APPLICATION OF FUZZY LOGIC IN TRANSPORT PLANNING

International Journal on Soft Computing (IJSC) Vol.3, No.2, May 2012
DOI: 10.5121/ijsc.2012.3201 1
APPLICATION OF FUZZY LOGIC IN TRANSPORT
PLANNING
Amrita Sarkar1
G.Sahoo2
and U.C.Sahoo3
1
Research Scholar, Department of Information Technology, B.I.T Mesra, Ranchi
amrita.manna@gmail.com
2
Professor and Head,Department of Information Technology, B.I.T, Mesra, Ranchi
gsahoo@bitmesra.ac.in
3
Assistant Professor, Department of Civil Engineerng, I.I.T, Bhabaneswar
ucsahoo@iitbbs.ac.in
ABSTRACT
Fuzzy logic is shown to be a very promising mathematical approach for modelling traffic and
transportation processes characterized by subjectivity, ambiguity, uncertainty and imprecision. The basic
premises of fuzzy logic systems are presented as well as a detailed analysis of fuzzy logic systems developed
to solve various traffic and transportation planning problems. Emphasis is put on the importance of fuzzy
logic systems as universal approximators in solving traffic and transportation problems. This paper
presents an analysis of the results achieved using fuzzy logic to model complex traffic and transportation
processes.
KEYWORDS
Fuzzy Logic, Transportation Planning, Mathematical modelling
1. INTRODUCTION
Transportation planning is a wide human-oriented field with diverse and challenging problems
waiting to be solved. Characteristics and performances of transport systems – services, costs,
infrastructures, vehicles and control systems are usually defined on the basis of quantitative
evaluation of their main effects. Most of the transport decisions take place under imprecision,
uncertainty and partial truth. Some objectives and constraints are often difficult to be measured by
crisp values. Traditional analytical techniques were found to be non-effective when dealing with
problems in which the dependencies between variables are too complex or ill-defined. Moreover,
hard computing models can not deal effectively with the transport decision-makers’ ambiguities
and uncertainties.
In the past several decades a variety of deterministic and stochastic models have been developed
to solve complex traffic and transportation engineering problems. These mathematical models use
different formulae and equations to solve such problems. During their education and training,
engineers are most often directed to the use of exclusively objective knowledge (formulae and
equations). However, when solving real-life engineering problems, linguistic information is often
encountered that is frequently hard to quantify using ‘classical’ mathematical techniques. This
linguistic information represents subjective knowledge (linguistic information).

2
The primary goal of this paper is the applications of fuzzy logic in traffic and transportation
engineering, and to indicate the directions for future research in this area.
2. FUZZY LOGIC SYSTEMS FOR TRANSPORTATION
Basic results linked to the development of fuzzy logic date back to Zadeh (1973) [51] and
Mamdani and Assilian (1975) [24]. Introducing a concept he called ‘Approximate Reasoning’,
Zadeh successfully showed that vague logical statements enable the formation of algorithms that
can use vague data to derive vague inferences. Zadeh assumed his approach would be beneficial
above all in the study of complex humanistic systems.
What is a fuzzy logic system? Mendel (1995) [25] explains the concept of a fuzzy logic system-a
(FLS) as follows: ‘In general a FLS is a nonlinear mapping of an input data (feature) vector into a
scalar output (the vector output case decomposes into a collection of independent multi-input/
single-output systems)’.
Pappis and Mamdani (1977) [28] published the first paper in which a practical traffic and
transportation problem was solved using fuzzy logic. In the mid- and late-1980s, a group of
Japanese authors made a significant contribution to fuzzy set theory applications in traffic and
transportation. Nakatsuyama et al. (1983) [27], Sugeno and Nishida (1985) [36] and particularly
Sasaki and Akiyama, (1986, 1987, 1988) [33, 34, 35] solved complex traffic and transportation
problems indicating the great potential of using fuzzy set theory techniques. At the end of the
1980s and beginning of the 1990s, the fuzzy set theory in traffic and transportation became
extensively used at American universities. The pioneering work of the research team of the
University of Delaware deserves special attention, headed by Professor Shinya Kikuchi
(Chakroborthy, 1990 [5]; Chakroborthy and Kikuchi, 1990 [6]; Perincherry, 1990 [29];
Perincherry and Kikuchi, 1990 [30]; TeodorovicÂ and Kikuchi, (1990, 1991) [41, 42]; Kikuchi et
al.,(1991, 1993) [19, 20]; Kikuchi, 1992 [18]). In the early and mid-1990s, interest in fuzzy logic
applications in traffic and transportation increased in other world universities. Different traffic
and transportation problems successfully solved using fuzzy set theory techniques were presented
in the works of Chen et al. (1990) [8], Tzeng et al. (1996) [44], Lotan and Koutsopoulos
(1993a,b) [22, 23], Xu and Chan, 1993a,b) [49, 50], TeodorovicÂ and BabicÂ (1993) [38],
Akiyama and Shao (1993) [1], Chang and Shyu (1993) [9], Chanas et al. (1993) [7], Akiyama
amd Yamanishi (1993) [2], Deb (1993) [10], Perkinson (1994) [31], Hsiao et al. (1994) [12],
VukadinovicÂ and TeodorovicÂ (1994) [47], TeodorovicÂ et al. (1994) [39], TeodorovicÂ
(1994) [37], TeodorovicÂ and KalicÂ (1995) [40], MilosavljevicÂ et al. (1996) [26], Teodorovic
and Pavkovic (1996) [43] and Tzeng et al. (1996) [44].
Let us try to explain why fuzzy logic systems were developed for transportation engineering.
Great many of the problems in the field of transportation planning and traffic control are often ill
defined, ambiguous and vague. As already mentioned, many traffic and transportation problems,
phenomena and parameters are characterized by subjectivity. It is hard to disregard the fact that
subjective judgment is present in problems dealing with the choice of route, mode of
transportation and carrier, a driver's perceptions and reactions, an established level of service,
defining safety standards, defining criteria to rank alternative transportation plans and projects,
etc. It should also be emphasized that both deterministic and stochastic models that have been
developed to solve a variety of complex traffic and transportation problems are characterized by
mathematics based on binary logic. Without denying the importance of binary logic as the basis
for the development of many scientific disciplines and technology leading to the prosperity of
man's society, we must note that it cannot deal effectively with passengers’, dispatchers’ or
drivers’ feelings of uncertainty, vagueness and ambiguity. Since the fuzzy set theory recognizes
the vague boundary that exists in some sets, different fuzzy set theory techniques need to be used

