Professional Documents
Culture Documents
Abstract-one of the major issues of our daily life is traffic proposed system also identifies the effects of improvements
congestion. To help solve the traffic congestion problem, a on other roads. The system can be integrated with other
traffic advisory system is proposed to identify the threshold measures of fighting traffic related issues like traffic
capacity of a road segment. Threshold capacity measures the signaling and delay estimation etc.
number of vehicles which can be allowed to pass over a certain This paper is organized as follows : section II describes
road segment without causing congestion. Threshold capacity the proposed approach , section III presents the results and
is used for suggesting improvements of road segments. discussions and conclusion is briefed in section IV.
Congestion information can subsequently be provided to
traffic management authorities for improvement of the II. PROPOSED ApPROACH
conditions of road segments. The traffic advisory system is
designed using fuzzy logic to determine the threshold capacity The proposed system requires (1) determination of
of a road segment. To estimate the tendency of traffic flow at threshold capacity, (2) calculations of current flow and
different junctions a neural network is designed. Furthermore tendency of traffic flow at traffic junctions. Fuzzy logic is
an algorithm is proposed, which measures the effects of these used for determination of threshold capacity for a certain
improvements. Testing of proposed system is accomplished road segment, current flow is measured using sensors and a
through simulation. neural network is designed for the determination of tendency
of flow.
Keywords: traffic congestion; fuzzy logic; neural network
1. INTRODUCTION
Traffic congestion problems are commonly encountered
in our daily life. Congestion may occur due to various
conditions such as inadequate number of lanes, rough road ~ -,-- ...
surface and poor road visibility. The consequences of traffic
congestion are known to be rather expensive .
By improving road conditions, traffic congestion
problems can be solved to get smooth traffic flow on roads.
Traditionally road conditions are improved on the basis of
experience of transportation experts. Considerable work is
done on delay estimation [1, 2] [6], using traffic signaling
techniques [3,4] and traffic management techniques [5, 7] &
[8]. However, there is still a great deal of research effort
required on improvement of traffic conditions .
In this research we propose a method for identifying
roads which need to be improved as well as the conditions
required to improve those roads. In this way traffic
management authorities are made capable of improving
conditions in a systematic manner. To develop our proposed Eff ects of Improv ements
193
Authorized licensed use limited to: NED UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on November 26,2020 at 08:04:02 UTC from IEEE Xplore. Restrictions apply.
A. Determination ofThreshold Capacity a) Ranges of fuzzy inputs: First input of the fuzzy
We define threshold capacity as the number of vehicles inference system is the number of lanes and its range is from
which can pass through a road segment without causing zero to five. Second one is road surface and its range start
traffic congestion, i.e. smooth flow of traffic is maintained. from zero up to one. Third and final input of the FIS is
. Fuzzy logic is used to determine the threshold capacity, visibility and its range is from zero to one.
In that road conditions are input variables for fuzzy inference b) Membership functions of fuzzy inputs: Number of
system and the output of fuzzy inference system is threshold lanes consists of three membership functions i.e. minimum
capacity. In coming subsections we have discussed the number of lanes ranging from (1, 2.5), average number of
conditions or inputs of the fuzzy inference system which lanes with a range of (1, 5) and maximum number of lanes
include number of lanes, road surface and visibility. which has a range from (3 to 5). Road surface is divided into
1) Number of lanes of roads: One important parameter three membership functions i.e. rough, its range is from (0.5,
influencing traffic congestion is the structure of the road. 1), moderate with a range from (0, 1) and smooth with a
Road structure is described by the number of lanes on that range of (0, 0.5). Visibility has also three membership
particular road. As the numbers of lanes are increased, traffic functions. These are bad visibility and its range is from (0.5,
congestion level improves. 1), average visibility with a range from (0, 1) and good
A road where tendency of traffic volume is large should visibility has its range from (0, 0.5). We have constructed the
contain more number of lanes, which will help the flow of membership function by using triangular shape. It has well
traffic and will negate congestion. Number of lanes, as fuzzy defined boundaries, faster computational speed, easy design
input is provided by three fuzzy sets. These are minimum and demands less memory.
number of lanes, average number of lanes and maximum c) Graphical and mathematical representation of
number of lanes. In the proposed method we limit the linguistic variables: In this subsection we present some
maximum number of lanes up to five. A road containing only graphical and mathematical models which are developed in
one lane is considered as minimum number of lane road a fuzzy inference system for threshold capacity determination.
