View
226
Download
0
Category
Preview:
DESCRIPTION
kelompok makalah
Citation preview
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 1/59
Chapter 5
Link Travel Time: Simulation Analysis
5.1 Simulation description
To verify the qualitative analysis proposed in the previous section, a simple network
is modeled using VISSI ! a microscopic traffic simulation model "y #TV $%
&http'((www.english.ptv.de(). The imaginary time period is morning and the imaginary
network is a corridor towards city center.
The network is illustrated in *igure 5!1. +ach intersection is controlled "y a two!
phase fied!time signal with a cycle length of 1- s. The offset time of two consequent
intersections is / s. The signal scheme is shown in *igure 5!1. Turning rate at each
intersection is also shown in the figure.
The network is simulated for / h with time varying demands. Vehicles are generated
at origin 0ones &112, see *igure 5!1) with #oisson distri"ution. Ta"le 5!1 shows the
"ase vehicle inputs of each 0one and the temporal variation.
5/
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 2/59
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 3/59
Table 5-1. ehi!le inputs
3ase Vehicle Inputs &Veh(h) Temporal Variation
4rigin one Vehicle Inputs Time &s) #roportion of 3aseVehicle Inputs
1 1 6 .1
211 7 6 17 .2
12 5 17 28 .5
28 /9 1
/9 -5 1.2
-5 9/ 1.5
9/ 82 1
82 71 .7
71 17 .5
In the network, only the links from link 2 to link 5 are interested. *igure 5!1 shows
the data collection points &$ *), which are located at the reference points of the
entrance and eit of link. :hen a vehicle pass through any data collection points, the
vehicle I;, pass time, and spot speed of the vehicle are collected and recorded into a
log file. Then, the num"er of accesses &only through movement), mean spot speed at
upstream &only through movement), and mean travel time &only TT) of the interested
links are aggregated in every signal cycle using the log file.
5.2 Simulation result
*igure 5!2 shows the aggregated result. The num"er of accesses &only through
movement), mean spot speed at upstream &only through movement, unit' m(s), and
mean travel time &only TT) in each signal cycle are illustrated "y three solid lines. The
figure also shows the individual travel time reports &<) and the temporal variation of
vehicle inputs &dot line).
55
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 4/59
.No of Accesses
60 50 40 30 20 10 0
t r a v e l t i m e a c c e s s e s s p o t s p e e d
9 9 0 0
1 0 8 0 0
1 5 . 1
m e a n n o . o
f m e a n
9 0 0 0
7 2 0 0
8 1 0 0
r e s u l t " l i n k # $
4 5 0 0
5 4 0 0
6 3 0 0
S i m u l a t i o
n i m e ! s "
% l l u s t r a t i o
n o & s i m u l a t i o n
5 9
3 6 0 0
5 - # " a $ .
1 5 . 1
2 7 0 0
F i g u r e
1 3 .
3
1 8 0 0
9 0 0
0
160 140 120 100 80 60 40
ravel ime !s"
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 5/59
.No of Accesses
60 50 40 30 20 10 0
t r a v e l t i m e a c c e s s e s s p o t s p e e d
9 9 0 0
1 0 8 0 0
1 3 . 7
m e a n n o . o
f m e a n
7 2 0 0
8 1 0 0
9 0 0 0
r e s u l t " l i n k ' $
4 5 0 0
5 4 0 0
6 3 0 0
S i m u l a t i o
n i m e ! s "
% l l u s t r a t i o n o & s i m u l a t i o n
5 8
3 6 0 0
5 - # " b $ .
2 7 0 0
F i g u r e
1 8 0 0
1 5 . 1
9 0 0
0
120 100 80 60 40
ravel ime !s"
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 6/59
.No of Accesses
60 50 40 30 20 10 0
1 4 .
6
t r a v e l t i m e a c c e s s e s s p o t s p e e d
9 9 0 0
1 0 8 0 0
m e a n n o . o
f m e a n
9 0 0 0
9 . 4
7 2 0 0
8 1 0 0
r e s u l t " l i n k (
$
4 5 0 0 5 4 0 0
6 3 0 0
S i m u l a t i o n i m e ! s "
% l l u s t r a t i o n o & s i m u l a t i o n
5 7
3 6 0 0
5 - # " ! $ .
1 5 . 3
2 7 0 0
F i g u r e
1 3 .
1
1 8 0 0
9 0 0
0
200 150 100 50
ravel ime !s"
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 7/59
.No of Accesses
60 50 40 30 20 10 0
t r a v e l t i m e a c c e s s e s s p o t s p e e d
9 9 0 0
1 0 8 0 0
1 4 . 8
m e a n n o . o
f m e a n
9 0 0 0
1 0 .
2
6 .
2
4 5 0 0 5 4 0 0
6 3 0 0
7 2 0 0
8 1 0 0
S i m u l a t i o n i m e ! s "
% l l u s t r a t i o n o & s i m u l a t i o n r e s u l t " l i n k 5 $
5 6
3 6 0 0
5 - # " ) $ .
2 7 0 0
F i g u r e
1 3 .
4
1 8 0 0
9 0 0
0
200 150 100 50
ravel ime !s"
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 8/59
+ach intersection in the simulation network has same geometric properties and thus
the capacity of each intersection should "e identical. *rom *igure 5!2d, it can "e
concluded that the capacity is a"out 55 vehicle(cycle.
$s shown in *igure 5!2, when the num"er of accesses is lower than the capacity of
upstream intersection, mean travel time on link / ehi"its large variation while mean
travel times on other links are relatively sta"le. That is, though the pro"a"ility of =ase
II in #hase I is small, it still happened. In this simulation, the offset time of two
consequent intersections is set as / s and the length of the links is 5 m. Thus, when
a vehicle>s speed eceeds 19.8 m(s &9 km(h), it is possi"le that the vehicle will pass
through a link within / s and does not stop at the downstream intersection of the link
even if it through the upstream intersection of the link in the end of the green period or
in the am"er period. ?owever, when the offset time is set smaller than /s, for eample
2 s, the vehicle will stop at the downstream intersection, and the pro"a"ility of =ase II
in #hase I will increase and mean travel time will ehi"it more variation. ore
importantly, mean travel time at all links cannot trace the change of the num"er of
accesses in #hase I. That is, in #hase I, mean travel time is not a good indicator of link
performance and the num"er of accesses should "e used for tracing the change of link
performance.
$t a"out 5- s on link - and link 5, the spot speed start to decrease. The decrease
of the spot speed at upstream of a link will increase air pollution and it can "e
considered as undesira"le situation. $t same time, as shown in *igure 5!2d, the num"er
of the accesses is not reduced o"viously. =hoosing the spot speed or the num"er of
accesses as the criterion of the undesira"le situation should "e determined "ased on the
researchers> @udgment. $t this phase, mean travel time is o"viously larger than other
phase and the distri"ution of travel time tends to one peak. Therefore, it is easy to
9
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 9/59
identify the phase using small si0e #V reports. ?owever, as mentioned earlier, the
phase is undesira"le situation and should "e avoided.
To avoid the a"ove phase, it is important to identify the previous phase &#hase II) of
the a"ove phase. $s shown in *igure 5!2c and *igure 5!2d, in #hase II, mean travel
time increases gradually. In this phase, there are two sets in travel time reports' one
without intersection delay and with intersection delay. The proportion of the set with
delay increases and the means of the two sets also increase over time. It causes the
increase of mean travel time. Asing small si0e #V reports, it is hardly epected that the
proportion of the set with delay &or without delay) can "e estimated. ?owever, the
variance of each set is relatively small and the means of the two sets can "e estimated
"y small si0e sample. This is the key idea of the estimation method proposed in
=hapter 8.
5./ Summary of this chapter
In this chapter, the microscopic traffic simulation model VISSI is used to confirm
the qualitative analysis proposed in the previous section.
The num"er of accesses, mean spot speed at upstream &only through movement,
unit' m(s), and mean travel time &only TT) in each signal cycle are aggregated rather
than mean travel time only. The aggregation shows that &1) in #hase I, mean travel time
is not a good indicator of link performance and the num"er of accesses should "e used
for tracing the change of link performance, and &2) in #hase II, though mean travel
time can trace the link performance, individual travel time reports show that the link
travel time distri"ution is two!peak, which makes the estimation of mean travel time "y
small si0e travel time reports from #Vs difficult.