3
in order to properly model traffic and transportation problems characterized by ambiguity,
subjectivity and uncertainty.Some applications of Fuzzy Logic in transport planning are discussed
below.
2.1. Trip generation
Trip generation constitutes the first stage in the traditional transportation planning process. Trip
generation defines that how many people want to go out of their homes and what their purposes
are. Researchers emphasized that more non-congestion trip generation models have better
prediction and adaptation ability. So in many years to validate different ideas, experts have
invented variety of models and they always try to estimate the number of generated trips in a
better method helping newer model presentation. Also they try to decrease the previous models
problems and defects as much as possible.
Trip generation problem was solved using fuzzy logic by Kalic and Teodorovic (1997a, b) [16,
17]. Fuzzy rule base was generated by learning from numerical examples. For this purpose, the
procedure proposed by Wang and Mendel (1992a, b, c) [46, 47, 48] was used. Firstly, the
available set of data was divided into two subsets: the first was used for generation of the fuzzy
rule base, and the other was intended to be a control data subset. After the fuzzy rule base was
created, the obtained fuzzy system was tested on both subsets of data. The estimate of the number
of trips for the subsets was also determined using artificial neural networks, as well as by multiple
linear regressions. The training of the neural network was done using simulated annealing
technique. After the testing, the fuzzy logic approach proved to give the closest estimate of the
actual number of trips generated in a given area.
One of the most important problems in the field of transportation planning is the problem of
Origin-destination estimation from link counts. In order to decrease the cost of passenger surveys,
traffic count are undertaken on certain links of the transportation network. In the next phase an
origin-destination matrix is estimated from the link counts. In the last two decades a greater
number of models have been developed for origin-destination estimation from link counts. Xu
and Chan (1993a, b) [49, 50] were the first to use the fuzzy set theory techniques when analyzing
the problems arising from the poor quality of link count data. As known, traffic counts are often
subject to errors. Xu and Chan (1993a, b) [49, 50] particularly pointed out the problem of traffic
counts in countries such as China in which mixed urban traffic flow with heavy bicycle volume is
common. In such situations the heavy bicycle volume also makes precise vehicular counts very
difficult. Xu and Chan (1993a, b) [49, 50] estimated an origin-destination matrix with fuzzy
weights.
2.2. Trip distribution
Trip distribution constitutes the second stage in the traditional transportation planning process.
Trip distribution models are used to determine the number of trips between pairs of zones when
the number of trips generated attracted by particular zones is known. Thus, the prediction of trip
distribution involves the prediction of flows in a network regardless of a possible transportation
mode or travel route.
Traffic flows and trip distribution resulted from human choices that are affected by social and
individual variables of the commuters. Due to this fact that human decision making are more
consistent with fuzzy logic in comparison with crisp mathematics, it seems that fuzzy logic could
be a logical tool to map such areas. Modelling a trip distribution system with fuzzy inference
systems would enjoy the exploration of subjective pattern of decision makers. Meanwhile, one of
the most important parts of forecasting procedure is to predict the trips flow between a given

4
Trip Generation
Variables
Trip Attraction
Variables
Input Space Transfer
Function
Output Space
ijT
origin-destination pair. Different models have been developed to forecast trip flows in recent
years. The key to a successful forecasting is to recognize the existing patterns correctly. Trip
distribution problem was also solved using fuzzy logic by Kalic and Teodorovic (1996, 1997a, b
[15, 16, 17] Like in the case of trip generation problem, Kalic and Teodorovic (1996) [15]
generated fuzzy rule base by learning from numerical examples.
A three phases fuzzy inference system (FIS) was proposed by Jassbi et al.(2011) [14] to map
social and demographic variables to total number of trips between origin-destination (OD) pairs.
Fuzzy rule bases in the model are in fact the exploration of transportation experts’ subjective
patterns.
Figure 1. Conceptual model of trip forecasting input-output space mapping
Figure 1 illustrates the conceptual framework of the trip forecasting problem. The main concern
of the problem was to approximate the transfer function that is capable of mapping trip attraction
and generation variables to number of trips between two given regions. As it addressed in
introduction section, Different kinds of transfer functions had been used in this problem. Here, a
fuzzy inference system is used to map input space to output space.
The process of fuzzy inference involves:
i. Defining If-Then rules
ii. Defining membership functions
iii. Appling logical operations
There are two types of fuzzy inference systems that usually can be implemented: Mamdani-type
and Sugenotype. These two types of inference systems vary somewhat in the way outputs are
determined.
Input space is ‘Trip Generation variables and Trip attraction variables’. These are the variables
that have direct impacts on the generation and attraction of trips of any given region. Output
space is ‘the number of trips between two given regions’. The number of trips would be
approximated by a transfer function that is formed based on the relations of trip generation
capabilities of origin and trip attraction capacities of destination.

5
Table 1. Trip generation variables (input space variables)
Input space variables Generation variables
1I Occupation in the residence area
2I Population
3I Residential space
4I Population density
5I Number of households
6I Car ownership rate
7I Average price of one square meter of land
8I Students population
9I Traffic zone space
10I Number of residential buildings
11I Distance to entertainment complexes
Table 2. Trip attraction variables (input space variables)
Input space variables Attraction variables
1I Occupation in the working area
2I Business/ Administrative/Agricultural/Industrial
Land Space
3I Administrative building space
4I Number of Administrative/Business/Industrial
Buildings
5I Schools’ space
6I Number of Students/Schools/Classes

