You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224230507

Estimating critical gap of roundabouts by different methods

Conference Paper · November 2010


DOI: 10.1049/cp.2010.1107 · Source: IEEE Xplore

CITATIONS READS
15 253

1 author:

Rui-Jun Guo
Dalian Jiaotong University
12 PUBLICATIONS 86 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rui-Jun Guo on 30 January 2020.

The user has requested enhancement of the downloaded file.


Estimating Critical Gap of Roundabouts by
Different Methods
1,2
Rui-jun Guo
1 School of Transportation Engineering, Dalian Jiaotong University, Dalian, China 116028; email: rjguo@163.com
2 School of Traffic and Transportation, Beijing Jiaotong University, Beijing, China 100044

rejected gap when the headway is smaller than critical


abstract gap. The critical gap usually follows a certain

Critical gap is an important parameter which affects the distribution but a definite value because of difference of

capacity and delay of unsignalized intersections. There driver operation, geometry of roundabouts and traffic
operation character.
are different critical gap values for drivers in different
geometry parameters and traffic conditions. Critical gap Many different methods estimating critical gap at
can be generally described by Erlang distribution, unsignalized intersections have been published in the
lognormal distribution, etc. Various methods international literature. Raff’s method (1950) involves
calculating critical gap at roundabouts have been the distributions of accepted gap and rejected gap. The
presented. Many studies on gap acceptance at sum of cumulative probabilities of accepted gap and
roundabouts have been carried out in America and rejected gap is 1 when the headway is equal to critical
Europe, whereas the same works have rarely been done gap. Critical gap can be obtained by drawing graph of
in China. This paper gives an overview of many cumulative probability. This method has been applied in
methods calculating critical gap. On the base of video many countries, whereas it lacks strict theoretic basis.
survey of Shuma Square roundabout in Dalian, we Robert Ashworth’s method (1969) is related to the
acquired a large number of rejected gaps and accepted distribution of accepted gap under the circumstances of
gaps in rush hour. The critical gaps can be calculated by different flow rates of circulating stream provided that
use of different methods including Ashworth’s method, the circulating headway follows exponential
Raff’s method and the maximum likelihood method. As distribution, critical gap and accepted gap follows
a result, the maximum likelihood procedure and the normal distribution. Miller (1974) deduced the formula
revised Raff’s method can be recommended for of critical gap provided that it follows Gamma
practical application, and Ashworth’s method will distribution based on Ashworth’s method.
result in larger result because it only uses values of M.M.Hammed et al. (1997) consider the distribution of
accepted gap. critical gap is related to the life background of driver,
waiting time and trip purpose etc. and the mean critical
1 Instruction gap is related to conflicting flow, the lane number of
entry, the proportion of left-turn lane and velocity of
Critical gap is an important parameter which influences
circulating stream. They built multiple regress model of
the capacity and delay of roundabouts, and it is usually
critical gap. The maximum likelihood method (Tian et
defined as a threshold whether drivers in entry vehicles
al. 1999) supposed the accepted gap, maximum rejected
run into the roundabout or not. If the headway between
gap and critical gap follow normal distribution, and the
the continued two vehicles of circulating stream is
mean and variance of critical gap can be calculated
larger than the critical gap, it is named as accepted gap
based on probability theory. Brilon et al. (1999) had a
since entry vehicles can accept the headway and run
comprehensive analysis on different methods and made
into roundabout. On the other hand, it is named as
a conclusion that maximum likelihood method (MLM)
and Hewitt’s method calculate well and truly. Abishai
Polus et al. (2003) considered critical gap decreased
with the increase of waiting time of entry drivers and
2
built an exponential model. where  c is the variance of critical gap(s2)。 The
Based on the video survey of Shuma Square computer iteration can be applied to calculate critical
roundabout in Dalian, China, we calculated and gap using equation (3). We calculate critical gap by use
compared critical gap by use of varied methods based of equation (2) in this paper.
on the accepted gaps and rejected gaps of a certain 2.3 Maximum likelihood method
waving section during the rush time. Troutbeck and Brilon (1999) considered the MLM as
the most reliable of varied methods. Troutbeck
2 Calculation method assumed critical gap followed lognormal distribution,
while Brilon assumed critical gap followed ultra-Erlang
The critical gap is a hypothetical value which can’t be
distribution. The paper uses the former.
measured based on the field samples, it can be only
estimated by measuring the accepted gap and rejected MLM uses two terms of gaps including accepted gaps
gap. There are many methods to estimate critical gap and maximum rejected gaps. The maximum rejected
such as regression method, MLM, Raff’s, Harder’s, gap is the maximum value of all rejected gaps during a
Ashworth’s, Siegloch’s and so on.
driver waits to running into the roundabout. The mean
2.1 Raff’s method
and variance of critical gap can be calculated used of
Raff consider the number of rejected gap larger than
the maximum likelihood function of probability theory.
critical gap is equal to the number of accepted gap
smaller than critical gap. He created the formula as We assume critical gap follows lognormal distribution,
that is to say, t ~ LN(u, ) . The maximum
2
follow. c

