IET Intelligent Transport Systems

Research Article

Control strategies for solving the problem of ISSN 1751-956X

Received on 5th January 2016
Revised 18th May 2016
traffic congestion Accepted on 12th August 2016
doi: 10.1049/iet-its.2016.0003
www.ietdl.org

Yi-Sheng Huang1 , Yi-Shun Weng2, Weimin Wu3, Bo-Yang Chen1

1Department of Electrical Engineering, National Ilan University, Ilan City, Taiwan
2Department of Electronic Engineering, Army Academy, Zhongli City, Taiwan
3State Key Laboratory of Industrial Control Technology and Institute of Cyber-Systems and Control, Zhejiang University, Hangzhou 310027,

China
E-mail: [email protected]

Abstract: Once a road is with a high traffic density over its loading, the traffic management has to face the fact that how to
alleviate the traffic congestion and how to prevent new traffic jams from being happening in the neighbourhood. This study
presents a new control strategy for alleviating the traffic congestion. For this purpose, the authors employ VISSIM to simulate
and analyse of an urban traffic congestion problem. In addition, two kinds of traffic light orders are involved. On the other hand,
an urban traffic network with three kinds of intersection is used to evaluate their new congestion control policy. Finally, they also
demonstrate three scenario cases low, middle, and heavy traffic flow works to compare with normal regular traffic light control
rules, the multiline control strategies and their new control policy. Simulation results show that their new proposed traffic control
policy is the best one among the others. It is worthy of note that their new proposed policy not only alleviates successfully the
congestion problem but also prevents new traffic jams from being happening in the neighbourhood.

1 Introduction [7] mentioned a variety of strategies have been done, which are
divided into two type strategies: (i) fixed-time strategies and (ii)
Since more and more vehicles are used, the traffic congestion is a traffic-response ones. We know that the former is employed in
growing problem in many metropolitan areas. A traffic jam can most industrialised countries for urban traffic control. However,
start in one of three ways [1]: (i) a temporary obstruction, (ii) a several traffic fundamental modelling technologies issues remain
permanent capacity constraint in the network itself, and (iii) a open. This paper's traffic signal controller is with a predetermined
stochastic fluctuation in the demand within a particular sector of timing scheme. In pioneer works, there is only one-way [2, 3]
the network. Here, we are aiming at traffic congestion problem and/or two-way [4] grid network discussed in an urban net.
which can be categorised under the first type. Roberg [2] proposed However, no any type of phase transitions is discussed in their
accident simulation models which focused on the traffic jam to works. Recently, a three-phase and four-phase traffic signals are
form a comprehensive view of occurred. Wright and Roberg [1] proposed [8] for an urban traffic controller. However, the vehicles
proposed a simple analytical model for incident-based jam growth heading in their complex intersection need not be mentioned. We
and discussed the effect of the length of the channelised part of can understand that traffic light is not only with green, yellow, and
roads and stop line width assignment on jam formation. red but there is also with a green left turn arrow, green right turn
One knows that the incident-based control strategy for arrow, and/or green straight arrow. Therefore, there may occur a
dispersing jams can be divided into two categories [3]. One is red signal turn on together with a left turn arrow on green in a
called static prevention strategy and the other is called dynamic westbound traffic light and also a red signal turn on together with a
control strategy. The former issues on how the road layout features left turn arrow on green in an eastbound one. Obviously, suitable
can be used to alleviate the jam spreading. The latter can be used to modelling tools can make it easy to examine the performance
delay the jam propagation. However, one drawback of the static indices of a traffic control system such as vehicles stopped on a red
strategy is its inflexibility [4]. signal and the following loss of time.
Recently, Long et al. [4] proposed a cell transmission model to Discrete event system (DES) specification is a formalism for
simulate incident-based jam spread and dissipation in a two ways modelling and analysis of DESs. Among a variety of DESs, Petri
network. However, there remain many unsolved problems in nets (PNs) formalism is with strong control logic and also with a
normal urban traffic network, particularly using the same type of clear means of presentation. Hence, it has been proved to be a
phase transitions in all intersections, which limit the usability of powerful modelling tool for DESs [9]. Nevertheless, traditional PN
their contributions. Additionally, their link channelisation is not is not the ability of determining the exact time of the transition
suitable to model a real-world urban traffic network. In a real- firing. For this reason, time PNs is proposed to improve their
world traffic network, systems are with different numbers of line. capability [10]. Recently, PNs have been successfully employed in
Beside, several alternative strategies have developed to handle the urban traffic network control system [11–15] and the railway level
traffic congestion issues by Roberg-Orenstein et al. [3]. However, crossing modelling work [16, 17]. In addition, timed coloured PNs
the above research is lack of realism due to using one kind of road (TCPNs) are also developed for modelling complex DESs. Huang
models. On the other hand, intelligent transportation system (ITS) et al. successfully employed TCPNs to model intelligent urban
is able to improve the efficiency of the transport system and traffic light control systems [11, 17]. In sum, a variety of PNs
enhance its safety and security. Hence, ITS technologies naturally modelling tools are used in urban traffic light controller. However,
are used in traffic congestion issues. Additionally, Figueiredo et al. the behaviour of the traffic light phases is difficult to present in an
[5] use new communication and advance technologies to process easy way. Moreover, lots of mathematical programming methods
traffic congestion problems. Besides, one of hot research topics in are proposed to perform traffic control logic and traffic flow [16,
ITS is the development of distributed traffic information systems 18]. Unfortunately, the mentioned work lacks of addressing how to
(TISs) [6]. However, most of their control strategies are necessary achieve a traffic light signal controller. For overcoming this
through traffic light control systems. Therefore, Papageorgiou et al. drawback, Huang et al. [12] proposed a modular design technology