road containing two or three lanes is considered as average There are three fuzzy sets i.e. minimum, average and
number of lane road. Similarly a road containing four or five maximum. The ranges are also shown such as ranges of the
lanes is considered as maximum number of lane road. Minimum start from 1 to 2.5, Average from 1 to 5 and
2) Road Surface: Road surface depicts the condition of Maximum from 3 to 5. Following is the mathematical model
road. There are chances of a road being in a bad condition for fuzzy input number of lanes, where Jl describes the
even if it contains maximum number of lanes. For good road membership of element x. Mathematical models for each of
condition the surface of road should be smooth and not the other fuzzy sets are developed in similar fashion.
spoiled. Rough and uneven roads are always subject to traffic I for x ~ 1
congestion, and ill traffic flow. For better traffic flow, the
surface of roads must be smooth. Our fuzzy inference system J..lMinimum = 2.5 - x
{ for 1 ~ x ~ 2.5
is considering road surface as an input variable. It can be 1.5
rough, average and smooth and these considered as fuzzy
x-I
sets of this fuzzy input. for 1 ~ x ~ 3
3) Visibility: The visibility factor is considered only in 2
J..l Average
night mode because at day time, visibility of a road is 5-x
considered as always good. In night mode the visibility is for 3 ~ x ~ 5
very effective for flow maintenance. Visibility in context of
2
{
this study is the level of a visibility of a certain road in night
mode over a period of time (months, years) not a particular
~ for 3 ~ x ~ 5
day. Clouds, dust etc come into play only in a system which
u Maximum = 2
is on a daily basis. Traffic flow can be affected in case of 1 for x ~ 5
poor condition of light on a road. The visibility of a road can
Membership function plots
be improved by providing extra light poles at roads. The
determination of threshold capacity is based on visibility
only in evening, night and morning. To implement the factor
of visibility we have split it into three fuzzy sets which are
poor visibility, average visibility and good visibility.
All of these inputs (number of lanes, road surface and
visibility) are used to design a rule base Fuzzy Inference
System (FIS).
4) Development of FIS for threshold capacity: This
section describes how to construct the fuzzy inference
system in MATLAB by using the fuzzy logic tool box. FIS 4 5 6 7 10
consists of five steps, as discussed below. output variable "Optimalcapac~y"
194
Authorized licensed use limited to: NED UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on November 26,2020 at 08:04:02 UTC from IEEE Xplore. Restrictions apply.
Figure 2 show the graphical representation of the fuzzy AB. Now we have to determine how many vehicles are
inference system for output i.e. threshold capacity. It moving to what road segment after leaving junction B. If we
consists of five fuzzy sets that include very small , small, suppose that there are three roads leaving junction B, i.e. BC,
medium, high and very high. The ranges of these sets are BD and BE, then some of those extra vehicles will move to
also shown in figure 2, such as ranges of very small starts BC, some will move to BD and some will move to BE. This
from 1 to 3.25; small starts from 1 to 5.5, medium from 3.25 is in fact the tendency of traffic flow at junctions which is
to 7.75, high from 5.5 tolO and very high from 7.75 to 10. very important to be known .
moving straight, turning right and turning left from a Hidden lilyet" Re1:OPunend",tio n
195
Authorized licensed use limited to: NED UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on November 26,2020 at 08:04:02 UTC from IEEE Xplore. Restrictions apply.
network. Furthermore the trammg data used for the Figure 7 is depicting the training, validation and testing
implementation of the neural network is a set of one of results as well as their comparison with target vector.
hundred instances. One can train a similar network on more
or less one hundred instances of data. So it is not necessary " I ,
to train the network upon a specific set of data instances.
We have found satisfactory results on training data of one
hundred instances. T~ningOutputsvsTar9ets,R=O ,:a:67 YalidationOlJtputs rs. TarQets ,R=O~$77 TestOutputsvs Targets,R=0 4:!&1
Another important consideration of the problem of ~~T;::::!'~±::::::::::;-I :6 ;;; :6rr=::::==!i:-=ii
~ 0 Training DataPoints 1. 0 ValidatiooD.taPomls 2: 0 TestOat,poinl'!
tendency of flow at junctions is the time of day which is ~ lJ - BestlinearH ~ lJ - Besl Uoe.r Fd t lJ - SestlinearFa
~ 25 ·······A=T
considered in a training set. The proposed neural network is B ~
~
······· Aoj '<t .··.··.A=T
not specific to the training data of a specific time of day. For ~ 2J !JJ
real implementation of this research the training data should "i tl / o ..",.",,·
be collected for a fixed interval. Usually this interval is fixed J 10 0 r-
ADVISORY_ALGO(u ,v,OptimaICapacity,TendencyOfFlow,
Changedoptimalcapacity)
Q=empty
ParentQ = empty
Diff(u,v) = Changedoptimalcapacity(u,v) - OptimaICapacity(u,v)
Set threshold
Set color all junctions to write
Color{ul = black
Figure 5. Definition of input vector p Push v in Q
Push u in ParentQ
b) Target Vector: Target vector is a matrix consisting of While Q is not empty and diff > threshold
multiple values of target of the neural network. It shows the Pop V from Q
Pop u from ParentQ
dispersion of vehicles from the particular junction.