likelihood function is as follow.


n
(4)
[F (ai )  F (ri )]
i 1

1 Fr (t)  Fa (t) (1) where ai is the logarithm of the gap accepted by the
i th driver; ri is the logarithm of the maximum gap
where Fa (t) is the cumulative proportion of
rejected by the i th driver, r = 0 if no gap was
i
accepted gap; Fr (t) is the cumulative proportion of
rejected ; f (t ) , F(t) are the probability density
rejected gap; t is the headway of two continued
function and cumulative distribution function for the
vehicles of circulating stream.
normal distribution respectively.
The critical gap is the cross point of two cumulative
The logarithm of function (4) is as follow.
proportion curves of accepted gap and rejected gap.
n

2.2 Ashworth’s method L  ln[ F (ai )  F (ri )] (5)


i 1
Ashworth gave the calculative formula as follow.
The likelihood estimators  and 
2
are the solutions
t t V  2 (2)
c a p a
that the partial derivative of equation (5) is equal to
where tc is critical gap(s); Vp is the flow rate of zero. They can be simplified as follow.
n
f (ri )  f (ai )

circulating stream(pcu/s);
ta
is the mean of accepted   0 (a)
 i1 F (ai )  F (ri )
2
gaps(s);  a is the variance of accepted gaps(s2)。  n (r  ˆ ) f (r )  (a  ˆ ) f (a )
 i i i i

Similarly, Miller gave the calculative equation (3) 



i1 F (a )  F (r )  0(b)

 i i

based on the hypothesis that critical gap follows gamma


(6)
distribution.

t t V 2 Where f (ri ) , f (ai ) , F (ri ) , F (ai ) are all the
c a p c function of  and  2 .  and  2 can be calculated
 t (3)
c
 
 c a by programming computer (Troutbeck, 1992). So the
 ta
mean and variance of critical gap can be derived by

conflicting stream considering the inner two lanes of the


equation (7). waving section as a lane. After video survey of continued
t  eu0.5 2 40 minutes from 16 :30-17:10 on April 28, 2009, the
c maximum rejected gaps and accepted gaps including 100
2 terms of data which are revealed in the table 1. The flow
2 2
rate of circulating stream is 1842pcu/h,or 0.512 pcu/s.
s  tc (e 1)

3 Measure of accepted gap and maximum


rejected gap
Shuma Square roundabout is an unsignalized
intersection located the cross of Shuma Road, Youjia
Road and Wuyi Road. The diameter is 62 meters and
there are two lanes in entry. The survey section is the
waving section which length is 45 meters between east
Wuyi Road and Youjia Road as figure 1. The section is
constituted by 3 lanes in which the outer lane is right-
turn lane and the other two lanes are the waving section
where confluence operations occur.