91
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 10/59
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 11/59
Chapter *
Appli!ability o& the Formulations o& the A)e+uate ,umber
The success of pro"e!"ased $TIS highly depends on the relia"ility of pro"e reports,
and the cost and capacity of the communication "etween #Vs and the operation center
impose restrictions on the num"er of #Vs. Therefore, it is impossi"le to make all
vehicles as #Vs. In other words, a set of travel time reports from #Vs on a link in a time
interval is a sample. The adequate sample si0e required to estimate mean travel time of
all vehicles &population) relia"ly has "een an imperative issue since pro"e vehicle was
recogni0ed as a method to collect traffic information and there are many literatures, for
eample 3oyce et al. &1661), =hen and =hien &2), =heu et al. &22), ?ellinga and
*u &1666), Buiroga and 3ullock &1667) and Srinivasan and Covanis &1669).
Two methods standard deviation formulation and confidence interval method are
commonly accepted, and were used as the relia"ility criterion. It should "e noted that
these formulations are "ased on =entral Dimit TheoremE that is, the result is dou"tful
when the population is severely nonnormal and sample si0e is small. This chapter will
eamine the applica"ility of the two methods on a signali0ed link.
9/
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 12/59
9.1 Sample si0e formulations
$ num"er of researchers have investigated the adequate num"er of #Vs at network
level and at link level. $t link level, two methods standard deviation formulation &+q.
1) and confidence interval method &+q.2) were proposed and are commonly accepted,
and were also used as the relia"ility criterion for each links in network level studies. In
the two methods, the former is used for links with normal distri"ution and the latter is
used for links without normal distri"ution.
t 2
sα (2, n#1
&1)n $
ε
x #t α ( 2,n#1 s ( n %µ% x &t α ( 2,n#1 s ( n &2)
9.2 $ssumption of travel time distri"ution
4"viously, the percentage of Set I in the queue, a""reviated as p, is a critical factor
that affects the average travel time of the vehicles in the queue &*igure -!8). Thus p is
considered as a performance indicator in this chapter, and travel time distri"ution at a
certain performance level & p F 85G is chosen @ust as an eample) is inferred "ased on
*igure -!8 and historical data descri"ed in section -.-. In the figure, it is assumed that
the vehicles spread during green period with uniform. In reality, the situation is more
comple than the illustration in the figure. There are some vehicles with other turning
movements and some vehicles partly through the link in the queue. ?owever, the
effects of these vehicles are ignored.
9-
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 13/59
*or p F 85G, the distri"ution can "e approimated as a composite of two normal
distri"ution and the mean and standard deviation of Set I and II can "e determined
from the historical data.
The means of %roup I and II in historical data are considered as the means of Set I
and Set II in real!time data at p F 6 G. Ander uniform assumption, when p decreases
p , the mean of Set I and Set II will increase p G &G is length of green time).
=onsequently, the mean of Set I and Set II at p F 85 G can "e calculated "y adding
15G 'G to the mean of Set I and Set II at p F 6 G, respectively. In this section, G is
set as 9 s.
Anlike historical data, the day!to!day variation is not included in real!time data and
it is reasona"le to consider that the variation of each set in real!time is smaller than the
variation in historical data. $dditionally, the adequate sample si0e is a function of the
standard deviation of link travel time, and the standard deviation is mainly contri"uted
"y the fact that there are two sets simultaneously "ut rarely contri"uted "y the variation
in each set. Thus, the standard deviations of all sets at p F 85 G are specified as one!
half of the standard deviation of %roup I.
Ta"le 9!1 summari0es the mean and the standard deviation of all and each set at p F
85 G. 3ased on the assumption of normal distri"ution in each set, the distri"ution of
travel time on the study link at p F 85 G is o"tained.
Table *-1. Travel time )istribution at p 5 /
mean &s) sd
Set I &85G) 1/5 -.92
Set II &25G) 226 -.92
$ll 157 -1.12
95
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 14/59
9./ $pplica"ility
$ simulation method is provided as a reference method against which the
applica"ility of standard deviation formulation and confidence interval method is
eamined. In the three methods, 15 s is selected as the pre!specified permitted error
& ), which is a"out 1 G of mean at p F 85G.
$ sample with si0e 1 is drawn from the distri"ution at p F 85 G o"tained from
previous section and is considered as the population and further sampling with sample
si0e n are performed from the population.
The simulated method is straightforward. *or each sample si0e n, 1 samples are
taken and the percentage of accepted samples is calculated. $ccepted sample means
that the sample mean fall into &µ , µ ) & is population mean). The percentage
for sample si0e from 1 to - is shown in Heference column in Ta"le 9!2. The
percentage increases as sample si0e increase and achieves 65 G when n F 28.
*or standard deviation formulation &+q. 1), / samples with sample si0e n are taken
and the formulation is eamined &α .5 ). The accepted(re@ected, the percentage of Set
I, sample mean and sample standard deviation of each sample are shown in Ta"le 9!2.
In the ta"le, the accepted samples are highlighted "y gray "ackground. :hen n 28,
there are some accepted samples. The cause is the critical underestimation of standard
deviation. The underestimation leads the right side of +q. 1 to "ecome small and the
inequality to "e satisfied. The cause of the underestimation is the sampling error. $s
shown in the ta"le, p of accepted samples is much larger than the percentage of Set I of
population &85 G). In contrast, when n JF 28, there are lots of re@ected samples. The
cause of most re@ect is the overestimation of standard deviation, and the cause of the
overestimation is also sampling error. In summary, the standard deviation
99
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 15/59
formulation is sensitive with sampling error and cannot provide consistent result.
*or confidence interval method , one sample is taken for each n and the
accepted(re@ected, the percentage of Set I, range of confidence interval, lower!
confidence "ound and upper!confidence "ound are calculated &Ta"le 9!2). $s shown in
the ta"le, the criterion that is the 1&1 α ) percent confidence interval calculated from
samples contains the population mean is satisfied in most n &α (.5 ). ?owever, when
n is small, the range of confidence interval is etremely large. It indicates that the a"ove
criterion is not enough to provide correct @udgment and the range of confidence interval
should "e checked. :hen n JF 28, the most confidence interval ranges are smaller than
2 and the confidence interval method with checking the confidence interval range can
"e considered as a good estimation method.
It is enough to present the shortcomings of these two methods using three times
sampling in standard deviation formulation and one time sampling in confidence
interval method . Thus further sampling is not performed in this study.
In the o"servation descri"e in the net chapter, the accesses with through movement
at Sakurayama in each signal cycle is a"out 5 vehicles and if we consider three signal
cycles as aggregation time interval the accesses is a"out 15. 3ecause we only
consider the vehicles with TT, if 7 G of accesses with through movement at %okiso,
the si0e of population is a"out 12. =onsequently, n F 28 means that a"out 22.5 G
vehicles are needed as #Vs to estimate the population mean of the study link with
F 15 s.
98
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 16/59
Table *-#. Appli!ability o& prevalent &ormulations
n HeferenceStandard deviation formulation =onfidence interval method
Sample 1 Sample 2 Sample /&G)
$(H p mean sd $(H p mean sd $(H p mean sd $(H p range low high
1 87 1 1 1// /.5 . 8 199 -5.6 . 8 19/ --.9 1 5 1/ 1/511 88 . 72 152 /6. . 8/ 157 -5.5 . 72 152 /9.9 1 8/ 58 1// 16
12 89 . 5 172 -8.8 . 7/ 15 /7./ . 57 18/ -8.9 1 85 51 1// 17-
1/ 72 . -9 179 -8.6 . 96 19/ --.8 . 96 195 --.6 1 96 5- 1/8 16
1- 86 . 79 1-6 /5.1 . 86 158 /6.6 . 86 15- /6.6 1 86 -8 1/2 186
15 7- . 7 155 -.1 . 8/ 191 --.5 . 78 1-6 /2.- - 5- 199 2219 79 . 9/ 181 -8.7 . 96 19/ -5.8 . 96 19/ -9.8 1 96 -6 1/6 176
18 77 . 72 152 /5.7 . 77 1-- /.- . 72 151 /7.6 1 72 - 1/1 181
17 61 . 87 159 -1.2 . 7/ 15 /9.2 . 7/ 15 /8.- 1 91 -7 1-7 165
16 6 . 86 155 /7. . 97 195 -9.8 . 57 185 -8.- 1 86 /7 1/9 18-
2 61 . 8 192 -2.1 . 75 1-7 /5.2 1 85 - 1/7 1871 65 1/6 21.8
21 61 . 81 19 -5.1 . 81 191 -/.6 1 65 1-1 22.1 1 52 -/ 158 21
22 62 . 88 159 -1. 1 61 1-/ 27. . 88 158 -.6 1 72 /- 1/8 181
2/ 6/ . 91 182 -9.2 . 8 19- -/.9 . 8- 156 -/.1 1 8- /8 1-1 187
2- 6/ . 86 155 /7.6 . 81 191 -/.8 . 7/ 152 /5.6 1 77 28 1// 19
25 6/ . 97 195 --.9 . 89 157 -2.5 1 82 /9 1-/ 1861 7- 1-6 /5.
