(1) Pre-investigation: search the candidate exits on Google
Earth. Remove the sites that do not meet the criteria. Shot the
individual speed of exiting vehicles with radar guns at the
locations form to , as shown in Fig. 2.
(2) Observer assignment: assign two observers at each
location. The radar-gun holder is in charge of taking speed,
remembering the last three digits of the plate number, and
informing his partner of these information immediately. The
observers are hidden to minimize the effect of their presence
on passing vehicles.
(3) Record match: The last three digits are for record match.
The record is valid if and only if the observation that has the
same three digits recorded at the four locations. Otherwise, it
is a mismatch, which should be discarded.
37 freeway exits were selected for the data collection,
through which we obtained both the individual speed and the
operating speed for each segment. After removing 766
mismatch record, a total of 6 972 individual speeds were
collected, 189 for each site on an average. Meanwhile, 829
records of heavy vehicles, 22 for each exit, were also
collected.

Model

Totally 37 exit ramps were investigated, among which 30


are for modeling and the remaining 7 for validation. Site
selection and model development are based on the following
assumption: (1) the state of exiting traffic is free-flow or
pseudo-free-flow, among which the interference is minor; (2)
The sight condition is good, so as not to cause abrupt driving
behavior; (3) Roads are asphalt paved and dry, with a grade
less than 5%; (4) only one lane on ramp; and (5) the
deceleration type is direct as shown in Fig. 2, without
paralleled type.

GUO Tangyi et al. / J Transpn Sys Eng & IT, 2010, 10(6), 7681

I, II, III, IV, V: Driving stages for exiting traffic

: Speed profile for exiting traffic

Speed profile for through traffic on mainline

, , , : Speed taking point

Fig. 2 Diagram of driving stages and speed profile during exiting


Table 2 Models of V 85 and 85( V ) on different elements
Predictor
Operating
speed V85
Speed differential
85(V)

Model

R2

Variable explanation

Decel. Lane: V85_0=-62.642+2.021V85_FW-0.215LDL+0.073SLRP

0.993

V85_FW: V85 on mainline

Upper Ramp: V85_1=56.735-1312.095/R1

0.946

LDL: length of deceleration lane

Lower Ramp: V85_2=53.590-1174.699/R2

0.867

SLRP: post speed on ramp

FM-DL: 85(V) _0=61.597-1.032V85_FW+0.245LDL

0.884

R1: radius of upper ramp

DL-UR: 85(V) _1=24.085-0.010R1-1.921D1

0.845

R2: radius of lower ramp

UR-LR: 85(V) _2=5.092+1578.647|1/R1-1/R2|

0.779

D1:deflection of upper ramp

The operating speed differential can be obtained by simply


subtract the operating speed on successive elements. Taking
ramp length, ramp radius and base speed on upstream element
into consideration, the models for operating speed, and the
85th percentile individual speed reduction are developed as in
Table 2.
According to our survey, the operating speed on freeway
mainline is about 10 km/h higher than the post speed. So the
base speed on freeway mainline can be replaced by 1.10 times
of post speed. The models in Table 2 are validated using the
data of the remaining 7 exit ramps. Results show that the
model fit the data well under the confidence level of 90%.

Safety Evaluation

The speed reduction on the linkage of deceleration


lane-upper ramp is larger than that on the linkage of upper
ramp-lower ramp. However, this is not a necessary indication
that the latter linkage is safer, as the base speed on the
upstream element is different. Take the following case for
instance. Case A: Speed reduces from 60 km/h on deceleration
lane to 45 km/h on ramp. Case B: On an urban arterial, speed
reduces from 40 km/h on a tangent to 25 km/h on a curve. In
both cases, the absolute speed reductions are the same (15
km/h). The speed reduced by 25% in case A compared with

37.5% in case B. One can see that drivers on the arterial need
to take more driving workload to reach their desired comfort
and safety. From this point of view, Case B may be dangerous.
To consider base speed, we use the 85th percentile speed
reduction rate 85(V)R as safety evaluation measure. 85(V)R
is computed dividing the 85th percentile speed reduction
85(V) by the base speed on upstream element. The criterion
is given in Table 3.
Table 3 Safety evaluation criterion based speed reduction ratio
Measures

Criterion

Safety level

(V 85-1-V 85-2)/ V 85-1


(V85-VD)/V85
(V1-V2)/V1

20%

Good

20~40 %

Fair

40 %

Poor

According to Tables 1 and 3, we can appraise the safety


level of the 37 exits by means of V85, 85(V) and 85(V)R.
To show all the results of the three measures in one figure, we
halved the value of 85(V)R. Take the linkage of freeway
mainline-deceleration lane as an example. The safety level
evaluated using three different measures are distributed in
three safety zones, namely, Safe Zone, Fair Zone and Poor
Zone, as shown in Fig. 3. It also shows the relationship
between V85 and 85(V), which indicating the overestimation
of conventional measure V85.