Figure 1. Survey section in Shuma Square roundabout

The accepted gaps and rejected gaps are measured with


the headways between the continued vehicles in
Table 1. The measure of accepted gap and the
maximum rejected gap(s)
ta tr ta tr ta tr ta tr ta tr ta tr
5.26 1.68 3.29 0 3.86 0 4.11 0 2.31 0 1.8 1.43

3.46 1.16 3.35 1.93 2.97 0 5.21 3.01 5.11 2.45 3.22 0

4.76 2.08 4.22 2.44 3.49 0 3.14 0 3.69 0 5.21 2.18

1.97 0 4.06 2.44 2.48 0 3.54 0 4.99 3 2.46 1.68

5.75 0 4.01 1.93 3.78 1.92 2.01 0 5.54 2.36 3.67 1.93

2.72 0 3.32 1.68 2.89 2.61 6 0 4.21 0 2.59 0

5.05 0 2.89 0 2.59 2.41 4.51 0 3.32 0 2.94 0

4.26 1.55 3.13 0 3.06 1.4 3.33 0 3.68 2 6.82 4.06

3.22 0 4.66 3.35 3.97 2.77 3.56 2.24 3.96 2.1 7.24 2.18

2.18 1.16 4.26 2.94 4.22 2.18 4.25 2.14 5.25 1.89 3.49 2.32

3.49 0 4.22 3.6 3.6 2.69 5.21 2.54 5.47 2.58 4.1 2.72

3.74 2.72 5.26 2.34 3.13 1.39 3.45 0 4.12 2.47 5.16 2.56

2.72 0 2.86 0 4.53 1.93 2.88 2.01 2.36 0 2.32 2.08

1.93 0 3.78 0 5.05 2.16 3.23 0 2.98 0 3.42 2.03

2.18 1.42 2.56 0 4.42 0 4.1 0 3.56 0 4.21 1.56

4.12 0 3.55 2.11 3.03 1.48 2.64 0 4.25 2.45 5.21 2.22

4 2.63 2.88 0 5.5 3.12 4.26 2.41

4 Calculation and compare of critical gap

4.1 Raff’s method

Raff’s method does not consider the maximum


rejected gap equal to zero, so 58 rejected gaps can be
derived from table 1. The accumulative proportions
of accepted gap and rejected gap are showed as table
2.

Table 2. Accumulative probability of accepted gap and


rejected gap
Accumulative Accumulative
Range of Acceptance Rejected Acceptance Rejected
probability of probability of
gap(s) number number probability probability
acceptance rejection

0-1.1 0 0 0 0 0 1

1.1-1.4 0 3 0 0 0.05 1

1.4-1.7 0 9 0 0 0.16 0.95

1.7-2.0 3 7 0.03 0.03 0.12 0.79

2.0-2.3 3 13 0.03 0.06 0.22 0.67

2.3-2.6 8 13 0.08 0.14 0.22 0.45

2.6-2.9 8 6 0.08 0.22 0.10 0.22

2.9-3.2 8 4 0.08 0.3 0.07 0.12

3.2-3.5 14 1 0.14 0.44 0.02 0.05

3.5-3.8 11 1 0.11 0.55 0.02 0.03

3.8-4.1 8 1 0.08 0.63 0.02 0.02 gaps from table 1 in six ranges.
4.1-4.4 13 0 0.13 0.76 0 0

4.4-4.7 4 0 0.04 0.8 0 0

4.7-5.0 2 0 0.02 0.82 0 0

5.0-5.3 11 0 0.11 0.93 0 0

5.3-5.6 3 0 0.03 0.96 0 0

5.6-5.9 1 0 0.01 0.97 0 0

5.9-6.2 1 0 0.01 0.98 0 0

6.2-6.5 0 0 0 0.98 0 0

6.5-6.8 0 0 0 0.98 0 0

6.8-7.1 1 0 0.01 0.99 0 0 Figure 3. Distribution of accepted gap in Shuma Square


7.1-7.4 1 0 0.01 1 0 0 roundabout
Total 100 58 1 1

2
Table 3.  test on normal distribution of accepted
The accumulative probability of accepted gap is the
probability which the accepted gap is smaller than the gap
upper limit of range, for example, the accumulative
Range(s) vi pi npi vi2 / npi
probability of accepted gap, 0.14 at the range of 2.3-
2.6s refers to the probability which accepted gap is 0-2 3 0.048 4.824 1.866

