MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

An Ant Colony Optimization Solution to the Optimum Travel Path Determination

Problem in VANETS: A NetLogo Modelling Approach
1

Kponyo Jerry ,2 Yujun Kuang,

Opare Kwasi, 3 Nwizege Kenneth,

Abdul-Rahman Ahmed,

Koudjo Koumadi,

Kwame Nkrumah Univ. of Sci. and Tech, KNUST, Kumasi, Ghana

University of Electronic Sci and Tech of China, UESTC, Chengdu, China
3
Swansea University, Swansea, UK
4
University of Ghana, Legon, Ghana

Emails: jjkponyo.soe@knust.edu.gh, kyj@uestc.edu.cn, opare@knust.edu.gh,

s.k.nwizege@ieee.org, aarahman.soe@knust.edu.gh, koumadi@gmail.com,

AbstractA Dynamic Travel Path Optimization System

(DTPOS) based on Ant Colony Optimization (ACO) for the prediction of the best path to a given destination is presented. The
system is modeled in a multi-agent multi-purpose framework
in NetLogo and experiments conducted on 100 test vehicles
under different traffic scenarios. The test vehicles are released
from a fixed location in the simulation environment and given
a predefined destination. The route taken by the vehicles to
reach the destination gives a measure of the intelligence in
the system. Two variations of the system, DTPOS with ACO
(DTPOS+ACO) and DTPOS without ACO (DTPOS-ACO) are
investigated to establish the effect of ACO on the solution. For
every vehicle which successfully makes it to the destination,the
mean travel time is recorded. The results have shown that
at all traffic densities the mean travel time of vehicles in
DTPOS+ACO was always lower than those for DTPOS-ACO.
Also it was shown that for the same percentage of vehicles
arriving at the destination, DTPOS+ACO vehicles took less
time than DTPOS-ACO vehicles.
Keywords-Intelligent Transport Systems (ITS); Vehicular Ad
hoc Networks (VANETS); Swarm Intelligence (SI); Ant Colony
Optimization (ACO); Dynamic Travel Path Optimization System(DTPOS)

I. I NTRODUCTION
Traffic congestion has been a serious problem on roads
around the world ever since ancient Roman times, when the
streets of Rome became so congested that all non-official
vehicles were prevented from entering the city [1]. In recent
times, traffic congestion has been responsible for problems
like long delays, wasted time, and increased pressure [2],
not only to drivers, but also to passengers and even to
pedestrians.
In the United States for instance, research has shown that
4.2 billion hours are wasted daily just waiting in traffic; this
converts to 2.8 billion gallons of fuel [3]. Similar results
have been recorded in Europe, Japan, and Australia [4][5][6].
Research has demonstrated that lower speed vehicles emit
more CO2 , for instance, vehicles traveling at 60Kph emit
40 percent fewer carbon emissions than vehicles traveling

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

at 20Kph and vehicles traveling at 40Kph emit 20 percent

fewer emissions than the 20Kph baseline [6].
Intelligent Transport Systems (ITS) provide attractive
methods for reducing congestion, which involves the use of
modern electronic information systems to control, manage,
and regulate traffic flows according to inputs from traffic
flow status prediction systems, through dynamic signal timings [7][8].
In this paper, we propose a Dynamic Travel Path Optimization System (DTPOS) which is an ITS solution to
the traffic problem based on Ant Colony Optimization
techniques. Ant colony optimization is a classic example of Swarm Intelligence (SI), in which case ants using
pheromone relay information from one ant to the other to
enable them determine the shortest and optimum path from
a new food source to the nest. Initially, the ants travel on
all possible paths while depositing pheromone on their trail.
After some time when more ants use the shorter paths, more
pheromone is deposited to act as positive feedback which
quickly results in the shortest trail being selected as a better
option due to its high pheromone concentration. The ACO
algorithm mimics the behavior of ants foraging for food
[9][10].
The remainder of the paper is organized as follows:
Section II briefly looks at some related works; Section III
mathematically models the ant colony optimization solution
to the traffic problem; Section IV introduces the proposed
ACO inspired DTPOS model and Section V touches on the
simulation of the model by NetLogo and Section VI discusses the results. Finally Section VII presents conclusions
on our key research findings.
II. R ELATED W ORKS
The first instance where an ant based system was used for
dynamic problem solving was in [11]. Ant Colony Optimization (ACO) has been applied to dynamic path optimization
in [12]. The authors in [12] have also demonstrated how the