6
7I Number of Universities/Students
8I Number of Retailers
9I Number and Capacity of Cinemas/Mosques/
Exhibitions/
10I Parks/Hospitals
Figure 2. Conceptual fuzzy model: Mapping trip generation and attraction variables to total
number of trips between two given regions
To map input space to output space, different fuzzy scenarios could be considered. Figure 2 above
illustrates the framework of conceptual fuzzy model that is used to forecast the number of trips in
this work. As it illustrated in above, three independent fuzzy inference systems are used. The first
FIS is aimed to map generation variables to number of generated trips by the given region
( iG ).The second FIS is designed to map attraction variables to number of attracted trips by the
given region ( jA ). At the end, the third FIS is designed to combine the output of the first and
second FIS for aggregation of two given regions iG and jA and calculating the final total number
of trips between two regions ( ijT ).
2.3. Modal split
The third phase of transport planning named as modal split is intended to know which modes of
transportation (bus, train etc.) are used by how much fraction of people. Teodorovic and Kalic
(1996) used fuzzy logic to solve the mode choice problem. The authors illustrated the possibilities
of fuzzy logic in solving the mode choice problem using a hypothetical numerical example. The
fuzzy rule base was generated using available numerical data on the differences between the
travel times and travel costs of competitive modes. In other words, learning from examples
generated the fuzzy rule base. The technique proposed by Wang and Mendel (1992a) was used to
generate the fuzzy rule base by learning from examples. There was extremely good agreement
between the results obtained using fuzzy logic and real results.

7
Quadrado and Quadrado (1996) [32] used fuzzy logic to determine the accessibility of different
transportation modes in the Lisbon Metropolitan Area. The authors first pointed out that all
variables used in the ‘classical’ way of calculating accessibility are characterized by fuzziness. A
fuzzy rule base was developed for each mode of transportation. Accessibility was shown as a
percentage between 0 and 100%. One of the typical rules to calculate accessibility reads:
If the number of people involved in movements is VERY BIG and the time used in those
movements is VERY SMALL
Then the accessibility is HYPER BIG
The authors thus put an end to the tradition of treating accessibility as a deterministic concept and
offered the possibility of a better representation of accessibility.
The purpose of modal split step is to determine the share of traffic of vehicle and Public
transportation means in urban area. One of the most well-known model that called “Logit model”
is often used.
))(exp(1
1
rs
tran
rs
rs
CC
qq
−−+
=
θ
This is a sort of nonlinear function model that based on random utility theory. In this formulation,
the share of mass transit between the OD pair (r-s) is determined by the general cost of mass
transit tran
rsC and general cost of car rsC .
The outline of the triangular fuzzy number for each link travel time is illustrated in Figure 8
below. Figure 3 presents an example of a fuzzy number with a triangular member-ship function
corresponding to the label “travel time difference is medium”.
Figure 3. Description of fuzzy travel time difference rs
Transit
rs CCT −=∆
where T∆ is the time difference on each OD pair, Transit
rsC is mass transit travel time, rsC is car
travel time are denoted.
As mentioned earlier, when attempting to represent a mode choice model that considers about
human behavior, the fuzzy reasoning is applied. Here are three simple rules that used in this
model:
Rule 1: If T∆ is big, then Transit TransitP is big.
Rule 2: If T∆ is medium, then Transit TransitP is medium.

8
Rule 3: If T∆ is small, then Transit TransitP is small.
where TransitP is the rate of transit.
Here, in each OD pair, the value of the linguistic variable “big”, “medium” and “small” will be
defined. Each membership function of linguistic variable is summarized in Figure 4.
Figure 4. Membership function
Each OD pair has a range of travel time difference, the ranges of some representative OD pairs is
listed in table 3. Figure 9 shows the membership function of an OD pair and (-30, 30) is the range
of travel time difference for the objective OD pair. In this description, the travel time difference is
a crisp value. This is also the medium rule in fuzzy rule sets that supposed in this model.
Figure 5.The share of mass transit
Figure 5 presents the change of mass transit’s share and the time difference in one OD pair,
assume that the range of travel time difference is (-20, 20). The results obtained from both
models, logit mode choice model and Fuzzy mode choice model are shown in the figure, the
share of mass transit share computed by logit mode choice model changes rapidly when the
absolute value of travel time difference is small, in contrast to that computed by Fuzzy mode
choice model.
For each OD pair, the time different ( ∆ T) between mass transit travel time and car travel time of
iteration number n will be calculated by the combine modal in loop number n –1, until the
condition is met. For each OD pair, the range is defined before hand, and they are listed in table 3.
Because it becomes too long to list the ranges of all the OD pairs in the network, in the table, only
some representative OD pairs generated from origin zone 1, 11 are listed.

9
Table 3. Zone pairs and ranges
As mentioned above, shares of Car and mass transit will be estimated by modal split with fuzzy
reasoning, in this section, the “Min – Max – Gravity “method is applied. The method is simply
represented as figure 6. For example, in this figure, the range of one OD pair is defined as 20
minutes, the time different is 6 minutes, the share of Car is 63% and share of mass transit is 27%.
Figure 6. The “Min – Max- Gravity” method
2.4. Route choice
In the past four decades, the route choice problem, along with the traffic assignment problem, has
been considered by a large number of authors worldwide. Teodorovic and Kikuchi (1990) [41]
were first to model the complex route choice problem using fuzzy logic. They used fuzzy
inference techniques to study the binary route choice problem. Akiyama et al. (1993) [2] also
developed a model for route choice behavior based on the fuzzy reasoning approach. Lotan and
Koutsopoulos (1993a, b) [22, 23] developed models for route choice behavior in the presence of
information based on concepts from approximate reasoning and fuzzy control. The research of
Lotan and Koutsopoulos (1993a, b) [22, 23] is particularly important within the context of
ongoing research in Intelligent Vehicle Highway Systems (IVHS). Teodorovic and Kalic (1995)
[40] developed an approximate reasoning algorithm to solve the route choice problem in air
transportation. Akiyama and Tsuboi (1996) [4] studied route choice behavior by multi-stage fuzzy
reasoning.