IET Intell. Transp. Syst. 1

© The Institution of Engineering and Technology 2016
 In this paper, we assume the duration of phases A2–A4, A6–A8 is
25 s and phases A1 and A5 is 15 s. It states that phase A1 changes to
A2 after 15 s and phase A2 changes to A3 after 25 s. Consequently,
the duration of a set period of time (i.e. phase A1 changing to A8) is
180 s.
Fig. 2b shows the rules of the six-phase transition. One can also
refer to our previous work [22], we will leave out the details.
We assume the duration of phases B1, B3–B4 is 15 s and phases
B2 and B5 is 30 s. It means that phase B1 changes to B2 after 15 s
and phase B2 changes to B3 after 30 s. Consequently, the duration
of a set period of time (i.e. phase B1 changing to B6) is 120 s.
Fig. 2c shows the rules of the six-phase transition. We assume
that the duration of phases C1 and C2 are with 45 s. Hence, the
duration of a set period of time is 90 s.

3 Control strategies for urban traffic network

system
Long et al. [4] proposed several control strategies that can be used
in a two-way traffic network. Their control strategies involve the
application of vehicle movement bans at a number of critical
intersections in the traffic network. For example, if an ahead ban is
employed, then the ahead vehicles will be limited to make a right
Fig. 1 Example with 7 × 7 intersections turning movement and a left turning movement proportionally, and
the stop line width devoted to the queue storage area will be
to implement an urban traffic lights controller. A variety of proportionally reallocated to the left turning queue storage area and
reducing traffic jam methodologies [19–21] are proposed. the right turning queue storage, respectively [4]. However, the link
However, their work did not explain how to perform the issues of channelisation is not suitable to model a real-world urban traffic
the traffic light changing. In sum, our proposed approach is network. In a real-world traffic network, systems are with different
distinguished from the previous related work in the following three numbers of line such as as shown in Fig. 1. According to the above
aspects: vehicles movement bans, Long et al. [4] proposed four categories
of control strategies: single-line control, multiline control, area
i. We construct a variety of road crossing, and propose control, and diamond control. In short, their single-line control
sophisticated transitions of the road crossing traffic light strategy takes a reroute method to avoid the vehicles away from the
systems. blocked area. Moreover, their multiline control strategy is very
ii. We propose a novel control policy that is not only able to avoid close to single-line one. For convenience, we depicted their
the traffic congest dispersing but also to avoid new traffic multiline control strategy (i.e. Line 1, Line 2, and Line 3) in
congestions being happened. Fig. 3a. However, the two control strategies are too strong
iii. We evaluate the effectiveness of our control strategies for limitation for vehicles movement. All vehicles are banned to go to
solving the problem of traffic congestion in three kinds of the incident direction. Obviously, the strategies might cause new
traffic flow situations by using VISSIM. traffic congestions being happened nearby. In this paper, we try to
release the restriction of the ahead ban movement. The main reason
The rest of this paper is organised as follows: Section 2 presents a is that our new policy is still able to avoid the traffic congest