smaller than 2.6s. On the contrast, the accumulative 2-3 22 0.182 18.211 26.577

probability of rejected gap is the probability larger than 3-4 34 0.344 34.374 33.630

the lower limit of range. 4-5 23 0.293 29.269 18.074

The curves of accumulative probability can be drawn 5-6 16 0.112 11.234 22.789

as figure 2. The critical gap is 2.91s which is the cross >6 2 0.021 2.089 1.915

of acceptance curve and rejected curve. Total 100 1 100 104.851

Figure 2. Critical gap of Raff’s method

We call the method which uses all rejected gaps


including zero as revised Raff’s method. It considers
the vehicles immediately running into the roundabout
without waiting. The critical gap is 2.78s by use of
revised Raff’s method.
4.2 Ashworth’s method The progress of  2 test is as table 3. where v
i

Test of hypothesis to normal distribution of accepted is the frequency of accepted gaps in each range; n is
gap. The figure 3 shows the frequencies of accepted the total frequency and n 100 , pi is the
2
probability of distribution in the range and
2
 104.851-100  4.851. The equation can be
2
derived in search of  distribution table as the

significance level is 0.05 and free degree is 3 as follow.

 2
0 .953  7.815  4.851
So the hypothesis can be accepted and the accepted gap
  0.9485 follows normal distribution.

4.2.1 Test of hypothesis to exponential distribution


of circulating stream.

The progress of test is as table 4.


circulating stream vehicles follows exponential
distribution and the accepted gap follows normal
 2 test on exponential
Table 4. distribution distribution. The critical gap is 3.20s by use of
equation (2) of Ashworth’s method.
n 4.3 Maximum likelihood method
No. Range(s) v Frequence pi np v2 p
i i i
We calculate the probability density and accumulative
i
probability by use of the NORMDIST function in
1 0-1.5 0.143 15 0.083 8.730 25.773 Microsoft Office Excel2007. Firstly the mean, 1.29s
2 1.5-2 0.143 15 0.165 17.283 13.019 and the standard variance, 0.29s of the logarithm of
3 2.5-3 0.248 26 0.246 25.815 26.187 accepted gaps are considered as the preliminary values
4 3-3.5 0.086 9 0.091 9.546 8.485 of  and  . Then the values of  and  are changed
5 3.5-4 0.048 5 0.075 7.832 3.192 step by step in order that left functions of (a) and (b) in
6 4-4.5 0.038 4 0.061 6.426 2.490 equation (6) tend to zero. Finally the values of  and
7 4.5-5 0.057 6 0.050 5.272 6.828
 are confirmed if the error is in the range of a little
8 5-6 0.057 6 0.075 7.875 4.571
value. The progress of calculation is as table 5.
9 6-7.5 0.057 6 0.069 7.262 4.958

10 7.5-9 0.057 6 0.038 4.011 8.976

11 >9 0.067 7 0.047 4.948 9.902

合计 1 105 1 105.000 114.381

2
 114.381-105  9.381
 0
2
.95(8) 15.507  9.381.