29 6- 1 75 1-7 //.7 . 71 152 /7.2 . 95 198 -5.5 1 75 27 1/8 195
28 65 . 8- 156 -2.5 . 8 19- --.6 1 76 1-5 27.- 1 71 / 1/5 199
27 69 1 79 1-6 /1.- . 85 157 -1./ . 86 155 /6.6 1 85 /2 1-/ 18526 69 . 99 198 -5.6 . 99 197 -9.5 . 99 198 -9.9 1 86 26 1/6 197
/ 69 . 98 199 -5.5 . 58 185 -5.9 . 88 157 -.7 1 98 /- 1-6 17/
/1 65 . 81 191 -/.- 1 95 /5 151 1791 7- 151 /5.5 1 88 158 -.
/2 68 . 82 19 -/./ 1 85 157 -./ 1 71 152 /8.2 1 71 28 1/6 198
// 68 1
89 156 -.9 1
89 157 -.2 .
8 19- --.8 1 72 29 1/7 195/- 68 . 59 189 -7.1 . 81 19/ --.2 1 81 192 -2.5 1 72 25 1/7 19-
/5 67 . 99 198 -5. . 81 192 --./ 1 8- 156 -2.5 1 81 26 1-8 189
/9 67 1 7/ 25 1/6 19/1 82 19 -/.8 1 82 191 -/.6 1 71 155 -.2
/8 67 1 8/ 156 -2.7 . 92 181 -9.5 1 89 157 -1.8 1 71 29 1- 199
/7 67 1 86 155 /6.8 1 89 157 -1.- . 9/ 196 -9.2 1 99 / 15/ 17/
/6 67 1 98 / 152 1711 82 192 -2.8 1 8- 156 -2. 1 86 15- /7.7
- 66 1 7/ 151 /5.2 1 7 15/ /7.6 1 8/ 19 -/.- 1 7 25 1-2 198
97
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 17/59
9.- Summary of this chapter
Though standard deviation formulation and confidence interval method are
considered as good estimation methods for the adequate num"er of #Vs at link level
and are used in several network level studies, this chapter shows that these methods are
not availa"le for a signali0ed link due to travel time has multi!peak distri"ution. The
standard deviation formulation is sensitive with sampling error and cannot provide
consistent result, and the confidence interval method is needed to add additional
criterion to provide correct @udgment.
96
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 18/59
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 19/59
Chapter
0er&orman!e stimation
+ven if the time period is very short &e.g., 5 min), the travel times of all vehicles on
a signali0ed link always can "e divided into several groups "y turning movements at
upstream and downstream intersection and the delay at downstream intersection, and
each group has different travel time characteristics. #ro"e reports on a link within a
time interval are considered as a sample. :hen the proportion of #Vs is different over
these su"groups, sampling error arises and "ecomes serious in small si0e sample.
This chapter introduces a new performance indicator and proposes a method to
estimate the new performance indicator using small si0e pro"e reports on signali0ed
link. The new performance indicator is essentially equivalent to conventional time!
"ased performance indicators such as mean travel time or space!mean speed and has
some desira"le features.
8.1 #erformance indicator
3y pro"e vehicle technique, route or link travel time can "e o"tained directly.
Therefore, time!"ased performance indicators such as mean travel time or space!mean
81
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 20/59
speed are commonly suggested for pro"e!"ased $dvanced Traveler Information
Systems &$TIS). Buiroga &2) also indicated that time!"ased indicators are
etremely powerful, versatile and desira"le. ?e compared three commonly used
categories of performance indicators &highway capacity manual &?=) "ased,
queuing!related and time!"ased) and provided the reasons why time!"ased
performance indicators are preferred.
Though time!"ased performance indicators have numerous advantages, it is difficult
to estimate the conventional time!"ased indicators relia"ly "y limited num"er of #Vs
as shown in the previous chapter.
Therefore, this chapter introduces a new performance indicator and proposes a
method to estimate the new performance indicator using small si0e pro"e reports on
signali0ed link. *urther, it will "e shown that the new performance indicator is
essentially equivalent to conventional time!"ased performance indicators such as mean
travel time or space!mean speed and has some desira"le features.
:hen demand eceeds capacity at a link>s upstream intersection &e.g., morning
peak), it is reasona"le to assume that the vehicles with through movement will access
the link with uniform distri"ution over each green period. If we further assume that the
vehicles with TT also access the link with uniform distri"ution, the ratio "etween the
num"er of Set I and the num"er of vehicles with same access green period, p, can "e
considered as a performance indicator &+q. /).
N Set I
p ( N Set I & N Set II &/)
where N Set I and N Set II represent num"ers of vehicles in Set I and Set II,
respectively. The performance indicator has some desira"le features.
82
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 21/59
*irstly, as a performance indicator, p is essentially equivalent to conventional time!
"ased performance indicators such as mean travel time or space!mean speed. $s p
decreases, mean travel time increases and space!mean speed decreases monotonously.
Secondly, using p and information a"out traffic signal &such as cycle length, green
time and offset time "etween downstream and upstream intersections), corresponding
travel time distri"ution can "e calculated approimately. *or a given p, the distri"ution
can "e approimated as a composite of two normal distri"utions. +q. - 7 provide a
possi"le form of formulations to calculate the means of two normal distri"ution and
total mean using p and traffic signal. These formulations are acquired from *igure 8!1,
which is illustrated "ased on uniform assumption. The vehicles with through access
movement and left(right departure movement will queue in left(right turning lanes at
downstream intersection and thus the length of queue of vehicles with TT at
Figure -1. A +ueue o& vehi!les that a!!ess a link in same green perio)
8/
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 22/59
downstream will "ecome shorter than the length at upstream. The in +q. - 5
reflects the reduction of the length and is dependent on the ratio of the vehicles with
through access movement and left(right departure movement to the vehicles with TT.
In this study, is simply specified as .8 and the variation with space and time are
not discussed precisely due to lack of real data for several links in road network. *or
δ ,
an estimated value is given "ased on an o"servation in net section. The variances
of two normal distri"utions are in!queue variance and can "e treated as constant. In
contrast, mean travel time or space!mean speed has no a"ility to calculate the
Head Set I ,Head Set II
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 23/59
distri"ution.
µSet I
( Head Set I #.5 '&1#α ) ' p ' EGin
&-)
(offset & EGout #α ' p ' EGin #.5 '&1 #α ) ' p ' EGin
µSet II
( Head Set II #.5'&1#α ) '&1# p) ' EGin
&5)(offset &cycle # p ' EGin #.5'&1#α ) '&1# p) ' EGin
µall ( p 'µSet I &&1 # p) 'µSet II &9)
EGin (Gin &δ &8)
EGout (Gout &δ &7)
where
p ( performance indicator
(travel times of head in Set I and Set II
µSet I , µSet II , µall (mean of Set I, Set II, all
cycle,offset (signal length and signal offset "etween upstream and
downstream instersection
Gin ,Gout (green time of upstream and downstream intersection
for through movement
α (reduction rate of length of vehicles with TT "etween
downstream and upstream
δ (constant
8-
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 24/59
Thirdly, using the knowledge a"out distri"ution, it is possi"le to develop a method
that can minimi0e the effect of the sampling error arisen from intersection delay and
efficiently estimate p from small sample. $ simple estimation method is presented in
the su"sequent section.