24

MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

ACO algorithm can be structured so as to adapt to changes in

the initial constraints of the optimization problem. In [13]
an optimum traffic system for the reduction of vehicular
traffic congestion in an urban environment is proposed. The
algorithm proposed in this system has the limitation of
performing well only when the number of agents is above
100. A dynamic system for the avoidance of traffic jams
(DSATJ) is also proposed in [14]. This system gives an
alternative path whenever there is a traffic jam at any section
of the road and resumes to the original route when the traffic
situation gets better. An ant colony system for a dynamic
vehicle routing problem has been proposed in [15], this
system provides a means to route a fleet of vehicles with
the objective of visiting a set of customers in minimum
time. A hybrid ACO technique for dynamic vehicle routing
is introduced in [16]. In [17], an improved ant colony
optimization algorithm by Previous Path Replacement (PPR)
which the authors term path crossover for optimal path
planning is introduced.
In [18], a preference based shortest path determination
using ACO is investigated. In this case, shortest path is
obtained taking into consideration the preferred paths of
the agents. In [19] ACO has been applied to a multicriteria vehicle navigation problem. In this case an exact
shortest path solution is not the ultimate objective but a
compromised set of best solutions taking into consideration
different preferences by different drivers.
ACO has been extensively applied to shortest path problems in [20]-[22]. In [23], the time for obtaining shortest
path solution is shortened by using a modified version of the
AntNet routing algorithm. In [24], a traffic congestion forecasting algorithm based on a pheromonal communication
model is proposed. This algorithm allows vehicles to react to
dynamically changing traffic situations through information
disseminated in the pheromone model. A system for travel
time prediction which takes into consideration past, present,
and future traffic trends is presented in [25].
In our previous work in [26], we introduced a Distributed
Intelligent Traffic System (DITS) which proposes a solution
to the traffic problem using ACO. In that paper, we investigated how ACO reduces the global traffic situation through
cooperation among the vehicles. In our recent work in [27],
similar to the work proposed in this paper an optimum path
prediction is proposed using ACO. The results are however
limited to a single traffic density and traffic distribution. In
this paper, we vary the traffic density as well as the traffic
distribution and investigate the effect of ACO on the Mean
Travel Time (MTT) as well as the percentage of vehicles
arriving at the destination, and the time taken. To implement
information relay among the vehicles in DTPOS we refer to
our previous work in [28].

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

III. M ATHEMATICAL M ODELING OF THE ACO

S OLUTION TO THE P ROBLEM
In defining the path determination problem, we consider
a minimization problem (S, f, ), where S is the set of
candidate solutions, f is the objective function which assigns
an objective function value f(s,t) to each candidate solution
s S and (t) is the set of constraints. The parameter t
indicates that the objective function, and the constraints can
be varying with time. The goal is to find a globally optimal
feasible solution s* which is a minimum cost feasible
solution to the problem.
In determining the candidate solution set S, we begin by
first defining a solution set R such that R = {r1 , r2 , r3 ...rn }.
The constraint is such that (s,d) is the best path given
a source to destination pair (s,d). In building a complete
solution S we begin from an empty set and then build a
complete solution S stepwise by adding one new component
ri R at every step.
The step by step decision depends on a stochastic decision
policy () which depends on a set of pheromone variables
ij . A policy in this context is a rule which links the resultant
status to an action based on given constraints. The stochastic
policy ij is characterised by a distribution probability over
all the likely actions. Each feasible action is connected to a
selection probability. An action is therefore taken based on
the selection probability for that action. ij therefore gives
an indication of how good the decision to use a particular
path is. For example ij can represent the desirability of
having (ri , rj ) in the anticipated solution sequence.
The ultimate objective is to get a good complete solution
which satisfies the primary goal of having the best path from
a source s to a destination d. ij therefore represents the
desirability of moving from si to sj in order to reach a
destination d such that the final path from a source s to a
destination d is the optimum, given a list of candidate paths.
The ij values are used to calculate the selection probability
Pij of each solution.
If N (ri ) is the set of all likely components in state ri ,
then the probability of each ri N (ri ) is calculated as Pij .
The road network is represented as a directed graph G such
that:
G = (N, E)