brief description of the traffic light system models. Section 3 dispersing and also to avoid new traffic congestions being
describes our control strategies for an urban traffic network system. happened. Here, we are omitted to discuss with the other control
Section 4 presents our simulation works. Section 5 depicts our strategies because the multiline control is the kernel of the others.
simulation results and discussion. Conclusions will be explained in It is well known that traffic congestion begins when more
Section 6. vehicles try to use a road than the road can move. For alleviating
the problem of traffic congestion, we propose a new control
strategy that can be employed to avoid most of vehicles to pass the
2 Basics of traffic light system models
traffic congestion zone in an urban traffic network. Here, we
In this paper, the definition of intersections follows our previous extend the incident position (i.e. incident position, ▾ as shown in
work [12] which is divided into three classes A, B, and C. Here, the Fig. 3a) to a traffic congestion zone (i.e. zone ab as shown in
intersection of two main roads crossing, one main road crossing Fig. 3b) because the extended version is closely the real-world
with minor one, and two minor road crossing, is called classes A, urban traffic network. In fact, that original traffic incident causing
B, and C, respectively. According to the above definitions, one can traffic jam is included in our traffic congestion zone.
point out one A, 12 B's, and 36 C's in Fig. 1. We use five traffic Fig. 3b is helpful one to figure out our traffic congestion control
signals [e.g. a left turn arrow on green (GL), a right turn arrow on policy. Assume that the traffic congestion happens in zone ab on
green (GR), a straight arrow on green (GS), a yellow (Y), and a red the main road. In this case (i.e. Fig. 3b), all right bans are applied
light signal (R)] for the two classes A and B. Therefore, six-phase from North to South on intersection B (i.e. Line 2), all left bans are
and eight-phase transitions can be conducted by the five traffic applied from South to North on intersection B (i.e. Line 3), and all
lights. As for the class C, we employ three signal lights G, Y, and left and right bans are applied on main road intersection B (i.e. Line
R. Using the three signals light, the two-phase transitions can be 1 and Line 4). It is worthy of note that our control policy is
obtained. Considering the real world of the traffic flow, we assume different from Long et al. [4]. No any ahead bans are applied in our
that a traffic light has eight-phase transitions at A, six-phase ones at control policy. Besides, our policy is able to prevent most vehicles
B, and two-phase ones at C. entering zone ab from the other ways (i.e. Line 4).
One can realise that the 7 × 7 intersections are controlled by one
eight-phase (A), 12 six-phase (B), and 36 two-phase (C) transitions. 4 Simulation of the control strategies for traffic
It is worthy of note that the initial state of the traffic light control
system is assumed the traffic lights all on red states. Fig. 2a shows congestion
the rules of the eight-phase transitions. One can refer to our Once the vehicles drop in traffic congestion problems, VISSIM is
previous work [22], we leave out the details. able to catch the vehicles behaviour that is plausible seems likely to