:hen demand is lower than capacity &e.g., off!peak), the uniform assumption
doesn>t hold and it is possi"le that mean travel time in peak hour is smaller than that in
off!peak. =onsequently, the proposed performance indicator is not applica"le in off!
peak and in distinguishing peak and off!peak situation.
8.2 4"servation of vehicle accesses at upstream intersection
The o"servation was made on Kovem"er 7th, 25 &Tuesday). The link from
Sakurayama intersection to %okiso intersection &north "ound only) is chosen and the
two intersections are investigated &*igure 8!2). The link is a primary arterial link in
Kagoya, Capan. The link has three lanes and is a"out 65m. Kagoya has a high density
road network and there are five signali0ed intersections on the link ecept Sakurayama
and %okiso intersections. Such link is fundamental element of arterial road and the
o"servation spot
Sakurayama
. 1 2
m
%okiso
;H!Dink Type
$rterial Hoad
inor Hoad
Figure -#. Stu)y link an) observation spot
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 25/59
85
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 26/59
properties are important.
The green, am"er and red time of Sakurayama for through movement is 5/ s, / s and
7- s, respectively and the %okiso is 91 s, - s and 85 s, respectively. The cycle length of
the two intersections is 1- s and the offset time is 12/ s. The signal time is an
o"servation of one day and day!to!day variation is unknown. In peak hour, the traffic
signals should operate in optimi0ed scheme and the variation may "e very small.
The access times of all vehicles during 7' 6' am &29 signal cycles) were
o"served at the o"servation spot shown in *igure 8!2. Then the access times are
organi0ed "y signal cycle &green starting for through movement as cycle starting). The
num"er of accesses, through accesses and left(right turnings &the sum of left and right
turning) are summari0ed in each signal cycle and the access times are converted to the
offset times from the corresponding signal cycle "eginning. *igure 8!/ shows the
num"er of accesses, through accesses and left(right turnings in each signal cycles.
3ecause these flows decrease during the last half, only the fist 12 cycles are used in the
8 0
6 0
r a
f f i c ) l o *
4 0
2 0
0
1 2 3 4 5 6 7 8 9
All+rou,+
-eft./i,+t
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
sclende
Figure -'. Tra&&i! &low o& ea!h signal !y!les
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 27/59
89
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 28/59
Table -1. Summary o& a!!esses o& signal !y!les
sclende All -/ ffset time from
1 76 58 18 101112...5961: -
2 72 54 18 111213...6162: -
3 66 50 16 91011...6161: -
4 72 58 14 91112...6162: -
5 67 54 13 81112...5959: -
6 68 52 16 8910...5859: -/
7 73 57 16 6810...6161: -/
8 70 56 14 91012...6162: -
9 66 54 12 81011...5858: -
10 73 56 17 9911...6061: -
11 60 48 12 91113...6166: -12 68 48 20 91213...5859: -
Avera,e 70 54 16 ;e,innin, of 9 s
simulation, which needs sta"le flow input to keep a certain performance level. Ta"le 8!
1 shows the flows and offset times of the first 12 cycles and the average. The offset
time in the ta"le are separated into two groups "y turning movement' through
movement &T) and left(right turning &D(H). In *igure 8!/ and Ta"le 8!1, the cycle inde
of o"servation is termed 4"sL=ycleLInde and is distinguished from the cycle inde of
simulation used in the net section, which is termed SimL=ycleLInde.
8./ Simulation at performance levels
To o"tain travel times of vehicles with TT across a link at different performance
levels, a simple network that consists of a link &link 2 in *igure 8!-), upstream and
downstream intersections, and ad@acent links are modeled in a simulation developed
"ased on an o"servation.
*igure 8!- shows the position of reference points of entrance and eit of link 2. $t
these reference points, the access and departure of a vehicle can "e @udged easily and
88
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 29/59
Figure -(. Simulation network
thus the time of access and departure can "e measured accurately. *urthermore,
acceleration process of vehicle is mostly finished at the reference points and running
time and intersection delay are included "etween two reference points completely.
The "ehaviors of vehicles at the entrances of link 2 and link / are considered similar,
and an o"servation at entrance of a real link is used to model the "ehaviors at the
entrances of link 2 and /. The entrance of link / is also the eit of link 2 &*igure 8!-).
The o"servations of each signal cycle, that consists of the num"er of accesses,
through accesses and left(right turnings and the offset times from the corresponding
signal cycle, are used on the two intersections &I and II) in the simulation. It is
equivalent to that the signals of the two intersections are set as same as the signal of
Sakurayama &cycle F 1- s, green F 5/ s, am"er F / s, and red F 7- s). The offset time
of the two intersections is set as same as the offset time "etween Sakurayama and
%okiso &offset F 12/ s).
Three levels of performance & p is 1G, 85G and 5G, respectively) are simulated.
That is the range of p is from 1 G to 5 G and the resolution of performance level is
25 G. In each performance level, the simulation is run for 25 signal cycles and the
vehicles that have TT and access link 2 during first 2- signal cycles are analy0ed.
87
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 30/59
In the simulation program, each link is considered as *I*4 &*irst In, *irst 4ut)
queue M a type of data structure. 3efore running simulation for each performance
level, the ad@acent links at upstream intersection &id' 1, -, and 5) are initiali0ed "y
sufficient num"er of vehicles to provide the vehicles that access link 2 within 25
signal cycles. Dink 2 is also initiali0ed "y different num"er of vehicles to reali0e three
performance levels. $ll initiali0ed vehicles are specified a unique numeric identifier.
The signal of upstream intersection &id' I) for through movement is switched to
green at &&k #1) 1- sec &k is SimL=ycleLInde of intersection I and 1 25).
The o"servations of 12 signal cycles are repeated 2 times to determine the num"er of
through accesses & N T , from link 1 to link 2), left(right turnings & N RL, from link -, 5 to
link 2) and offset times for each k . *or eample, the 2nd o"servation in 12 signal
cycles are used for k F 2 and /rd o"servation is used for k F 15. $t green starting of
each cycles & &&k 1) '1- sec), two operations are performed in the simulation' &1)
N T vehicles are transferred from link 1 to link 2 and the access times of the vehicles
&through movement) are calculated using the green starting and the offset times from
the relevant o"servation, and &2) N RL vehicles are transferred from link -, 5 to link 2
and the access times of the vehicles with left(right turning are set as E the travel times
from vehicles without TT is not considered in this study.
The signal of downstream intersection &id' II) for through movement is switched to
green at 12/ && #1) '1- sec & is SimL=ycleLInde of intersection II and 1 25).
The o"servations of 12 signal cycles are repeated as descri"ed a"ove. The num"er of
departures & N all , link 2 to link /, 9, 8) and the num"er of through departures & N T , link 2
to link /) in the simulation are determined using the num"er of accesses and through
accesses in the o"servation, respectively. $t green starting of each cycles, N all vehicles
are moved out from link2, then N T vehicles are chosen randomly and moved into link /
86
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 31/59
&through movement) and remainder are moved into link 9, 8 &left(right turning). The
departure times of vehicles with through movement are calculated using the green
starting and the offset times and the departure times of vehicles with left(right turning
are set as .
:hen the simulation of each performance level is terminated, vehicle id, access
cycle inde, and travel time of each vehicles with TT &from link 1 to link /) are
recorded into a file.
8.- The simulation output
*igure 8!5 illustrates the mean of Set I, the mean of Set II, and performance
indicator p of each signal cycle for first -7 cycles from simulation for p F 85G. The
minimum value and maimum value of Set I and Set II are also illustrated "y dot line.
N all , N T , N RL change over signal cycles in the simulation, so p is not invariant and
varies around the mean of p as shown in the figure.
r a v e l i m e ! s "
mean of Set mean of Set p
1 3 5 7 9 11 14 17 20 23 26 29 32 35 38 41 44 47 Simclende
Figure -5. Travel times an) p "Simulation &or p 5/$
7
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 32/59
0 . 0
8
100<Set Set
0 . 0
6
75<50<
e n s i t
0 . 0
450<
d
75<
0 . 0
2
0 . 0
0 100<
100 120 140 160 180 200 220 240
ravel ime !s"
Figure -*. Link travel time )istribution " p 122/3 5/3 an) 52/$
*igure 8!9 illustrates the travel time distri"utions for p F 1 G, 85 G, and 5 G
using travel time reports during 2- cycles. $s shown in the figure, link travel time
distri"ution at a certain performance level can "e approimated as a composite of two
normal distri"utions. The figure also shows that as p decreases, the proportion of the
set with delay increases and the means of the two sets increase. =onsequently, mean
link travel time increases as p decreases.