10
Let us consider the simplest case when the user is to choose one of two possible paths between
the origin and destination of movement. Perceived travel time is very often ‘fuzzy’ and can be
treated as a fuzzy set. In other words, when subjectively estimating travel time between two
points, expressions are used such as ‘it takes about 20 minutes from point C to point D’, ‘you'll
get there in about half an hour’, etc. it is assumed that users choose their paths based on a
comparison of the characteristics of alternative paths. Despite the fact that travel time is a
measurable parameter, a driver's notion of travel time when he/she makes the route choice is often
fuzzy. Thus, driver's preference for a route is based on a comparison of two fuzzy numbers. As a
result, his/her certainty of choosing a route is also not crisp. The route choice model can represent
the decision process under two layers of uncertainty: driver's judgment of time and driver's
preference (Teodorovic and Kikuchi, 1990) [41].
Let us denote by TA and TB, respectively, fuzzy sets representing estimated travel time (by one
user) along path A and path B. When comparing travel times along paths A and B, the user might
estimate, for example, that travel time along path A, TA is ‘much shorter than TB’, ‘shorter than
TB’, ‘equal to TB’, ‘longer than TB’, and ‘much longer than TB’. These fuzzy sets are denoted in
the following manner: MLTB-‘much less than TB’, LTB-‘less than TB’, GTB-‘greater than TB’,
MGTB-‘much greater than TB’. Let us assume that a user has a specific preference regarding the
choice of each of the possible paths through the network. This preference can be ‘stronger’ or
‘weaker’. Let us introduce into the discussion a preference index that can take values on the
interval from 0 to 1. When the user has an absolute preference for a specific path we consider the
preference index to be equal to 1.
This preference index decreases along with a decrease in the strength of the preference. Fuzzy
sets such as ‘very strong’, ‘strong’, ‘medium’ can express the strength of the user's preference,
‘weak’ and ‘very weak’. The following rules of approximate reasoning are proposed to establish
the transport network user's preference strength:
If TA=MLTB then PA=VERY STRONG
If TA=LTB then PA=STRONG
If TA=TB then PA=MEDIUM
If TA=GTB then PA=WEAK
If TA=MGTB then PA=VERY WEAK
where PA is the preference index associated with path A.
An approximate reasoning algorithm is used to establish the preference strength for every
network user. Once the preference index has been determined for each network user, an algorithm
must be developed that will determine the number of users along individual links. Such
algorithms are known in the literature as network ‘loading’ algorithms developed a network
‘loading’ algorithm is developed in which the basic input data are the preference indexes of
individual network users. It is assumed that the perceived travel times are distributed around a
previously measured most likely travel time. The ‘loading’ algorithm is based on the concept of
hybrid numbers (Teodorovic and Kikuchi 1990) [41].
Multi-stage fuzzy reasoning is used to describe the driver decision making process on road
networks. Their term paper considers the multi-route choice problem. The first step for multi-
route choice is made a survey and collected data. Experiment participants were asked about the
number of alternative routes they use going from their origin to their destination (the authors
asked the respondents to cite at most three alternatives they may use). For each route considered
they defined its characterizing factors (travel time, degree of congestion and risk of accidents).
All the participants estimated the values of all the factors using their own perception and

11
experience. Appropriate fuzzy numbers could then represent these values. The third survey
question dealt with the values of the utilities of the nominated routes. The authors assumed the
existence of two stages in the driver decision making process. The firrst approximate reasoning
algorithm they proposed determined the utilities of the alternative routes (Akiyama and Tsuboi,
1996) [4]. A typical rule in this approximate reasoning algorithm reads:
If T is large and CR is small and RA is medium
Then V is small
where:
(a) T=perceived travel time
(b) CR=congestion of a route
(c) RA=risk of accident
(d) V=utility for the route.
The first input variable DF in the second approximate reasoning algorithm is the difference
between the utilities associated with the shortest path and the second shortest path. The second
input variable DS is the difference in the utilities between the second shortest path and the third
shortest path. The third input variable N refers to the number of alternative routes. When N=2,
variable DS does not appear in the premises. A typical rule of the second approximate reasoning
algorithm reads:
If N=3 and DF is medium and DS is medium
Then FR(1) is large
FR(2) is medium
FR(3) is medium
where:
(a) DF=difference of utilities between the shortest path and the second shortest path
(b) DS=difference of utilities between the second shortest path and the third shortest path
(c) FR(i)=degree of frequency for route i
In Fuzzy Logic Model, membership functions of input and output parameters are determined
based on survey results of a particular area and by the help of an expert’s knowledge. For input
parameters (travel time, traffic safety, congestion and environmental effects), triangular and
trapezoidal membership functions are used, while for output parameter (route utility), triangular is
used (Figure 7). Most convenient membership functions are defined by examination of statistical
dispersion of the survey results and after some trials.

12
Figure 7. Fuzzy logic route choice model parameters and membership functions
Input parameters that are obtained from survey study were used in the first stage, given values of
these parameters are converted to fuzzy using membership functions in the second stage, the rule
base was used for inference in the third stage, using rule base a fuzzy output that shows the utility
of route was obtained and the fuzzy output was converted to crisp numbers using a
deffuzzification method in the fourth stage. At the end, route utility was obtained as a crisp
number. After these processes, choice probabilities of routes were determined by multinomial
logit formula.
Rule base used in the fuzzy logic model is formed after statistical analysis of the input and output
parameters. Rule base consists of 3*3*3*3=81 rules; 3 for traffic safety; 3 for congestion, 3 for
travel time and 3 for environmental effects. These IF-THEN rules are collated with AND
operator. Samples of fuzzy logic route choice model rule base are:
IF TT is Short and TS is High and Co is Low and EE is Good THEN RU is High
IF TT is Short and TS is Med.and Co is Low and EE is Good THEN RU is High
IF TT is Short and TS is Med. and Co is Med and EE is Med THEN RU is Medium
IF TT is Long and TS is Med. and Co is Low and EE is Med THEN RU is Medium
IF TT is Med. and TS is High and Co is Low and EE is Med.THEN RU is Med.
IF TT is Long and TS is Low and Co is High and EE is Med THEN RU is Low
The model output is gained where Mamdani inference mechanism is used. In order to compare
fuzzy results of Fuzzy Logic Model with crisp results of mathematical models, fuzzy output is
defuzzified using Centroid method as:

13
∫
∫ ⋅
=∗
dzzu
zdzzu
z
c
c
)(
)(
&&
&&
where, z= crisp value(value for x axis), cu =membership value for given crisp value.
2.5. Traffic assignment
Many papers devoted to the traffic assignment problem were published in world literature. There
are two characteristics common to most of these papers: they are based on Wardrop's principle
and they are travel time crisp. Akiyama et al. (1994) observed that ‘in real world, however, the
driver can only use fuzzy traffic information even if several types of information are available’.
These authors succeeded in formulating the fuzzy user equilibrium with fuzzy travel time. They
confirmed their assumption about drivers' perception of time as triangular fuzzy numbers by
organizing the survey on the network comprised of Hanshin Expressway and urban streets in the
Osaka area. In their paper, the authors introduced ‘the descriptive method of route choice
behavior to design the traffic assignment model’. They showed that ‘the state of user equilibrium
is also generated even if fuzziness of travel time exists’. Akiyama et al. (1994) [3] developed
Fuzzified Frank-Wolfe algorithm to solve the problem considered. This pioneer paper
undoubtedly presents a significant study on the relationship between traffic information and
drivers' behavior.
2.5.1 Fuzzy Sets of Perceived Link Travel Time
The main concern of drivers traversing in the road networks is travel time, travel time of each link
is the basic component for assigning traffic. The fuzzy sets of perceived link travel time will be
based on deterministic link travel time and actual link travel time
a. Deterministic Link Travel Time
Deterministic link travel time is determined by deterministic features of the link, like link length,
capacity, etc., and also the current deterministic traffic conditions, like traffic flow on this link.
The following equation (1) is usually chosen for computing the deterministic link travel time:
a
x
x
C
u
Ttxtuctc
a
a
a
a
faaaaa ∀++== ])(1[)](),([)(
max,
,
α
β (1)
where α and β are coefficients; Ta f, is the free flow travel time on link a; au is its inflow rate;
Ca is its capacity; xa,max is its maximum holding capacity.
b. Actual Link Travel Time
In practice, it is difficult to formulate the equation for actual link travel time due to the
tremendous factors that may impact the actual link travel time greatly, such as different time
period we focused on, road condition, traffic signal, congestion, incident, construction, weather,
special events, and so on. As mentioned above, this property is called the stochastic characteristic
of traffic condition (link travel time). However, in order to provide a reasonable approximation of
actual link travel time, some researchers using certain distribution to represent the stochastic part

14
of link travel time. Normal Distribution was applied in Liu and Ban’s paper [9]
as shown in
Equation (2) and (3):
)()()( ttct aaa ξτ += (2)
),(~)( 2
σµξ Nta (3)
where aτ is actual link travel time and aξ is its stochastic term.
In this paper, equation (1), (2), (3) is used to estimate actual link travel time.
c. Fuzzy Sets of Perceived Link Travel Time
1.0
.0.8
0.6
0.4
0.2
O
0.9 aτ aτ 1.1 aτ 2 aτ 2.15 aτ
NORMAL CONGESTED
Figure 8. Membership function for fuzzy sets of perceived link travel time
Fuzzy Sets for Perceived Link Travel Time
Based on the actual link travel time, we can construct fuzzy sets for perceived link travel time. In
this paper, we use linguistic descriptions to represent the fuzzy sets. Due to driver’s experience of
similar links, several linguistic descriptions of perceived link travel time can be formulated as
below:
o “travel time is normal” --- NORMAL
o “link is congested” --- CONGESTED
o “link has incident” --- INCIDENT
o “link has construction” --- CONSTRUCTION
o “link has special event” --- SPECIALEVENTS
These fuzzy sets represent different level of traffic conditions and more sets may be added if
necessary.
Membership Function for Fuzzy Sets of Perceived Link Travel Time
Generation of membership functions arouses a lot of interests in the literature. However, no
standard method exists yet. The membership functions can be constructed based on a survey. To
model driver’s perceptions, in most applications of fuzzy set theory, TFN (Triangular Fuzzy
Numbers) or TrFN (Trapezoidal fuzzy Numbers) are usually used. In this paper, triangular

15
function is used to form the membership functions of the fuzzy sets in 2-3-1. Figure 13 shows the
membership functions for NORMAL and CONGESTED fuzzy sets.
Fuzzy Rules for Perceived Link Travel Time
Only one simple rule will be used for these fuzzy sets of link travel time. “If the travel time of link
A is less than that of link B, link A will be chosen.” This rule implies that if link A is faster than
link B, then link A, instead of link B, will be used by the driver.
d. Fuzzy Sets of Perceived Path Travel Time
In crisp case, path travel time equals to the summation of the travel time of those links that
construct this path. However, for fuzzy case, we need to define how to formulate fuzzy sets of
perceived path travel time based on fuzzy sets of perceived link travel time. Therefore, here we
need to define fuzzy addition formulation.
In this paper, Dubois and Prade’s method is used. Four steps are needed to perform fuzzy addition
operation: Flattening, Decomposition, Operation * and Union. Here operation * is operation +.
Generally, Dubois’s method is a little complex to compute. But for fuzzy sets with triangular
membership function, it is quite simple.
As shown in Figure 9, if we represent fuzzy set S that has triangular membership as S (L, M, R),
its membership function can be expressed as:








∉
∈
−
−
∈
−
−
=
],[,0
],(,
],[,
)(
RLx
RMx
MR
xR
MLx
LM
Lx
xAµ (4)
And from Dubois’s theory[12]
, the addition of fuzzy set S1(L1,M1,R1) and S2(L2,M2,R2) will be:
),,( 2121212121 RRMMLLSSS +++=+=+ (5)
From equation (4) and (5), we can easily get fuzzy path travel time and its membership function
given the fuzzy travel time of those links that form this path.