2 IET Intell. Transp. Syst.

© The Institution of Engineering and Technology 2016
 ii. the length of each East–West road is 1650 m, each block length
is 200 m;
iii. the length of each South–North road is 1400 m, each block
length is 175 m;
iv. the eight-phase transitions of the traffic lights located at
intersection A with a periodic time 180 s (s), the six-phase is at
intersection B with a periodic time 120 s, and the two-phase is
at intersection C with a periodic time 90 s;
v. the proportion of vehicles starting at the main road (i.e. see
Table 1 and Fig. 5): the proportion of vehicles travelling in the
going straight direction is 50% (i.e. path 1 in Fig. 5), the right
turning to the main road is 4.28% (path 2), the right turning to
the minor roads is 2.14% (path 3), the left turning to the main
road is 2.86% (path 4), and the left turning to the minor roads
is 1.43% (path 5); and
vi. the proportion of vehicles starting at the minor road: the
proportion of vehicles travelling in the going straight direction
is 50% (path 6), the right turning to the main road is 3.44%
(path 7), the right turning to the minor roads is 1.72% (path 8),
the left turning to the main road is 3.44% (path 9), and the left
turning to the minor roads is 1.72% (path 10).

According to the definition of the above parameters, we

construct Tables 2 and 3 to list the detailed information of the
vehicles travelling distance [i.e. scale in metres (m)] and the
number of vehicles travelling in each exit, respectively. For
example, in Table 2, if a vehicle starts at entrance I (i.e. see Fig. 4)
and it exits the exit 1, then the vehicle travelling distance is 1600
m.
In this simulation work, we use 2000 vehicles to represent for a
less traffic flow situation, 4000 are for a middle one and 6000 are
for a heavy one. We list the number of vehicles in Table 3. For
convenience, the original traffic light control rules is called Normal
policy, the proposed multiline strategies of Long et al. [4] is called
Long policy and our proposed traffic congestion control policy is
called Huang policy. One knows that Normal policy is represented
that no any additional control strategy is applied in except its
regular traffic light rules. Here, we define the number of vehicles
from the entrances I, J, and K to exits 1–7 as I1–I7, J1–J7, and K1–
K7, respectively. Additionally, we define I1–7Normal (J1–7Normal, K1–
7Normal), I1–7Long (J1–7Long, K1–7 Long), and I1–7Huang (J1–7Huang, K1–
7Huang), as Normal, Long, or Huang policy is applied in exits 1–7
from the entrances I, J, and K, respectively. For example, 2000
vehicles starting at entrance I (main road), I1Normal = 2000 × 50% =
1000, I2Long = 2000 × 2.14% = 42.8 ≅ 43, and I3Huang = 2000 ×
1.43% = 28.6 ≅ 29. On the other hand, considering the real-world
traffic flow, the vehicles passing through the minor road is usually
less than the main road. Therefore, in Table 3, we assume the
number of vehicles is 2000 × (2/3) = 1333 those starting at the
entrances J and K. Here, J1Normal = 1333 × 3.44% = 45.8 ≅ 46,
J2Long = 1333 × 1.72% = 22.9 ≅ 23, and J1Huang = K1Huang = 0
because all right turn traffic lights are banned in the main road. It
hints no way to exit 1 from the minor road (J or K) by using Huang
policy. It is worthy of note that all items (Table 3) multiple two or
three times, one can obtain the information of 4000 or 6000
vehicles for each item, respectively. Note that there are four main
road entrances and 24 minor road entrances in Fig. 4. Here, I is the
Fig. 2 Rules of the eight-phase transitions
one of the four main road entrances, J/K is one of the 24 minor
(a) Models of eight-phase transition, (b) Intersection of the six-phase transition, (c)
road entrances. It is worthy of emphasis that all the number of
Models of the two-phase transition
vehicles in Table 3 is reference data in VISSIM.
be true. Subsequence, VISSIM based is employed to simulate
traffic congestion and evaluate the efficiency of our traffic 5 Simulation results and discussion
congestion control policy. In this section, a 7 × 7 urban traffic We try to simulate a traffic congestion zone in our proposed urban
network that is shown in Fig. 4 is used to evaluate our congestion traffic network that includes three kinds of intersections A, B, and
control strategy. C by using the VISSIM. However, the tool of VISSIM does not
The detailed information of the parameters for the VISSIM is address any function for the traffic congestion and does not support
set as follows: addition traffic light models (i.e. GR, GS, and GL), except a regular
traffic light (i.e. R, Y, and G). Consequently, we employ four stop
i. the main road is with three lanes and the minor road is with signs to enforce the passing vehicles to reduce their speed such that
two lanes; the vehicles form traffic congestion on a road. Additionally, in
order to meet the different traffic light phase (i.e. three