GUO Tangyi et al. / J Transpn Sys Eng & IT, 2010, 10(6), 7681

Fig. 3 Safety zone distribution (freeway mainline-deceleration lane


linkage)

From Fig. 3, and our data analysis, we can draw the


following conclusions: (1) Two lines, 10 km/h (20%) and 20
km/h (40%) divide the zone into three levels, Safe Zone, Fair
Zone and Poor Zone. (2) There are 20 consistent evaluation

results from the three different measures for freeway


mainline-deceleration lane linkage (the vertical lines in Fig. 3).
They are site 3, 4, 5, 9, 10, 11, 13, 15, 18, 19, 20, 23, 27, 28,
29, 31, 33, 34, 36 and 37. There are 16 and 17 consistent
results for deceleration lane - upper ramp linkage and upper
ramp lower ramp linkage, respectively. (3) The fact that the
line of 85(V) is higher than V85 indicates the overestimation
of V85. The ratio of 85(V)/V85 is 1.42 for freeway mainline
- deceleration lane linkage. The value is 1.69 and 2.02 for
deceleration lane - upper ramp linkage and upper ramp
lower ramp linkage, respectively. (4) The safety level of a
target linkage can be determined if the results are consistently
located. (5) if there is one result located in Poor Zone, the
target exit should be classified as poor safety level. Evaluation
results for the other two linkages are listed in Table 4.

Table 4 Safety evaluation for different elements using three measures


Safety level

Mainline-deceleration lane

Deceleration -upper ramp

85(V)

85(V)R

V85

85(V)

85(V)R

V85

85(V)

85(V)R

Good

10

20

13

28

19

15

Fair

24

20

23

12

15

12

11

14

Poor

14

14

12

It can be seen from Fig. 3 and Table 4 that conventional


measure V85 has the most Good sites compared with the
other two measures. The result of 85(V)R is between the
other two. We can also see that with the decrease of speed,
speed reduction becomes smaller.

[2]

[3]

Conclusions

The 85th percentile value of individual speed reduction,


85(V), is introduced to evaluate the safety performance of
freeway exit ramp. The finding that 85(V)V85 implies there
is overestimation of safety appraisal using conventional
measure V85. Another measure, speed reduction rate
85(V)R, proposed in this paper considers the base speed on
upstream element, which leads to its superiority over other
measures. Point speeds of individual vehicle on deceleration
lane, upper ramp and lower ramp were collected using radar
guns at 37 exits in Nanjing, China. Models were built to
predict the operating speed, speed reduction and speed
reduction rate. Based on the proposed models and evaluation
criteria, the safety level of the three linkages is determined in
terms of zone distribution. The recommended measure
85(V)R avoids pitfalls of ecological fallacy and
overestimation possessed by conventional ones, and it can be
adopted in practice conveniently.

References
[1]

Upper ramp-lower ramp

V85

Global Road Safety Partnership. Speed management: a road


safety manual for decision-makers and practitioners. Geneva,
Global Road Safety Partnership, 2008.

Lamm R, Choueiri E M. Recommendations for evaluating


horizontal design consistency based on investigation in the
state of New York. Transportation Research Record 1122,
Washington D.C., 1987, 6878.
Lamm R, Choueiri E M, Mailaender T. Comparison of
operating speed on dry and wet pavement of two lane rural
highways. Transportation Research Record 1280, Washington

D.C., 1990, 199-207.


Islam M N, Seneviratne P N. Evaluation of design consistency
of two-lane highways. ITE Journal, 1994: 64(2), 2831.
[5] Krammes R A, Rao K S, Oh H. Highway geometric design
consistency evaluation software. Transportation Research
Record 1500, Washington D.C., 1995, 1924.
[6] McFadden J, Elefteriadou L. Formulation and validation of
operating speed-based models using bootstrapping.
Transportation Research Record 1579, Washington D.C., 1997,
97103.
[7] Leisch J E, Leisch J P. New concept in design speed
applications, as a control in achieving consitent highway
design. Transportation Research Record 631, Washington D.C.,
1977, 414.
[8] Guo T Y. Research on investigation methods based on
disaggregate model. School of Transportation, Nanjing, 2005.
[9] McFadden J, Elefteriadou L. Evaluating horizontal alignment
design consistency of two-lane rural highways: Development
of new procedure. Transportation Research Record 1737,
Washington D.C., 2000, 917.
[10] Misaghi P. Modelling operating speed and speed differential
for design consistency evaluation. M.Sc. Thesis, Department
of Civil and Environmental Engineering, Carleton University,
Ottawa, Ontario, Canada. 2003.
[4]

También podría gustarte