16
M1L1
1.0
S2 (L2 ,M2 ,R2)
R1
L2 M2 R2
S1(L1 ,M1 ,R1)
O
O
L1+L2 M1+M2 R1+R2
S1+2=S1+S2
0.8
0.6
0.4
0.2
1.0
0.8
0.6
0.4
0.2
Figure 9. Fuzzy Addition Illustration
e. Fuzzy Shortest Path (FSP) Algorithm
Under crisp case circumstances, after we find the travel time for paths connecting origin and
destination, we can find the shortest path and assign all traffic to this path. However, under fuzzy
case, it is difficult to find only path that is the “shortest” in term of fuzzy travel time. Therefore,
we need some method to find a group of fuzzy “shortest” paths and assign traffic to all of them
based on their membership functions.
Several fuzzy shortest path algorithms have been proposed for solving the traffic assignment
problem. In this paper, Blue’s method is used.
According to Blue’s classification, the road networks we are working on are actually a pure Type
V fuzzy graph with crisp nodes and links as well as fuzzy weights (travel time). Denote G is a
Type V fuzzy graph, we can find the fuzzy shortest paths using the following three steps.
Step 1: Converting Fuzzy Graph to Crisp Graph
We first construct two crisp graphs G and G that are identical to G, except that the link travel
time is crisp. Link travel time for graph G is:
)}supp(sup{ aττ =a (6)
where sup means least upper bound, and supp means support of fuzzy set.
Equation (7) is the link travel time for graph G :
)}supp(inf{ aττ =a (7)
where inf means greatest lower bound.

17
Step 2: Find Shortest Path in Graph G
Find the shortest path in graph G . It is a traditional problem to find the shortest path in a crisp
graph and various algorithms are available to solve it [14]
. Denote K is the length of this shortest
path.
Step 3: Calculate Membership
Denote S is the set of paths inG , connecting origin to destination and with the length less than
K.
Denote S is a set of fuzzy paths in G. All paths in S are identical to those paths in S respectively,
except that the travel time is fuzzy. S is the set of fuzzy shortest paths found and will be assigned
traffic. Calculate the membership of K for each of the fuzzy path in S.
5. Fuzzy Traffic Assignment
From 4, we can get the set of fuzzy shortest paths S. We assume that S has M elements as
displayed below:
},...,....,{ 21 Mm ppppS = (8)
where pm is the fuzzy shortest path.
And the memberships of K in those fuzzy shortest paths are:
Mmfpf mm ,..2,1)( == (9)
Further, we assume the traffic flow from origin O to destination D is odT . In this paper, we assign
the traffic flow to these paths in S proportionally to their membership using equation (10):
∑=
= M
i
i
mm
od
f
f
T
1
(10)
For dynamic traffic assignment (DTA), we have to consider the flow propagation. In this case,
equation (11) is used for this purpose:
)]}()]([{)}()]([{)(
~
tEttEtxttxtx rs
pa
rs
p
pb
rs
bpa
rs
bp
rs
ap −++−+= ∑∈
ττ (11)
where p is the path a belongs to, p~ is the subpath from a to destination D.

18
3. CONCLUSION
Many problems encountered when studying complex transportation systems are highly non-
linear. As already noted, a fuzzy logic system is a nonlinear system that maps a crisp input vector
into a crisp scalar output. When solving a large number of different traffic and transportation
problems, this is what we actually do: map a crisp input vector into a crisp scalar output.
Fuzzy logic could be used successfully to model situations in which people make decisions in an
environment that is so complex that it is very hard to develop a mathematical model. Such
situations for example often occur in the field of traffic and transportation when studying the
work of dispatchers or modeling choice problems. Present experience shows that there is room for
the development of different approximate reasoning algorithms when solving complex problems
of this type. Until a few years ago the `trial and error' procedure was the customary method used
to develop fuzzy logic systems, and so researchers designed fuzzy logic systems independently of
numerical training data. Wang and Mendel (1992a,b) [46, 47]; Jang (1992) [13] and Horikawa et
al. (1992) [11] proposed separately from each other to tune fuzzy logic system parameters using
numerical training data.
This idea was accepted by a great many researcher and in the past two or three years a large
number of papers have appeared in which the fuzzy rule base and/or shape of the membership
function are determined using numerical training data. Designing fuzzy logic systems using
numerical training data is certainly one of the very important tasks in the area of fuzzy logic
systems for transportation engineering as well.
In the past several years, significant theoretical results have been achieved in the field of fuzzy
logic systems. Wang and Mendel (1992c) [48] and Kosko (1992) [21]showed that fuzzy logic
systems are universal approximators, which means that a fuzzy logic system `can uniformly
approximate any real continuous nonlinear function to an arbitrary degree of accuracy' (Mendel,
1995 [25]). This existence theorem shows us above all the possibilities offered by fuzzy logic, but
does not indicate the manner in which to create fuzzy logic systems. In other words, the number
of inputs, the number of fuzzy sets used to describe fuzzy variables, and the number of rules
essentially influence the quality of the solution generated by a fuzzy logic system. It should be
emphasized that feed-forward neural networks and fuzzy logic systems are techniques that can be
used to solve the same class of problems. Since fuzzy logic system parameters can be initialized
using expert knowledge, while weights in feed-forward neural networks are most often initialized
randomly, fuzzy logic systems are tuned much more quickly than the tuning of feed-forward
neural networks.
Fuzzy logic systems provide two other very important advantages. They can use existing
linguistic knowledge very successfully, and they treat uncertainty in an appropriate manner.
The basic goal of this paper was to classify and analyze results in the application of fuzzy logic
when modeling complex traffic and transportation processes. The results obtained show that
fuzzy set theory and fuzzy logic present a promising mathematical approach to model complex
traffic and transportation processes that are characterized by subjectivity, ambiguity, uncertainty
and imprecision. As already noted, the benefits from the fuzzy logic will be more accurately
assessed as the number of successful practical applications of the fuzzy logic in traffic control and
transportation planning increases.