For each adjacent junction w of v
Dispersion means the percentage of vehicles moving left, If color [w) = white
straight and right. In this way there are three outputs of one Diff (v,w) = Diff(u ,v) * (TendencyOfFlow (u,v,w) /100)
input. Target vector consisting of hundred columns is CurrentFlow (v,w) = CurrentFlow (v.w) + diff(v .w)
constructed as shown in figure 6. Each column represents Pushwin Q
Push v in ParentQ
three outputs.
Color[v) = black
196
Authorized licensed use limited to: NED UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on November 26,2020 at 08:04:02 UTC from IEEE Xplore. Restrictions apply.
traffic which is supposed to run from junction 3 to these two
0 - - -0====0 - -0 roads. Similarly this effect will spread to upcoming road as
826/838 828/814
I
''''~i 'I"~'
I
~"~i ~"'I'
well.
814/002 8121826
In this research we have proposed a traffic management
system called a traffic advisory system. This proposed
system is an innovative idea to assist traffic management
t OD , . n<l'" •• • f .... IInk '.j.- ~3J 81H847
= 0 ::::: 0) ==:::::::= 0
848/820 authorities. It will help them to improve the conditions where
i- !n>p,.'...' n "'nOQ. O(hnl l.l- >3f (' .. ~ 1.l- >3.
it is required. Importantly it tells the authorities about the
I :: Ncd>trofL-. -)-- ::;~~.
: Ro.-JS>.rl>o: e - -r:-. l.=I orV..t4ly
effect of improvements in traffic conditions on other roads.
These improvements will be effective in maintaining an
• " " ":~:""""D Levdorv..,.,. 1
ThrtlholdC., 1IC1f
r ,·-
~I 'Ilm b.IdC.. ..". ~~
...~
appropriate and proper flow of traffic at different roads of a
certain city.
Figure 8. Current flow/threshold capacity (link 4, 3) The proposed system is constructed using soft computing
techniques i.e. fuzzy logic and neural network. It
In figure 8 one can easily identify that the road (4, 3) is incorporates the benefits of both fuzzy and neural approaches
having the threshold capacity of 802 but the current flow is to solve the problem. The results and validations of proposed
843. It means that improvements are required at road link (4, system are provided using simulations.
3). It shows that the tendency of traffic flow at junction 3 As a future direction it would be interesting to integrate
(from 4 to 3) is 62 percent left and 38 percent straight. other traffic related solutions. One entity of integration with
To improve the conditions at road link (4, 3) first we take a the proposed system is signal controllers.
look at the conditions of (4, 3). Though the conditions of(4,
3) are quite satisfactory, but not enough for the traffic that is
REFERENCES
occupied on road segment (4,3). We improve the conditions
[I) Kerner , B.S, C. Demir , R.G. Herrtwich, S.L. Klenov , H. Rehborn , M.
by constructing an extra lane as well as by repairing the Aleksic , A.Haug and D.C. AG, "Traffic state detection with floating
surface of the road. According to the improved conditions, car data in road networks", Proceeding of the 8Ch international IEEE
the inputs for threshold capacity are changed (Lanes=5, conerence on intelligent transportation system , pp. 700-705, 2005.
Surface=0.85, visibility remains unchanged). After (2) Qiao, F, Yi P, H. Yang, S. Devarakonda, "Fuzzy logic based
improvements threshold capacity changes to 954 which is intersection delay estimation", Mathematical and computer modeling,
Volume 36. pp. 1425-1434,2002.
appropriate for the current state of traffic.
(3) Madhavan, .B.N and J.cai, "A fuzzy logic controller for isolated
signalized intersection with traffic abnormalities considered" ,
Intelligent vehicles sympos ium IEEE. pp. 1229-1233,2007.
0- - 8====8 - -0 (4) Niittymaki,.J and K. Ville, "Traffic signal controller based on fuzzy
I I logic", System, Man and cybernetics IEEE international conference,
826/838 834/814
197
Authorized licensed use limited to: NED UNIV OF ENGINEERING AND TECHNOLOGY. Downloaded on November 26,2020 at 08:04:02 UTC from IEEE Xplore. Restrictions apply.