(1)

where N =(N1 ,N2 ,N3 ,N4 ...Nn ) is the set of n nodes (i.e.
junctions) and E is the set of directed edges as shown in
Figure 1. The objective of the model is to route vehicles
so that they reach their destination in the quickest time
possible while avoiding heavy traffic portions of the road.
The modeling of the problem is subject to the following
limitations:
Let represent the delay in vehicular movement and
the congestion situation; it can clearly be seen that: ,
however, even though other factors like road accidents and

25

MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

the deposition rate ij (the rate at which cars arrive). The

pheromone or traffic update is governed by the equation:
Tij = (1 )Tij + ij

ij depends on whether a car used the edge aij or not,

ie whether aij or aji =1. The total amount of pheromone
added or traffic added can be calculated as follows:

Figure 1. A typical road scenario for best path determination.

road works can also cause delays, traffic delays in this case
are only limited to congestion. In this case all delays as a
result of other factors are quantified in terms of congestion.
Eij is characterized by the length xij and traffic Tij . Each
route in the network is represented by (2) below:
R = aij

(8)

(2)

ij =

(k)
N
X
aij
tm (k)

(9)

k=1

tm (k) is the time taken by the car k in covering that section

of the road and is a function of the speed of the car v(r).
N is the total number of cars in that section of the road
tm (k) =

xij
v(k)

(10)

where

aij =

1,
0,

if node j is visited after node i

otherwise

Substituting equation (10) into (9) and (8) yields:

(3)

where i,j=1...n and n is the total number of nodes in the

route R.
We assume that time taken to traverse Eij is independent
of time taken to traverse other edges. Therefore the total
transit time Tm (R) for a route R is given by:
Tm (R) =

n
n X
X

aij xij Tij

(4)

Tij = (1 )Tij +

(k)
M
X
aij
xij
k=1 v(k)

(11)

Selecting and such that

+ =1

(12)

Equation (7) is reduced to:

i=1 j=1

The objective is to choose a path to the destination which

minimizes Tm (R). Every path is assigned a score and the
path with the best score is attained under the condition:
Best Score(S) = Score at min(xij Tij )

(5)

The shorter the route the more traffic it attracts. The

relationship between distance and traffic can therefore be
represented as:
1
Tij (R)
(6)
xij
The ultimate objective is to get a solution which is a
trade-off between minimum distance and traffic. We play
around both parameters in arriving at the best solution to
the problem. In moving from node i to node j the cars make
a decision based on the probability below:
Pij = P

Tij xij

hQ

x
Tih
ih

(7)

where and give the influence of traffic (pheromone)

and distance on the solution and Q is the set of nodes
not yet visited. The traffic (pheromone) update depends on
the evaporation rate (the rate at which cars leave) and

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

(1)

Pij = P

Tij xij

(1)

hQ

x
Tih
ih

(13)

At every junction, the car computes the probability Pij in

(13) and selects the junction with the highest probability.
It continues until it gets to the destination, after which the
travel time is computed. In testing the model, it is compared
with a model that computes its probability by considering
only the distance of the candidate paths with no knowledge
of the traffic. In that case:
xij
Pij = PK PK
i=1

j=1

xij

(14)

where K is the number of candidate junctions. Before traffic

begins to build up all other cars arriving at the junction
choose their path based on (14). In that case, the path
with the shortest distance to the next junction is selected,
however; if the distances are the same, they all have an
equal probability of being selected. Portions of the above
mathematical model were introduced in [9] but the main
difference between this approach and previous methods is
the use of ABM techniques to investigate the effect of other
agents on the global optimum path determination problem.