19
REFERENCES
1. Akiyama, T., Shao, C.-F., (1993) “Fuzzy mathematical programming for traffic safety planning on an
urban expressway” Transportation Planning and Technology, Vol. 17, pp. 179-190.
2. Akiyama, T., Yamanishi, H., (1993) “Travel time information service device based on fuzzy sets
theory”, In: Ayyub, B.M. (Ed.), Proceedings of ISUMA '93, The Second International Symposium on
Uncertainty Modeling and Analysis. IEEE Computer Press, College Park, Maryland, pp. 238-245.
3. Akiyama, T., Shao, C-F., Sasaki, T., (1994) “Traffic flow on urban networks with fuzzy information”,
Memorial Faculty of Engineering, Kyoto University, Vol. 56, pp. 1-22.
4. Akiyama, T., Tsuboi, H., (1996) “Description of route choice behaviour by multi-stage fuzzy
reasoning”, Paper presented at the Highways to the Next Century Conference, Hong Kong.
5. Chakroborthy, P., (1990) “Application of fuzzy set theory to the analysis of capacity and level of
service of highways” MSc. thesis, University of Delaware, Newark, DE.
6. Chakroborthy, P., Kikuchi, S., (1990) “Application of fuzzy set theory to the analysis of capacity and
level of service of highways” In: Ayyub, B.M. (Ed.), Proceedings of ISUMA '90, The First
International Symposium on Uncertainty Modeling and Analysis. IEEE Computer Press, College
Park, Maryland, pp. 146-150.
7. Chanas, S., Delgado, M., Verdegay, J.L., Vila, M.A., (1993) “Interval and fuzzy extensions of
classical transportation problems”, Transportation Planning and Technology Vol. 17, pp. 203-218.
8. Chen, L., May, A., Auslander, D., (1990) “Freeway ramp control using fuzzy set theory for inexact
reasoning”, Transportation Research, Vol. 24A, pp. 15-25.
9. Chang, Y.H., Shyu, T.H., (1993) “Traffic signal installation by the expert system using fuzzy set
theory for inexact reasoning”, Transportation Planning and Technology, Vol. 17, pp. 191-202.
10. Deb, S.K., (1993) “Fuzzy set approach in mass transit mode choice”, In: Ayyub, B.M. (Ed.),
Proceedings of ISUMA '93, The Second International Symposium on Uncertainty Modeling and
Analysis. IEEE Computer Press, College Park, Maryland, pp. 262-268.
11. Horikawa, S., Furahashi, T., Uchikawa, Y., (1992) “On fuzzy modeling using fuzzy neural networks
with back-propagation algorithm”, IEEE Transactions on Neural Networks, Vol. 3, pp. 801-806.
12. Hsiao, C.-H., Lin, C.-T., Cassidy, M., (1994) “Application of fuzzy logic and neural networks to
automatically detect freeway traffic incidents”, Journal of Transportation Engineering, Vol. 120, pp.
753-772.
13. Jang, J.-S.R., (1992) “Self-learning fuzzy controllers based on temporal back-propagation”, IEEE
Transactions on Neural Networks, Vol. 3, pp. 714-723.
14. Jassbi J., Makvandi P., Ataei M. and Sousa Pedro A. C., (2011) “Soft system modeling in
transportation planning:Modeling trip flows based on the fuzzy inference system approach”, African
Journal of Business Management, Vol. 5(2), pp. 505-514.
15. KalicÂ ,M., TeodorovicÂ , D., (1996) “ Solving the trip distribution problem by fuzzy rules generated
by learning from examples”, Proceedings of the XXIIIYugoslav SymposiumonOperationsResearch,
Zlatibor,Yugoslavia, pp. 777-780 (in Serbian).
16. KalicÂ , M., TeodorovicÂ , D., (1997a) “Trip distribution modeling using soft computing
techniques”, Paper presented at the EURO XV/INFORMS XXXIV (Book of abstracts, p. 74),
Barcelona.
17. KalicÂ , M., TeodorovicÂ , D., (1997b) “A soft computing approach to trip generation modeling”,
Paper presented at the 9th
Mini EURO Conference Fuzzy sets in traffic and transport systems, Budva,
Yugoslavia.
18. Kikuchi, S., (1992) “Scheduling demand-responsive transportation vehicles using fuzzy-set theory”,
Journal of Transportation Engineering, Vol. 118, pp. 391-409.
19. Kikuchi, S., VukadinovicÂ , N., Easa, S., (1991) “Characteristics of the fuzzy LP transportation
problem for civil engineering applications”, Civil Engineering Systems, Vol. 8, pp. 134-144.
20. Kikuchi, S., Perincherry, V., Chakroborthy, P., Takahashi, H., (1993) “Modeling of driver anxiety
during signal change intervals”, Transportation Research Record, Vol. 1399, pp. 27-35.
21. Kosko, B., (1992) “Fuzzy Systems as universal approximators”, Proceedings of 1st IEEE
International Conference on Fuzzy Systems, pp. 1153-1162.
22. Lotan, T., Koutsopoulos, H., (1993a) “Route choice in the presence of information using concepts
from fuzzy control and approximate reasoning”, Transportation Planning and Technology, Vol. 17,
pp. 113-126.