Ta"le 8!2 summari0es the output of the 2- signal cycles. The ta"le presents the mean
of all & !all ), the mean of Set I & !Set I ) and the mean of Set II & !Set II ). The difference of !all
"etween two consecutive performance levels is a"out / s. The space!mean speed is also
presented. In the ta"le, the items with hat are the result from +q. -7 and
Table -#. Summary o& simulation output
p Speed&km(h)µ
all
µSet I
µSet II
µNall
µNSet I
µNSet II
1G 28 126 126 M 1/1 &!2) 1/1 &!2) M
85G 21 191 1- 22/ 192 &!1) 1-2 &!2) 222 &1)
5G 17 16/ 151 2/- 16/ &) 15/ &!2) 2// &1)
71
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 33/59
the red values are the difference from the simulated values. EGin in *igure 8!1 is the
range of offset times for through movement in each signal cycle and the average is 52 s
from Ta"le 8!1E that is δ F 1 s. In +q. -7, α .8 are used and the formulations provide
good estimates.
8.5 #roposed estimation method
Several studies estimated the smallest num"er of #Vs that are required to estimate
link travel time relia"ly' link level studies &Buiroga and 3ullock, 1667E ?ellinga and
*u, 1666) and network level studies &Srinivasan and Covanis, 1669E =hen and =hien,
2E =heu et al., 22). In these studies, sample mean &pro"e link travel time reports)
is directly compared to population mean &travel times of all vehicles). *or instance, if
estimation error &the difference "etween sample mean and population mean) is small
than allowa"le error &e.g., 1 G of population mean), it is considered that the
estimation result is relia"le.
*igure 8!9 and Ta"le 8!2 show that as performance "ecome worse &as p decreases),
mean link travel time increases and the means of the two sets &without delay and with
delay) also increase. It indicates that performance level can "e estimated "y the means
of the two sets of a sample &pro"e travel time reports) ecept sample mean of link
travel time. The essential of the proposed method is estimating performance level using
the means of the two sets of a sample instead of sample mean.
The travel time reports in each signal cycle o"tained from the simulation at
performance level 85G is treated as a population. Samples are taken from these
populations for each sample si0e n &2- times for each sample si0e n). *or each sample,
+q.6 &proposed method) and +q.1 &conventional method) are used to @udge the
72
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 34/59
Table -'. Comparison between propose) an) prevalent metho)
Sample si0e #roposed method #revalent method
/ 22 &6.2 G) 1/5 &59./ G)
5 6 &/.7 G) 76 &/8.1 G)
1 9 &2.5 G) -1 &18.1 G)
15 8 &2.6 G) 26 &12.1G)
2 - &1.8 G) 18 &8.1 G)
estimated performance level' these equations are eamined for p F 85 G and the two
ad@acent performance levels & p F 1 G and p F 5 G) &O p,Set I, O p,Set II, O p,all are
o"tained from Ta"le 8!2) and the p" that minimi0es these equations is considered as
the estimated performance level. If the performance level p F 85G is concluded "y a
sample, it is considered as success and if others & p F 5 G or 1 G) it is considered as
failed. Ta"le 8!/ shows the failed times and the error rate at five levels of sample si0e.
In +q.6, each set of sample and population are compared respectively. 3y this, the
influence of the sampling error arisen from intersection delay is eliminated. ;ue to the
o"vious difference "etween Set I and Set II, it is easy to @udge which set a sample case
"elongs to.
e p
(&µ # xSet I
)2 &&µ # xSet II
)2
&6) p,Set I p,Set II
where
e p
µ p,Set I , !
p,Set II
xSet I
, x
Set II
(the sum of the square of deviation of the means of the two sets of a sample
from the means of the two sets of a performance level p
(mean of Set I and Set II of a performance
level (mean of samples in Set I and Set II
#opulation mean at p F 85 G is different from the means of two ad@acent
7/
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 35/59
performance levels & p F 1 G and p F 5 G) a"out / s &see Ta"le 8!2). Thus, for
conventional method, if the difference "etween sample mean and population mean is
larger than 15 s &a"out 1 G of mean at p F 85 G), it will "e concluded that the
estimated performance level is p F 1 G or p F 5 G &failed).
e p
(&µ # x)2&1)
p,all
where
e p (the square of deviation of sample mean from true mean of a performance level p
µ p,all (mean of a performance level
x (mean of samples
$s shown in Ta"le 8!/, at sample si0e /, the error rate of proposed method is lower
than 1 G while the error rate of conventional method is higher than 5 G. +ven if the
sample si0e increases to 15, the conventional method can not provide the same quality
as proposed method at sample si0e /. In the simulation, average num"er of vehicles
with TT in each cycle is /9 &range' 25 -5). If 1 G is regarded as accepta"le error
rate, the adequate #V rate is a"out 7 G. The conventional method needs a"out -2 G to
o"tain the same quality. If traffic condition is considered invaria"le in consecutive three
signal cycles, the adequate #V rate in proposed method "ecomes a"out /G and in
conventional method "ecomes a"out 19 G. 4"viously, the adequate #V rate is affected
"y the resolution of performance level &e.g., 25 G in this study). ?igher resolution &e.g.,
1 G) needs more #Vs.
8.9 Summary of this chapter
It is epected that link travel time can "e estimated relia"ly "y relatively small
7-
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 36/59
num"er of #Vs in #ro"e!"ased $TIS. ?owever, when sample si0e is small, sampling
error makes difficult to estimate population mean using sample mean directly.
Sampling error arises from two sources' turning movement and intersection delay.
This study suggests that the effect of sampling "ias from the former should "e
eliminated "y redefining the population to only travel times from vehicles with TT.
?owever, the latter is inevita"le and this chapter proposes a new estimation method
that minimi0es sampling error from the latter.
$s mentioned in =hapter - and =hapter 5, as link performance decreases, the
proportion of the group with delay increases and the means of the two groups also
increase. *or estimating link performance, it is needed to estimate the proportion or the
means of the two groups or "oth. In the proposed estimation method, the means of the
two groups is estimated directly using pro"e reports instead of the proportion or mean
travel time. The failure rate of the proposed method is lower than 1 G at sample si0e /
and the conventional method can not provide the same quality even if sample si0e
increases to 15.
:hen the link performance was identified, mean link travel time can "e estimated
"y the means of the two groups and the relationship of the proportion and the means of
the two groups. *or estimating mean travel time, it is needed to estimate "oth the
proportion and the means of the two groups. *ortunately, when traffic signal is known
and under uniform assumption, the relationship "etween the proportion and the means
of the two groups can "e identified &see +q. - 7). The proportion can "e o"tained
indirectly "y the relationship and then the mean travel time can "e estimated.
$ new performance indicator is introduced and a set of formulations is proposed to
o"tain link travel time distri"ution at a certain performance level using information
a"out traffic signal and the new performance indicator. In the formulations, there is a
75
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 37/59
parameter & ) and the parameter might "e link specific. *urther research is needed to
verify the availa"ility of the formulations in different links.
79
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 38/59
Chapter 4
ehi!le-to-ehi!le Travel Time ariability
Travel time varia"ility plays an important role in travel decision and there is a
growing attention to travel time varia"ility measurement &%raves et al., 2E Di et al.,
29E 4h and =hung, 29). Travel time varia"ility has two components' one from the
varia"ility of performance level, and another from the varia"ility of vehicle!to!vehicle
at certain performance level. The performance levels of each link in a road network
may vary over day!to!day and time!to!time. It can "e measured "y traffic detectors
such as loop detector, video camera and pro"e vehicle. $t a certain performance level,
vehicle!to!vehicle varia"ility still arises from driver "ehaviour such as aggressiveness
and lane choice &Di et al., 29) and from intersection delay. In peak time period, the
"ehaviour of individual driver will "e limited and the intersection delay "ecomes the
ma@or source of vehicle!to!vehicle varia"ility.
The vehicle!to!vehicle varia"ility of corridor travel time is important to pro"e!"ased
estimationE if the varia"ility is relatively small, travel time reports from #Vs can "e
aggregated at corridor level directly and corridor travel time can "e estimated "y small
num"er of #Vs.