26

MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

IV. DTPOS D ESCRIPTION

The DTPOS proposed in this section is developed within
the framework of a Multi Agent Multi Purpose (MAMP)
system architecture to predict the optimum path to a predefined destination. The ultimate objective of DTPOS is
to reduce the travel time by intelligently selecting the best
path in terms of distance and traffic. The proposed system
accomplishes this task by computing a selection probability
Pij for the best path at every traffic intersection. In this paper
two variations of the DTPOS, DTPOS+ACO and DTPOSACO which differ only in the inclusion of ACO in one and
its absence in the other are developed and simulated. The
results are then discussed in later sections to get a better
understanding of the effect of ACO on VANETs.

There are two breeds of vehicles in the system; these

are test cars and passers. The test cars are assigned specific
destinations and the time taken for them to maneuver their
way through the traffic to reach their destination is recorded
as the travel time. The second breed of vehicles, the passers,
also travel within the study environment to create a traffic
situation but they are not monitored as is done for the test
vehicles. The system works in such a manner that when test
cars get to the junctions and a decision has to be made
as regards which path to take, the vehicles compute the
probability Pij as indicated in equation (13). The path with
the best probability at that instance is selected by the vehicle.
As soon as a test car arrives at its destination the travel
time is recorded, and the vehicle dies out of the simulation
environment. The stopping criterion for the simulation is for
all the test cars to arrive.
2) DTPOS-ACO Description: The DTPOS-ACO is similar to Figure 2 shown above. The major difference between
the two systems is in the way the selection probability
is calculated. Pij in this case is calculated only from the
distance without taking the traffic into consideration as
illustrated by (14).
V. S IMULATION F LOW D IAGRAM

Figure 2. NetLogo GUI with 6x6 road network topology.

The use of agent-based models(ABMs) or individual

based simulation models is growing rapidly in a number of
fields. The Logo family of platforms is currently one of the
most widely used ABMs. NetLogo is popular in this family
because it offers the option of studying the behaviour of
individual agents working in isolation as well as the effect
of the agents working in a community.
1) DTPOS+ACO Description: The ACO inspired system
relies on repeated sampling of multiple solutions to the
optimum path determination problem. The solutions from
these outcomes are used to update the value of pheromone
variables in the model. The major difference between this
system and ant behavior is that while ants choose the path
with the highest pheromone concentration as the best path,
the proposed system selects the path with the least traffic as
the best path. As the traffic concentration increases, the probability of selecting a particular path reduces. Figure 2 shows
the GUI of the NetLogo implementation of DTPOS+ACO.
The DTPOS+ACO mimics traffic behavior in an urban
environment. The system has a control environment which
permits the user to set the initial conditions for the experiment and an observation environment which gives a visual
display of what is happening in the simulation environment.
The system permits the number of cars, the maximum speed,
the destination and the traffic controls to be set by the user.
It relies on mobile agents which are vehicles in this case
and stationary patches which represent the road network.