So the hypothesis can be accepted and the headway


follows exponential distribution.
4.2.2 Calculation of critical gap
We can calculate the mean accepted gap, 3.8s and the
variance, 1.17 s2 by use of the data in table 1. Based
on the above conclusion that the headway of
Critical
Table 5. The iteration of MLM 2.65 3.2 2.91 2.78
  Left (a) Left (b) gap
1.29 0.29 -270.695 40.143
1 0.29 -45.832 -16.940
The order of different method’s results is Ashworth’s,
1 0.23 -52.610 -1.100
Raff’s, revised Raff’s and MLM. Ashworth’s result is
0.93 0.23 17.838 -0.056
the largest and other methods have a little difference
0.93 0.28 5.892 -13.631
because Ashworth’s method only uses accepted gap,
0.95 0.23 -2.061 -1.097
but other methods use rejected gap too. The revised
0.945 0.246 -0.020 -5.621
Raff’s method considers the rejected gap equal to zero,
0.949 0.227 -0.552 -0.114
so its result is less than Raff’s result and close to
0.9485 0.2268 -0.008 -0.030
MLM’s result. The revised Raff’s method and MLM is
more reliable than others because both accepted gap
When and   0.2268 ,left and rejected gap influence critical gap in the field
(a) traffic conditions.
and (b) tend to zero. The mean and variance of critical MLM is a pretty method Calculating critical gap of
gap is respectively 2.65s and 0.37s2 using equation confluence operation. It uses the accepted gap and
(7). rejected gap independently in spite of condition of
4.4 Compare of varied method entry stream and survey time. Ashworth’s method is
We can acquire the table 6 from the above calculations. simple to calculate critical gap but it should be assured
that the headway of circulating stream follows
Table 6. Critical gap of varied methods(s) exponential distribution and the accepted gap follows
Revised normal distribution. The distribution condition is
method MLM Ashworth’s Raff’s difficult to satisfy sometimes because the vehicle
Raff’s streams are influenced by the upriver signalized
intersections or the
[1] Abishai Polus, M.ASCE, Sitvanit Shmueli
flow rate is small in peacetime. So Ashworth’s method Lazar, Moshe Livneh, M.ASCE. (2003). “Critical Gap
demands strict adaptable condition in fact. Raff’s and as a Function of Waiting Time in Determining
revised Raff’s method are simple, intuitive and can get Roundabout Capacity”. Journal of Transportation
reliable result. It can be as a preliminary value of Engineering, 129(5), 504–509
MLM under the condition of a large size sample to [2] M.M.Hammed, S.M.Easa, Member, ASCE,
find the final result quickly. R.R.Batayneh. (1997). “Disaggregate gap-acceptance
Conclusion model for unsignalized T-intersections.” Journal of
It is assumed that critical gap follows a certain
distribution. The result values are usually different
using different methods for the same headway data. It
reveals revised Raff’s method and MLM are reliable.
Ashworth’s result is larger than others as it only uses
accepted gap. With the change of society and city
traffic, critical gap, in addition to the other parameters
such as follow-up headway and minimum headway,
changes too. The further researches are needed to
capture the change regular and build dynamic model
of critical gap based on a large number of real-time
samples so as to make use of roundabouts effectively.
reference
[7] Transportation Research Board. (2000). Highway
Transportation Engineering, 123(1), 504–509 capacity manual 2000. National Research Council,
[3] Robert Ashworth. (1969). “The analysis and Washington D. C., USA.
interpretation of gap acceptance data”. The first annual [8] Zongzhong Tian, Mark Vandehey, Bruce W.
conference of the British Universities transport study Robinson, Wayne Kittelson, Michael Kyte, Rod
group, University of Leeds Troutbeck, Werner Brilon, Ning Wu. (1999).
[4] Miller, Alan J. (1974). “A Note on the Analysis of “Implementing the maximum likelihood methodology
Gap--acceptance in Traffic.” Journal of the Royal to measure a driver's critical gap.” Transportation
Statistical Society: Series C (Applied Statistics), 23(1), Research Part A, 33,187~197
66~73 [9] Troutbeck, R.J. (1992). “Estimating the Critical
[5] Raff, M. S. and Hart, J. W.. (1950). “A volume Acceptance Gap from Traffic Movements.” Physical
warrant for urban stop signs.” Eno foundation for Infrastructure Center Report. Queensland University
highway traffic control: Saugatuck, Connecticut. of Technology, Australia.
[6] Werner Brilon, Ralph Koenig, Rod J. Troutbeck.
(1999). “Useful estimation procedures for critical gaps.”
Transportation Research Part A, 33, 161~186

View publication stats

You might also like