IET Intell. Transp. Syst. 3

© The Institution of Engineering and Technology 2016
 transition (i.e. see Fig. 2a), we need to create 12 traffic light
models in VISSIM system.
As shown in Fig. 1, a 7 × 7 urban traffic network is used to
evaluate our congestion control strategy. The urban traffic network
consists of three type crossroads; one A, 12 B’s, and 36 C’s (see
Fig. 3b). In this paper, we employ the notations ‘-·-·-·-·‘, ‘-------’,
and ‘¾¾’ to represent the vehicles travelling time under Normal,
Long, and Huang policies, respectively. The suffix ‘Save’ and ‘Tave’
are represented the average of travelling time for each segment and
the average of travelling time for total exits, respectively. Three
cases are simulated and discussed as follows.

5.1 Three case simulation results

5.1.1 Case I: low traffic flow network: In this case, we use a low
traffic flow network with 2000 vehicles/h (i.e. main roads) to
simulate the congestion of the urban traffic network. It can be
divided into three segments I1–I7, J1–J7, and K1–K7. For
convenience, we figure out the average of travelling time for each
segment in Fig. 6. For example, NormalSave, LongSave, and
HuangSave are represented the average of travelling time for each
segment by Normal, Long, and Huang policies, respectively.
Similarly, NormalTave, LongTave, and HuangTave are for the average
of travelling time for all exits.
According to Fig. 6, the performance of Long policy is the best
among Normal and Huang policies in the first segment (i.e. I1–I7).
The main reason is that there are few vehicles on the main road due
to ahead ban applied by Long strategy. Since all vehicles are
distributed to the other ways such that they do not need to pass
through the congestion zone. Consequently, those vehicles can save
much travel time. On the contrary, 50% vehicles are allowed to go
straight on the main road under Normal and Huang policies such
that those vehicles need much travel time to pass through the
congestion zone. However, our control policy is better than Normal
policy.
In the second segment (i.e. J1–J7), though the performance of
the three control policies are very close but our proposed control
policy is the best among Normal and Long policies. It is worthy of
mention that the average of travelling time in the second segment is
with shortest one because the entrance J is located on the right side
of the main road. It hints a great quantity vehicles need to make
their left turning across the main road. On the contrary, the
entrance K is located on the left side of the main road. Hence, more
Fig. 3 Multiline control strategy vehicles have to cross the main road such that the vehicles need
(a) Long et al. [4] multiline control strategy, (b) Our new traffic control strategy more time to get their place.
In the third segment (i.e. K1–K7), one can point that our
proposed control policy is the best among Long and Normal
policies. It is worthy of note that the performance of Long is better
than Normal.
In sum, according to NormalSave, LongSave, and HuangSave in
Fig. 6, one can realise that once ahead ban is applied in this case,
the motorists only can save their time-consuming cost in the first
segment. However, the others have to pay more time cost in the
third segment. It states that new traffic jam might be happening in
this area. On the contrary, i.e. the ahead ban is released; the
motorists need to be more patient than the above one. However, the
others can save more time-consuming cost in the second and third
segments. Therefore, our proposed traffic control policy is more
reasonable than Long. The key reason is that the motorists who
pass through the congestion zone have to pay more time-
consuming cost than no blocking area. Besides, according to
NormalTave, LongTave, and HuangTave in Fig. 6, we can conclude
that our proposed traffic control policy is the best among the others
two.