This chapter presents vehicle!to!vehicle varia"ility on a signali0ed corridor at three
performance levels in peak time period. $ simulation is developed to model the
78
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 39/59
Figure 4-1. Simulation network
vehicles on the corridor and the ad@acent links. Though the simulation is very simple,
performance level and vehicle movement at intersection are considered eplicitly,
which are two critical factors that influence travel time. $ performance indicator
proposed in =hapter 8 is employed to demonstrate the sta"ility of the performance in
each three levels.
7.1 Simulation network
To o"tain travel times of all vehicles that traverse a corridor completely at different
performance levels, a simple network is modeled in a simulation that is a simple
etension of the simulation shown in section 8./ &*igure 7!1). In this figure, the
intersections are num"ered with Homan numerals and the links are num"ered with
$ra"ic numerals. The cross links are num"ered using from 11 to 2 and the access link
in the cross links are num"ered using odd num"er and the departure link in the cross
links are num"ered using even num"er. The left and right turning movements at an
intersection are not distinguished in this study and thus two access cross links are
assigned "y the same id and two departure cross links are also assigned "y the same id.
:hen a vehicle accesses link 2 from link 1 and departs link 5 to link 9, the vehicle is
77
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 40/59
considered as traversing the corridor completely and the travel time from link 2 to link
5 is considered as corridor travel time. The signals of the intersections and the offset
times "etween consecutive two intersections are specified equally "ased on the
o"servation shown in 8.2.
The figure also shows the positions of reference points of entrance and eit of each
links. The entrance of a link is also the eit of previous link. *or eample, the entrance
of link / is the eit of link 2.
7.2 Simulation procedure
Three levels of performance & p of link 2 5 are 1 G, 85 G, and 5G,
respectively) are simulated. In each performance level, the simulation is run for 25
signal cycles and the link travel times of vehicles with TT and the corridor travel times
are calculated.
In the simulation program, each link is simply considered as *I*4 &*irst In, *irst
4ut) queue P a type of data structure. The access links &link 1, link 11, link 1/, link 15,
link 18, and link 16) are initiali0ed "y sufficient num"er of vehicles to provide the
vehicles that access the network during 25 signal cycles. The links that compose the
corridor &link 2 5) are also initiali0ed "y different num"er of vehicles to reali0e the
three performance levels. $ll initiali0ed vehicles are specified a unique numeric
identifier.
The information of 12 signal cycles in the o"servation are repeated several times to
determine the num"er of through accesses & N T ), the num"er of left(right turning
accesses & N LR), and offset times for a cycle inde of an intersection in the simulation.
*or eample, the information of 2nd signal cycle in the o"servation is used for cycle
76
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 41/59
inde F 2 and /rd signal cycle is used for cycle inde F 15 in the simulation. The
num"er of left(right turning departures is specified as same as the num"er of left(right
turning accesses N LR.
$t 12/'&i #1) &1- '&ci #1) sec, the signal of cycle inde ci of intersection i for
through movement is switched to green in the simulation &ci is 1 25). $t the
moment, two operations are performed in the simulation' &1) N T # N LR vehicles are
moved out from linkQiR , then N T vehicles are chosen randomly among these vehicles
and moved into linkQi &1R and the rest of vehicles are moved into linkQ1 &2 'iR .
*or the vehicles with through movement & N T ), the departure times of upstream link and
the access times of downstream link are calculated using the green starting time and
the offset times from the relevant signal cycle in the o"servation. *or the vehicles with
left(right turning, the departure times of upstream link and the access times of
downstream link are set as E the link travel times of vehicles without TT is not
discussed in this paper. &2) N LR vehicles are transferred from linkQ1 &2 'i #1R into
linkQi &1R . The departure times of upstream link and the access times of downstream
link of these vehicles are set as .
:hen the simulation of each performance level is terminated, the result is
summari0ed "y each vehicle. *or a vehicle, the links that the vehicle traverse with TT
are identified and link id, cycle inde of access and link travel time of each link are
recorded into a file.
In this simulation, performance level and vehicle "ehavior at intersection are
considered eplicitly, which are critical factors on travel time.
6
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 42/59
7./ Travel time varia"ility
$%&%' Link travel time varia(ility
In this section, travel times on link 2 generated "y the simulation are presented.
*igure 7!2 illustrates mean travel time of Set I, mean travel time of Set II, and
performance indicator p for the first -7 signal cycles at three performance levels. The
minimum value and maimum value of Set I and Set II of each signal cycle are also
illustrated "y dot lines. $s shown in the figure, p fluctuates over signal cycles around
epected value of each performance level. The causes are the variation of N T and N LR
over signal cycles in the simulation, and pseudo!random num"er generated "y the
simulation. The former is consistent with the real situation. In this study, sufficient
num"er of signal cycles &25 signal cycles) is simulated to counteract the influence of
the latter on travel time distri"ution.
*igure 7!/ shows link travel time distri"utions at three performance levels, which
are o"tained using link travel times during 25 signal cycles. $s shown in the figure,
link travel time distri"ution has two peaks at each performance levels. $s p decreases,
the peak shifts towards right side and the ratio of Set II increases. =onsequently, mean
travel time on the link increases monotonously.
7.- =orridor travel time varia"ility
*igure 7!- shows corridor travel time distri"ution at three performance levels. Dink
travel time will "elong to Set I or Set II. Ta"le 7!1 shows the percentage of vehicles in
each com"ination of the num"er of Set I and Set II on the four links.
61
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 43/59
r a v e l i m e ! s "
mean of Set mean of Set p
1 3 5 7 9 11 14 17 20 23 26 29 32 35 38 41 44 47 cle nde
p F 1 G &a)
1 2 0
1 1 0
1 0 0
p ! < "
9 0
8 0
r a v e l i m e ! s "
mean of Set mean of Set p
1 3 5 7 9 11 14 17 20 23 26 29 32 35 38 41 44 47 cle nde
p F 85 G &")
r a v e l i m e ! s "
mean of Set mean of Set p
1 3 5 7 9 11 14 17 20 23 26 29 32 35 38 41 44 47 cle nde
p F 5 G &c)
7 0
6 0
5 0
p ! < "
4 0
3 0
Figure 4-#. Link travel time an) per&orman!e in)i!ator
62
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 44/59
The distri"utions at each performance level are multi!peak distri"ution and the
difference "etween two peaks is fairly large &a"out 2 G of travel time mean). That is
travel time varia"ility is still produced even if the performance level is invaria"le and
the aggregated distance is relatively long. $s the num"er of links traversed "y vehicle
as Set II increases, there are more intersection delay and the corridor travel time
increases.
:hen performance level decreases, there are more Set II and the travel time
increases rapidly.
Table 4-1. 0er!entage o& !ombination o& the o& Set % an) Set %% on &our links "/$
- ' / ' 1 2 ' 2 1 ' / ' -
p F 1G 6.6 6.1
p F 85G 71./ 17.8
p F 5G .7 67.7 .-
0 . 0
8
100<Set Set
0 . 0
6
75<50<
e n s i t
0 . 0
450<
d
75<
0 . 0
2
0 . 0
0 100<
100 120 140 160 180 200 220 240
ravel ime !s"
Figure 4-'. Link travel time )istribution
6/
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 45/59
7 0 0
) r e = u e n c
5 0 0
3 0 0
1 0 0
0
450 500 550 600 650 700 750 800 850 900
ravel ime !s"
0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 ) r e = u e n c
p F 1 G &a)
450 500 550 600 650 700 750 800 850 900
ravel ime !s"
p F 85 G &")
0 2 0 0 6 0 0 1 0 0 0 ) r e = u e n c
450 500 550 600 650 700
ravel ime !s"
750 800 850 900
p F 5 G &c)
Figure 4-(. Corri)or travel time
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 46/59
6-
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 47/59
7.5 Summary of this chapter
In this chapter, vehicle!to!vehicle varia"ility of signali0ed corridor travel time at
three performance levels is presented using travel times from a simulation that
considers performance level and vehicle "ehavior at intersection.