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

In this section, the simulation flow diagrams for the two

DTPOS systems are discussed to give a better understanding
of how they are simulated. As has already been mentioned
the systems are implemented in NetLogo. To have a fair
comparison, both systems are simulated under the same conditions, and the results obtained are analyzed. The systems
are simulated based on the simulation flow diagram shown
in Figure 3. The simulation is initialized by choosing the
size of the grid, the number of cars, the maximum speed of
cars, the pheromone evaporation rate, the traffic duty cycle,
and the recording time interval. The pheromone values are
initialized to give a better picture of the traffic buildup as
the simulation progresses.
The simulation terminates when all the test cars have
arrived at a given destination. For every traffic density
selected, five separate traffic distributions are considered and
the simulation run until the stopping criteria is met. The
simulation parameters are as shown in TABLE I. 100 test
cars are made to travel a fixed distance under different traffic
densities and the mean travel times as well as the percentage
of cars arriving at the predefined destination are recorded and
compared.
VI. S IMULATION R ESULTS AND A NALYSIS
Let the time taken by a vehicle i to arrive at the predefined
destination be ti . Given that N vehicles successfully arrived
at the destination, the mean travel time given N vehicles is
given by:
PN
ti
(15)
tt = i=1
N

27

MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

Case 3: 400 Vehicles in 6x6 topology.

The experiment is conducted for DTPOS+ACO and
DTPOS-ACO to obtain tt and TK for three different traffic
densities. tt and TK are measured in ticks which translates
to seconds in the simulation. 1tick is equivalent to 1second
in this simulation.

A. Case1: 200 Vehicles

Figure 4 shows the mean travel time comparison for
DTPOS+ACO and DTPOS-ACO for a traffic density of 200
vehicles. It is evident that vehicles in the DTPOS+ACO case
have shorter mean travel times than those in the DTPOSACO case. What this means is that vehicles in the DTPOS+ACO case spend less time reaching their destination
compared to those in the DTPOS-ACO case travelling to
a common destination. Figure 5 also shows the percentage
of vehicles arriving at the specified destination and the
cumulative time taken. As can be seen from the figure it
takes about 50ticks for 50 percent of the vehicles to arrive
in the DTPOS+ACO case and vehicles take 100ticks which
is twice that time to get 50 percent of the vehicles arriving
in the DTPOS-ACO case. The figures double in both cases
to have 100 percent of the vehicles arriving.

Figure 3. Simulation Flow Diagram for DTPOS.

TABLE I
SIMULATION PARAMETERS.

PARAMETER
Model Simulator
Total number of vehicles
Total number of test vehicles
Road Topology
Maximum Speed
Trace volatility rate()

Terminating Criteria

SPECIFICATION
NetLogoV4.1RC5
200, 300 and 400
100
6x6
1.0(50Km/h)
8 percent
0.5
0.5
All test cars to arrive

Also let the number of vehicles arriving at the destination

be K and let TK be the time taken for all the vehicles to
arrive at the predefined destination. The time taken for all
vehicles to arrive is calculated as a sum of the individual
times. Hence the total time is given by:
TK =

K
X

ti

(16)

i=1

In the experiment, 100 test vehicles are set off and the mean
travel time for every N vehicles arriving is computed using equation (15). The time taken for a percentage of the vehicles
to arrive at the predefined destination is also recorded. Three
different cases are defined for the experiment as follows:
Case 1: 200 Vehicles in 6x6 topology;
Case 2: 300 Vehicles in 6x6 topology;

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

Figure 4. Mean Travel Time of Vehicles: Case 1.

B. Case 2: 300 Vehicles

Figure 6 shows a similar trend to what has been previously
discussed with DTPOS+ACO showing lower mean travel
times than DTPOS-ACO. The mean travel time figures
however are much higher for the 300 vehicles traffic density
than for the 200 vehicles.
Figure 7 also shows the percentage of vehicles arriving
at the predefined destination and the time taken for all of
them to reach their destination. For the case of 300 vehicles,
it takes 100ticks for 50percent of the vehicles to arrive at
the destination in the DTPOS+ACO situation and 200ticks
for the same percentage to arrive at the destination in the

28

MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

Figure 5. Percentage of Vehicles Arriving Vs Time: Case 1.