5.1.2 Case II: middle traffic flow network: In this case, we use a
Fig. 4 Traffic network with 7 × 7 intersections model created by VISSIM middle traffic flow network with 4000 vehicles/h (i.e. main roads)
to simulate the congestion of the urban traffic network.
intersections A, B, and C) transitions and their phases alternating In the first segment (i.e. I1–I7), the performance of Long is the
order, we construct the eight-phase transition and the six-phase, best among Normal and Huang. The reason is as the same as the
respectively. According to the definition of the eight-phase Case I. One can see that exit 3 is located below the congestion

4 IET Intell. Transp. Syst.

© The Institution of Engineering and Technology 2016
 area. On the contrary, i.e. the ahead ban is released, the motorists
can save more time-consuming cost in the second and third
segments. According to NormalTave, LongTave, and HuangTave in
Fig. 7, we can conclude that our proposed traffic control policy is
more suitable to be applied in the middle traffic flow network than
Long because HuangTave is with the lowest consuming time costs.

5.1.3 Case III: heavy traffic flow network: In this case, we use a
heavy traffic flow network with 6000 vehicles/h to simulate the
congestion of the urban traffic network.
Fig. 8 shows this experiment result is similar as the above cases.
Notably, our proposed control policy is with the optimal average of
travelling time of all exits. Here, two simulation results can support
the reasons: (i) HuangSave is closer with HuangTave than the others
for every segment. (ii) HuangTave is with the lowest of average of
travelling time. Hence, we can conclude that our proposed traffic
control policy is the best one in the heavy traffic flow network than
the others. The important reason is that our new proposed policy
not only alleviates successfully the congestion problem but also
prevent the other areas from forming new traffic jam problems.
Fig. 5 Percentage of vehicles travelling in the urban traffic network
6 Conclusion
zone. This factor make a sharp point I3 being formed in. Note that a
It is well known that vehicle road traffic congestions are one of the
sharp point I3 being formed in. It states lots of vehicles need to
common phenomena that motorists have to face in their trip. In this
cross the congestion zone from North to South due to exit 3 located paper, we proposed a new control strategy for solving the problem
below the congestion zone. In the second segment (i.e. J1–J7), of traffic congestion that is different with Long et al. [4] in
though our system performance is close with Normal policy but releasing the limitation of the ahead bans. We employ a tool of
ours is still the best among the other two. In the third segment (i.e. VISSIM to simulate and analyse a congestion zone happened in an
K1–K7), one can point that our proposed control policy is also the urban traffic network. For fitting the situation of the real-world
best policy. It is worthy of note that the average of the travelling traffic network case, we extend the link channelisation road model
time by using Long is higher than Normal policy. [23] to three kinds of roads and a 7 × 7 urban traffic network with
In sum, according to NormalSave, LongSave, and HuangSave in three kinds of intersections A, B, and C is employed to evaluate our
Fig. 7, one can realise that once ahead ban is applied in this case, congestion control policy. The applications of VISSIM to six-phase
the motorists only can save some time-consuming cost than and eight-phase traffic signal control systems are performed. As far
Huang's in the first segment. In the meantime, the other ways as we know, this is the first work that uses VISSIM to construct the
motorists have to pay most time-consuming cost in the third models of the six-phase and eight-phase traffic light control
segment. It states that new traffic jam might be happening in this systems. In addition, we demonstrate three scenario cases low,

Table 1 Proportion of vehicles starting at main/minor road

Exit
Ahead Right turn to main road, % Right turn to minor road, % Left turn to main road, % Left turn to minor road, %
Entrance Direction, %
main road 50 4.28 2.14 2.86 1.43
minor road 50 3.44 1.72 3.44 1.72