Though an artificial corridor is modeled in the simulation, the simulation can "e
used to o"tain travel time distri"ution &consequently, vehicle!to!vehicle varia"ility) for
a real signali0ed corridor. *or estimating travel time of a real corridor "y the proposed
simulation, the vehicle accesses of each link and the relationship "etween the initial
num"er of vehicles and the performance levels are needed. :hen demand eceeds
intersection capacity, the accesses of a link will "e controlled "y the signal of upstream
intersection and this is independent with performance level. That is the o"servation of
the accesses at each intersections in peak hour can "e used to identify the accesses for
several performance levels. :hen the performance levels of each link are specified, the
links can "e initiali0ed "y different num"er of vehicles "ased on the performance levels
and the corridor travel time can "e o"tained "y the simulation. The relation should "e
affected "y the geographic properties of each links and "e link!specific. The relation
"etween the initial num"er of vehicles and the performance level should "e studied in
the future.
65
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 48/59
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 49/59
Chapter
Con!lusions an) Future Stu)ies
6.1 =onclusions
)%'%' Statistical properties of link travel time
Travel time reports from #Vs should "e aggregated at link level instead of at path
level "ecause there are numerous paths in a city and the path!"ased method would
suffer from low pro"e o"servations. Though the statistical properties of link travel time
are important to implement pro"e!"ased real!time data collection system, little is
known a"out the statistical properties.
Turning movement is usually neglected in practice when aggregating traffic data at
linkE that is, all vehicles traveling a link during a time interval regardless of turning
movements at the link>s ends are defined as population. 3y the traditional definition of
link travel time, link travel time distri"ution will "e concluded as normal or
approimate normal. ?owever, it will inevita"ly accompany with large variance and it
is difficult to estimate mean link travel time using small si0e pro"e reports. Thus, the
research proposes a new definition of link travel time' only consider the vehicles with
through movement at upstream and downstream intersections.
68
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 50/59
3y new definition, historical pro"e reports on an arterial link show that link travel
time is two!peak distri"ution and mean travel time is larger in off!peak time period
than in peak time period. To identify the intrinsic reasons of the phenomenon, a
qualitative analysis and a simulation "ased analysis are performed and the following
conclusions are made'
1) :hen the demand is lower than the capacity of upstream intersection, for
eample in off!peak time period, mean link travel time ehi"its large variation
over time and more importantly cannot trace the change of the num"er of
vehicle accesses. That is, in this phase, mean travel time is not a good indicator
of link performance and the num"er of accesses should "e used for tracing the
change of link performance.
2) :hen demand eceeds capacity at upstream intersection, it is reasona"le to
assume that the vehicle accesses are uniform over green period. In this phase,
though mean travel time can trace the link performance, travel time
distri"ution is two!peak rather than asymptotically normal. ;espite this is not
prefera"le result, it is consistent with the o"servation from historical data and
the widely accepted "elief that link travel times "elong to at least two different
groups' one without delay at downstream intersection and the others with the
delay. $s link performance decreases, the proportion of the group with delay
increases and the means of the two groups also increase &see *igure 7!/).
Asing small si0e pro"e reports, though it is difficult to estimate the proportion
of the group with delay, it is possi"le to estimate the means of the two groups
due to their variances are relatively small. This fact is used to develop
performance estimation method using small si0e pro"e reports.
/) *or avoiding congestion, identifying the performance decrease in the second
67
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 51/59
phase is important.
)%'%* The formulations of the ade+uate num(er
Traditionally, it is considered that the relia"ility of pro"e!"ased estimation depends
on the num"er of the pro"e reports and the adequate sample si0e required to meet the
relia"ility has "een an imperative issue.
Two methods standard deviation formulation and confidence interval method are
commonly accepted. ?owever, these formulations are "ased on =entral Dimit Theorem
and the result is dou"tful when the population is severely nonnormal and sample si0e is
small.
The eamination descri"ed in =hapter 9 shows that these methods are not capa"le
for a signali0ed link due to travel time has multi!peak distri"ution. The standard
deviation formulation is sensitive with sampling error and cannot provide consistent
result, and the confidence interval method is needed to add additional criterion to
provide correct @udgment.
)%'%& erformance estimation
It is epected that link travel time can "e estimated relia"ly "y relatively small
num"er of #Vs in #ro"e!"ased $TIS. ?owever, when sample si0e is small, sampling
error makes difficult to estimate population mean using sample mean directly.
Sampling error arises from two sources' turning movement and intersection delay.
This study suggests that the effect of sampling "ias from the former should "e
eliminated "y redefining the population to only travel times from vehicles with TT.
?owever, the latter is inevita"le and =hapter 8 proposes a new estimation method that
66
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 52/59
minimi0es sampling error from the latter. The failure rate of the proposed method is
lower than 1 G at sample si0e / and the conventional method can not provide the
same quality even if sample si0e increases to 15.
$s mentioned earlier, as link performance decreases, the proportion of the group
with delay increases and the means of the two groups also increase. *or estimating link
performance, it is needed to estimate the proportion or the means of the two groups or
"oth. In the proposed estimation method, the means of the two groups are estimated
directly using pro"e reports instead of the proportion or mean travel time.
:hen the link performance was identified, mean link travel time can "e estimated
"y the means of the two groups and the relationship of the proportion and the means of
the two groups. *or estimating mean travel time, it is needed to estimate "oth the
proportion and the means of the two groups. *ortunately, when traffic signal is known
and under uniform assumption, the relationship "etween the proportion and the means
of the two groups can "e identified &see +q. - 7 in =hapter 8). The proportion can "e
o"tained indirectly "y the relationship and then the mean travel time can "e estimated.
6.2 *uture Studies
In this thesis, the statistical properties of link travel time and the transform of the
properties over the change of the traffic condition were investigated "y historical data,
qualitative analysis, and simulation "ased analysis. ?owever, the result of the analyses
should "e verified "y a field test. $ corridor that consists of several arterial links can "e
chosen as the target of the field study, such as the corridor used in simulation "ased
analysis &see *igure 5!1). *or the field test, a technique that can identify the most
vehicles &e.g., 6 G) at the entrance of each link &see *igure 8!-) is needed. $t present,
1
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 53/59
high!quality video camera is one of the techniques.
*or estimating path travel time, the effect of left(right turning movement should "e
studied. *urthermore, the travel time on local roads also should "e studied.
11
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 54/59
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 55/59
6e&eren!es
$sano, ., Kaka@ima, $., ?origuchi, H., 4neyama, ?., uwahara, ., oshi, .
$kahane, ?.&2/) $ real time traffic signal control "y self!evaluating delay. -ournal
of Infrastructure lannin. Revie/, 0ol%*1, No%2, pp. 786!779. &in Capanese)
3ertini, H. D.&29) Uou are the traffic @am' an eamination of congestion measures.
Transportation Research 3oard $4th 5nnual 6eetin. . 7ashin.ton, 89 .
3oyce, ;., ?icks, C. Sen, $.&1661) In!vehicle navigation system requirements for
monitoring link travel time in a dynamic route guidance system. Transportation
Research 3oard, :1th 5nnual 6eetin., 7ashin.ton, 89 .
=etin, ., Dist, %. *. hou, U.&25) *actors affecting minimum num"er of pro"es
required for relia"le estimation of travel time. TR3 *114 5nnual 6eetin. .
=hen, . =hien, S. I. C.&2) ;etermining the num"er of pro"e vehicles for freeway
travel time estimation "y microscopic simulation. Transportation Research Record
':')% TR3, National Research 9ouncil, 7ashin.ton, 8%9%, pp. 91!97.
=hen, . =hien, S. I. C.&21) ;ynamic freeway travel time prediction using pro"e
vehicle data' link!"ased vs. #ath!"ased. Transportation Research 3oard, $1th 5nnual
6eetin. .
=heu, H. D., ie, =. Dee, ;. ?.&22) #ro"e vehicle population and sample si0e for
arterial speed estimation. 9omputer;5ided 9ivil and Infrastructure En.ineerin., 18,
pp. 5/!9.
=hien, S. I. C. uchipudi, =. .&22) ;ynamic travel time prediction with real!timeand historical data. Transportation Research 3oard $'th 5nnual 6eetin. .
%raves, T. D., arr, $. *. Thakuriah, #.&2) Varia"ility of travel times on arterial
links' effects of signals and volume. National Institute of Statistical Sciences
Technical Report ''2.