20
23. Lotan, T., Koutsopoulos, H., (1993b) “Models for route choice behaviour in the presence of
information using concepts from fuzzy set theory and approximate reasoning” Transportation, Vol.
20, pp. 129-155.
24. Mamdani, E., Assilian, S., (1975) “An experiment in linguistic synthesis with a fuzzy logic
controller” International Journal of Man-Machine Studies, Vol. 7, pp.1-13.
25. Mendel, J., (1995) “Fuzzy logic systems for engineering”, A tutorial. Proceedings of the IEEE 83, pp.
345-377.
26. MilosavljevicÂ , N., TeodorovicÂ , D., PapicÂ , V., PavkovicÂ , G., (1996) “A fuzzy approach to the
vehicle assignment problem”, Transportation Planning and Technology Vol. 20, pp. 33-47.
27. Nakatsuyama, M., Nagahashi, N., Nishizuka, N., (1983) “Fuzzy logic phase controller for traffic
functions in the one-way arterial road”, Proceedings IFAC 9th Triennial World Congress. Pergamon
Press, Oxford, pp. 2865-2870.
28. Pappis, C., Mamdani, E., (1977) “A fuzzy controller for a traffic junction”, IEEE Transactions on
Systems, Man and Cybernetics SMC-7, pp. 707-717.
29. Perincherry, V., (1990) “Application of fuzzy set theory to linear programming” MSc. thesis,
University of Delaware, Newark, DE.
30. Perincherry, V., Kikuchi, S., (1990) “A fuzzy approach to the transshipment problem”, In: Ayyub,
B.M. (Ed.), Proceedings of ISUMA '90, The First International Symposium on Uncertainty Modeling
and Analysis. IEEE Computer Press, College Park, Maryland, pp. 330-335.
31. Perkinson, D., (1994) “Using automated vehicle location data to monitor congestion,: Fuzzy set
theory, ITE Journal, pp. 35-40.
32. Quadrado, J.C., Quadrado, A.F., (1996) “Fuzzy modeling of accessibility: Case study-Lisbon
metropolitan area”, Proceedings of the Fourth European Congress on Intelligent Techniques and Soft
Computing,Aachen, Germany, pp. 1307-1311.
33. Sasaki, T., Akiyama, T., (1986) “Development of fuzzy traffic control system on urban expressway”,
Preprints 5th IFAC/IFIP/IFORS International Conference in Transportation Systems, pp. 333-338.
34. Sasaki, T., Akiyama, T., (1987) “Fuzzy on-ramp control model on urban expressway and its
extension”, In: Gartner, N.H., Wilson, N.H.M. (Eds.), Transportation and traffic theory. Elsevier
Science, New York, pp. 377-395.
35. Sasaki, T., Akiyama, T., (1988) “Traffic control process of expressway by fuzzy logic,. Fuzzy Sets
and Systems, Vol. 26, pp.165-178.
36. Sugeno, M., Nishida, M., (1985) “Fuzzy control of model car”, Fuzzy Sets and Systems, Vol. 16, pp.
103-113.
37. TeodorovicÂ, D., (1994) “Invited review: fuzzy sets theory applications in traffic and transportation”,
European Journal of Operational Research, Vol. 74, pp. 379-390.
38. TeodorovicÂ , D., BabicÂ , O., (1993) “Fuzzy inference approach to the flow management problem
in air traffic control”, Transportation Planning and Technology, Vol.17, pp. 165-178.
39. TeodorovicÂ , D., KalicÂ , M., PavkovicÂ , G., (1994) “The potential for using fuzzy set theory in
airline network design”, Transportation Research 28B, pp.103-121.
40. TeodorovicÂ , D., KalicÂ , M., (1995) “A fuzzy route choice model for air transportation networks”,
Transportation Planning and Technology Vol. 19, pp.109-119.
41. TeodorovicÂ , D., Kikuchi, S., (1990) “Transportation route choice model using fuzzy inference
technique”, In: Ayyub, B.M. (Ed.), Proceedings of ISUMA '90, The First International Symposium on
Uncertainty Modeling and Analysis. IEEE Computer Press, College Park, Maryland, pp. 140-145.
42. TeodorovicÂ , D., Kikuchi, S., (1991) “Application of fuzzy sets theory to the saving based vehicle
routing algorithm”, Civil Engineering Systems, Vol. 8, pp. 87-93.
43. TeodorovicÂ , D., PavkovicÂ , G., (1996) The fuzzy set theory approach to the vehicle routing
problem when demand at nodes is uncertain. Fuzzy Sets and Systems Vol. 82, pp. 307-317.
44. Tzeng, G.-H., TeodorovicÂ , D., Hwang, M-J., (1996) “Fuzzy bicriteria multi-index transportation
problems for coal allocation planning of Taipower”, European Journal of Operational Research, Vol.
95, pp. 62-72.
45. VukadinovicÂ , K., TeodorovicÂ , D., (1994) “A fuzzy approach to the vessel dispatching problem”,
European Journal of Operational Research, Vol. 76, pp.155-164.
46. Wang, L-X., Mendel, J., (1992a) “Generating fuzzy rules by learning from examples”, IEEE
Transactions on systems, Man and Cybernetics, Vol. 22, pp. 1414-1427.

21
47. Wang, L-X., Mendel, J., (1992b) “Back-propagation of fuzzy systems as nonlinear dynamic system
identifiers”, Proceedings IEEE International Conference on Fuzzy Systems, San Diego, CA, pp. 807-
813.
48. Wang, L.-X., Mendel, J., (1992c) “Fuzzy basis functions, universal approximation, and orthogonal
least squares learning”, IEEE Transactions on Neural Networks, Vol.3, pp. 807-813.
49. Xu, W., Chan, Y., (1993a) “Estimating an origin-destination matrix with fuzzy weights”, Part 1:
Methodology. Transportation Planning and Technology, Vol. 17, pp. 127-144.
50. Xu, W., Chan, Y., (1993b) “Estimating an origin-destination matrix with fuzzy weights”, Part 2: Case
studies. Transportation Planning and Technology, Vol. 17, pp. 145-164.
51. Zadeh, L.,(1973) “Outline of a new approach to the analysis of complex systems and decision
processes”, IEEE Transactions on Systems, Man and Cybernetics SMC-3, pp. 28-44.
Authors:
Amrita Sarkar
Amrita Sarkar is a graduate Engineer in Information Technology with a post graduation
in Remote Sensing. She is presently a PhD Research Fellow at the Department of
Information Technology, Mesra, India. She has got few research publications in her area
of specialization. Her areas of interests include Soft Computing, Artificial Intelligence,
Data Mining, DBMS and Image Processing.
Dr. G. Sahoo
Dr. G. Sahoo received his MSc in Mathematics from Utkal University in the year 1980
and PhD in the area of Computational Mathematics from Indian Institute of Technology,
Kharagpur in the year 1987. He has been associated with Birla Institute of Technology,
Mesra, Ranchi, India since 1988, and currently, he is working as a Professor and Head in
the Department of Information Technology. His research interest includes theoretical
computer science, parallel and distributed computing, evolutionary computing,
information security, image processing and pattern recognition.
Dr. U. C. Sahoo
Dr. U. C. Sahoo is working as an Assistant Professor in the Department of Civil
Engineering, Indian Institute of Technology, Bhubaneswar and is an expert in the field of
Transportation Engineering. He has more than eight years of teaching and research
experience. Presently he is engaged in research in the area of transportation planning, road
safety and pavement engineering and published many papers in these areas.

APPLICATION OF FUZZY LOGIC IN TRANSPORT PLANNING

More Related Content

What's hot (15)

Similar to APPLICATION OF FUZZY LOGIC IN TRANSPORT PLANNING (20)

Recently uploaded (20)

APPLICATION OF FUZZY LOGIC IN TRANSPORT PLANNING