Table 2 Travelling distant from the entrances I, J, K to exits 1–7

Exit
1 2 3 4 5 6 7
Entrance
I 1600 1950 1950 1500 1900 1500 1900
J 1775 1775 2125 1325 1725 1675 2075
K 1775 2125 1775 1675 2075 1325 1725

Table 3 Number of vehicles in travelling from the entrances I, J, K to exits 1–7

Main roads = 2000 Policy 1 2 3 4 5 6 7
Minor roads = 1333
I Normal 1000 43 29 86 43 57 28
Long 0 86 56 172 86 114 56
Huang 1000 43 29 86 43 57 28
J Normal 46 23 23 45 23 45 23
Long 0 24 24 48 24 48 24
Huang 0 24 24 48 24 48 24
K Normal 46 23 23 45 23 45 23
Long 0 24 24 48 24 48 24
Huang 0 24 24 48 24 48 24

IET Intell. Transp. Syst. 5

© The Institution of Engineering and Technology 2016
 problem but also prevents the other areas from forming new traffic
jam problems. However, new traffic jams might be happening in
the other areas if one used a strong ban method. To summarise, this
paper has new contributions:

i. It has demonstrated how to use VISSIM to simulate control

strategies for the traffic congestion problems.
ii. It performs a novel control policy that is able to avoid the
traffic congest dispersing and ways to avoid traffic congestions
being happened.
iii. It identifies our control strategy is more efficient in the average
of travelling time than Long et al. [4].

On the other hand, there are many other traffic congestion control
policies that can alleviate the congestion problems such as TIS [6]
and traffic signal control [12, 17]. Hence, our future work aimed to
design a traffic signal light controller that is able to interrupt the
duration of the normal traffic light phases when traffic congestion
is happening.

Fig. 6 Average of travelling time for case I

7 Acknowledgment
This work was supported in part by the Ministry of Science and
Technology of Taiwan, under Grant MOST 104-2221-E-197-010-
MY and supported in part by the National Natural Science
Foundation of China Under Grant 61374066 and 61134007, the
Open Research Project of the State Key Laboratory of Industrial
Control Technology, Zhejiang University, China (No. ICT
1600261), National Key Basic Research Program of China under
Grant 2013CB035406 and the Research Project of the State Key
Laboratory of Industrial Control Technology, Zhejiang University,
China.