?ellinga, 3. H. *u, D.&22) Heducing "ias in pro"e!"ased arterial link travel time
estimates. Transportation Research art 9< Emer.in. Technolo.ies, 1, 258!28/.
?ellinga, 3. *u, D.&1666) $ssessing epected accuracy of pro"e vehicle travel time
1/
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 56/59
reports. -ournal of Transportation En.ineerin., 0ol% '*4, No% =, pp. 52-!5/.
?origuchi, H.&22) The advantage of event!periodic data recording for pro"e vehicle
system. roceedin.s of Infrastructure lannin., 0ol% *=, 98;R>6 .&in Capanese)
?origuchi, H.&22) Strategic disposition of tai pro"e to accomplish effective data
collection for travel information provision ! a theoretical framework and its practice.
roceedin.s of Symposium on ITS *11*.
?origuchi, H. :ada, .&25) Intelligent detection of irregular congestion with an
on!"oard pro"e equipment. roceedin.s of Symposium on ITS *114. &in Capanese).
Internet ITS Coint!Hesearch %roup &22) Internet ITS research report.
uwahara, .&2/) ;emand analysis considering traffic congestion. roceedin.s of
Infrastructure lannin., 0ol% *$, 98;R>6% &in Capanese)
Dester, #.&25) $W pathfinding for "eginners.
Di, H., Hose, %. Sarvi, .&29) Asing automatic vehicle identification data to gain
insight into travel time varia"ility and its causes. Transportation Research 3oard,
$4th 5nnual 6eetin. .
Diu, .&29) =ost!effectiveness of a pro"e vehicle system at lower frequencies. ;octor
dissertation, Kagoya Aniversity, Capan.
Diu, ., Uamamoto, T. orikawa, T.&29) $n analysis of the cost efficiency of pro"e
vehicle data at different transmission frequencies. International -ournal of ITS
Research, -.
akimura, ., ikuchi, ?., Tada, S., Kaka@ima, U., Ishida, ?. Tetsuro, ?.&22)
#erformance indicator measurement using car navigation systems. Transportation
Research 3oard, $'th 5nnual 6eetin., 7ashin.ton, 89 .
iwa, T.&25) Travel "ehavior analysis in road networks "ased on pro"e vehicle data.
;octor dissertation, Kagoya Aniversity, Capan.
iwa, T. orikawa, T.&2/) $nalysis on route choice "ehavior "ased on pro"e!car
data. roceedin.s of the '1th 7orld 9on.ress on Intelli.ent Transport Systems, 98;
R>6 .
iwa, T., Sakai, T. orikawa, T.&2-) Houte identification and travel time
prediction using pro"e!car data. International -ournal of ITS Research, 2, 21!27.
ontgomery, ;. =.&166-) 5pplied statistics and pro(a(ility for en.ineers, , Cohn :iley
Sons Inc.
1-
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 57/59
ousa, H. .&22) $nalysis and modeling of measured delays at isolated signali0ed
intersections. ->?RN5L >@ TR5NS>RT5TI>N ENGINEERING;5S9E, 127, /-8!
/5-.
4chieng, :. U. Sauer, .&22) Ar"an road transport navigation' performance of the
glo"al positioning system after selective availa"ility. Transportation Research art 9<
Emer.in. Technolo.ies, 1, 181!178.
4h, C. S. =hung, U.&29) =alculation of travel time varia"ility from loop detector
data. Transportation Research 3oard, $4th 5nnual 6eetin. .
4ls0ewski, #.&166/) 4verall delay, stopped delay, and stops at signali0ed intersections.
->?RN5L >@ TR5NS>RT5TI>N ENGINEERING;5S9E, 116, 7/5!752.
#an, =., Du, C., :ang, ;. Han, 3.&28) $ gps!"ased arterial roadway network travel
time(delay data collection method. TR3 *11: 5nnual 6eetin.%
Buddus, . $., 4chieng, :. U., hao, D. Koland, H. 3.&2/) $ general map
matching algorithm for transport telematics applications. GS Solutions, 8, 158!198.
Buiroga, =. $.&2) #erformance measures and data requirements for congestion
management systems. Transportation Research art 9;Emer.in. Technolo.ies, 7,
278!/9.
Buiroga, =. $. 3ullock, ;.&1667a) ;etermination of sample si0es for travel timestudies. ITE -ournal, 97&7), 62!67.
Buiroga, =. $. 3ullock, ;.&1667") Travel time studies with glo"al positioning and
geographic information systems' an integrated methodology. Transportation Research
art 9< Emer.in. Technolo.ies, 9, 11!128.
Buiroga, =. $. 3ullock, ;.&1666) easuring control delay at signali0ed intersections.
-% Transp% En.r.%, 125, 281!27.
Sen, $. Thakuriah, #.&1665) +stimation of static travel!times in a dynamic route
guidance!system. 6athematical and 9omputer 6odellin., 22, 7/!11.
Sen, $., Thakuriah, #. V., hu, . B. arr, $.&1668) *requency of pro"e reports and
variance of travel time estimates. ->?RN5L >@ TR5NS>RT5TI>N
ENGINEERING;5S9E, 12/, 26!268.
Sen, $., Thakuriah, #., hu, . arr, $.&1666) Variances of link travel time estimates'
implications for optimal routes. International Transactions in >perational Research,
9, 85!78.
Srinivasan, . Covanis, #.&1669) ;etermination of the num"er of pro"e vehicles
15
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 58/59
required for relia"le travel time measurement in an ur"an network. Transportation
Research Record '4&:, TR3, 7ashin.ton, 8%9%,15!22.
Taylor, . $., :oolley, C. +. ito, H.&2) Integration of the glo"al positioning
system and geographical information systems for traffic congestion studies.
TR5NS>RT5TI>N RESE5R9H 5RT 9;E6ERGING TE9HN>L>GIES, 7, 258!
275.
Turner, S. . ?oldener, ;. C.&1665) #ro"e vehicle sample si0es for real!time
information' the houston eperience.
Turner, S. ., +isele, :. D., 3en0, H. C. ?oldener, ;. C.&1667) Travel time data
collection hand"ook report no. *hwa!pl!67!/5. , *ederal ?ighway $dministration,
Teas Transportation Institute.Aesugi, U., Iryo, T., 4neyama, ?., ?origuchi, H. uwahara, .&2/) +stimation of
epectation and variance for a section travel time "ased on fragmentary pro"e
tra@ectories. -ournal of Infrastructure lannin. Revie/, 0ol%*1, No%2, 62/!626,
Capanese.
Van $erde, ., ?ellinga, 3., Uu, D. Hakha, ?.&166/) Vehicle pro"es as real!time atms
sources of dynamic o!d and travel time data. roceedin.s of the 5T6S 9onference
held in St% eters(ur., @lorida, 28!2/.
Vlahogianni, +., %olias, C. and arlaftis, .&2-) Short!term traffic forecasting'
overview of o"@ectives and methods. Transport Heviews, 2-, pp. 5//!558.
:hite, =. +., 3ernstein, ;. ornhauser, $. D.&2) Some map matching algorithms
for personal navigation assistants. Transportation Research art 9< Emer.in.
Technolo.ies, 7, 61!17.
Uamamoto, T., Diu, . orikawa, T.&29) Varia"ility of travel time estimates using
pro"e vehicle data. roceedin.s of the @ourth International 9onference on Traffic
and Transportation Studies AI9TTSB, 287!278.
Uang, C., ang, S. =hon, .&25) The map matching algorithm of gps data with
relatively long polling time intervals. -ournal of the Eastern 5sia Society for
Transportation Studies, Vol.9.
ito, H., ;este, %. Taylor, . $.&1665) %lo"al positioning systems in the time!
domain ' how useful a tool for intelligent vehicle!highway systems.
Transportation Research art 9;Emer.in. Technolo.ies, /, 16/!26.
4oyama, K., Shigeta, ., atsumoto, T. and Ikenoue, .&1689)' akuritsuhensu to
19
7/21/2019 Kelompok 2 Rek Lalulintas
http://slidepdf.com/reader/full/kelompok-2-rek-lalulintas 59/59
shiteno ryoko@ikan no tokusei ni tsuite &4n the characteristics of travel time as
pro"a"ilistic varia"le), #roceedings of /rd Hesearch =onference on Traffic
+ngineering, pp. 86!72. &in Capanese)
Recommended