8 References
[1] Wright, C., Roberg, P.: ‘The conceptual structure of traffic jams’, Transp.
Policy, 1998, 5, pp. 23–35
[2] Roberg, P.: ‘Distributed strategy for eliminating incident-based traffic jams
from urban networks’, Traffic Eng. Control, 1995, 36, (6), pp. 348–355
[3] Roberg, -Orenstein, P., Abbess, C.R., Wright, C.: ‘Traffic jam simulation’, J.
Maps, 2, 2007, pp. 107–121
[4] Long, J., Gao, Z., Orenstein, P., et al.: ‘Control strategies for dispersing
incident-based traffic jams in two-way grid networks’, IEEE Trans. Intell.
Transp. Syst., 2012, 13, pp. 469–481
Fig. 7 Average of travelling time for case II [5] Figueiredo, L., Jesus, I., Tenreiro Machado, J.A., et al.: ‘Towards the
development of intelligent transportation systems’. Proc. of IEEE Intelligent
Transportation Systems, August 2001, pp. 1206–1211
[6] Sussman, J.: ‘Perspectives on intelligent transportation systems (ITS)’
(Springer-Verlag, New York, 2005)
[7] Papageorgiou, M., Diakaki, C., Dinopoulou, V., et al.: ‘Review of road traffic
control strategies’, Proc. IEEE, 2003, 91, (12), pp. 2043–2067
[8] Febbraro, A.D., Giglio, D., Sacco, N.: ‘Urban traffic control structure based
on hybrid Petri nets’, IEEE Trans. Intell. Transp. Syst., 2004, 5, pp. 224–237
[9] Hruz, B., Zhou, M.: ‘Modeling and control of discrete event dynamic systems’
(Springer, London, UK, 2007)
[10] Huang, Y.S., Chung, T.H.: ‘Modeling and analysis of air traffic control
systems using hierarchical timed coloured Petri nets’, Trans. Inst. Meas.
Control, 2011, 33, (1), pp. 30–49
[11] Huang, Y.S., Chung, T.H.: ‘Modeling and analysis of urban traffic lights
control systems using timed CP-nets’, J. Inf. Sci. Eng., 2008, 24, (3), pp. 875–
890
[12] Huang, Y.S., Weng, Y.S., Zhou, M.: ‘Modular design of urban traffic light
control systems based on synchronized timed Petri nets’, IEEE Trans. Intell.
Transp. Syst., 2014, 15, (2), pp. 530–539
[13] Huang, Y.S., Su, P.J.: ‘Modeling and analysis of traffic light control systems’,
IET Control Theory Appl., 2009, 3, (3), pp. 340–350
[14] Huang, Y.S., Chung, T.H.: ‘Modeling and analysis urban traffic light control
systems’, J. Chin. Inst. Eng., 2009, 32, pp. 85–95
[15] List, G.F., Cetin, M.: ‘Modeling traffic signal control using Petri nets’, IEEE
Trans. Intell. Transp. Syst., 2004, 5, pp. 177–187
Fig. 8 Average of travelling time for case III [16] Dutilleul, S.C., Deffossez, F., Bon, P.: ‘Safety requirements and p-time Petri
nets: a level crossing case study’. IMACS-IEEE Multiconf. Computational
Engineering in Systems Applications, Beijing, China, October 2006, pp.
middle, and heavy traffic flow works to evaluate Normal, Long, 1118–1123
and our congestion control policies. According to the experiment [17] Huang, Y.S., Weng, Y.S., Zhou, M.: ‘Critical scenarios and their identification
results, we found that the proposed control policy can indeed in parallel railroad level crossing traffic control systems’, IEEE Trans. Intell.
alleviate successfully the congestion problem efficiently whatever Transp. Syst., 2011, 11, (4), pp. 968–977
[18] Chin, D.C., Spall, J.C., Smith, R.H.: ‘Evaluation of system-wide traffic signal
in the above mentioned cases. Hence, we can conclude that our control using stochastic optimization and neural networks’. Proc. of ACC 99,
proposed traffic control policy is the best one in this urban traffic June 1999, vol. 3, pp. 2188–2194
network than the others. The important reason is that our new [19] Al-Deek, H., Lochrane, T.W.P., Srinivasa Ravi Chandra, C.V., et al.:
proposed policy not only alleviates successfully the congestion ‘Diversion during unexpected congestion on toll roads: the role of traffic

6 IET Intell. Transp. Syst.

© The Institution of Engineering and Technology 2016
 information displayed on dynamic message signs’, IET Intell. Transp. Syst., [22] Huang, Y.S.: ‘Design of traffic light control system using statecharts’,
2012, 6, (2), pp. 97–106 Comput. J., 2006, 49, pp. 634–649
[20] Milanés, V., Godoy, J., Villagrá, J., et al.: ‘Automated on-ramp merging [23] Long, J., Gao, Z., Zhao, X., et al.: ‘Urban traffic jam simulation based on the
system for congested traffic situations’, IEEE Trans. Intell. Transp. Syst., cell transmission model’, Netw. Spat. Econ., 2011, 11, pp. 43–64
2011, 12, pp. 500–508
[21] Zhang, P., Wu, C.-X., Wong, S.C.: ‘A semi-discrete model and its approach to
a solution for a wide moving jam in traffic flow’, Phys. A, Stat. Mech. Appl.,
2012, 391, (3), pp. 456–463

IET Intell. Transp. Syst. 7

© The Institution of Engineering and Technology 2016