Figure 8. Mean Travel Time of Vehicles: Case 3.

Figure 6. Mean Travel Time of Vehicles: Case 2.

Figure 9. Percentage of Vehicles Arriving Vs Time: Case 3.

C. Case 3: 400 Vehicles

Figure 7. Percentage of Vehicles Arriving Vs Time: Case 2.

Figure 8 shows the mean travel time comparison of

vehicles for the two systems with a traffic density of 400
vehicles. The trend remains the same with higher values of
travel time as a result of the increased traffic density. Figure
9 however, shows an interesting trend when the percentage
of vehicles arriving and their cumulative times are studied.
It was realised that as the traffic density increases it becomes
almost impossible for all the vehicles to arrive in the
DTPOS-ACO case. For the 400 vehicles traffic density, it
was realised that while it took about 220ticks for all the
vehicles to arrive in the DTPOS+ACO case, it took almost
600ticks which is almost thrice the time for all the vehicles
to arrive in the DTPOS-ACO case.
VII. C ONCLUSION

DTPOS-ACO situation. It takes twice as much time in both

systems for all the vehicles to arrive. Once again it has been
shown that vehicles in the DTPOS+ACO case use less time
than those in DTPOS-ACO.

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

In this paper, a dynamic travel path optimization system

has been proposed. The system is implemented in NetLogo. Two cases of the DTPOS have been developed, one
with ACO and the other without ACO. The two systems

29

MOBILITY 2015 : The Fifth International Conference on Mobile Services, Resources, and Users

were simulated, and the results analyzed for comparison

of performance. The performance indicators studied have
shown that the DTPOS with ACO always gives results
which are better than the case without ACO. At all traffic
densities, the mean travel time of vehicles was always lower
for the DTPOS+ACO than the DTPOS-ACO. Also, the
percentage of vehicles arriving at the destination and their
cumulative time were compared, and it was evident that
DTPOS+ACO took less time for all the vehicles to arrive
than DTPOS-ACO. It can therefore be inferred that Ant
Colony Optimization solves the vehicular traffic problem by
substantially reducing the travel time.
R EFERENCES
[1] R. Salter and N. Hounsell, Highway Traffic Analysis and Design,
Third ed. Palgrave: Macmillan, 1996.
[2] United Kingdom, Department for Transport, Perceptions of congestion: report on qualitative research findings, London, UK, 2001.
[3] RITA, Intelligent Transportation Systems (ITS) Strategic Plan:
Background and Processes, January 8, 2010, pp. 18-19.
[4] E. Thielman, Head of Division for ITS, European Union, keynote
speech at the 15th ITS World Congress, New York City, November
17, 2008.
[5] S. Ezell,Report by The Information Technology and Innovation Foundation, Explaining International Leadership: Intelligent
Transportation Systems; January, 2010.
[6] Japan Highway Industry Development Organization, ITS Handbook Japan 2007-2008 pp. 89-90.
[7] L. Rao, L. Owen, and D. Goldsman, Development and application
of a validation framework for traffic simulation models presented
at the 30th conference on Winter simulation, WSC 98, Washington, DC, USA, 1998, pp.1079-1086.
[8] M. Slinn, P. Mathews, and P. Guest, Traffic Engineering:Design,
Principles and Practice, Second ed Butterworth-Heinemann, 1998.
[9] M. Dorigo and G. Di Caro, Ant Colony Optimization: A new
meta-heuristic Congress on Evolutionary Computation, 1999, pp.
1470-1477.
[10] K. Jones, Comparison of genetic algorithm with particle swarm
optimization Journal of International Conference on Computer
Systems and Technologies, vol. IIIA, pp. 1-6 2005.
[11] E. Bonabeau, Dorigo, M, and Theraulaz, G., Ed.,Swarm Intelligence:From Natural to Artificial Systems, Santa Fe Institute
Studies in Science of Complexity, Oxford University Press, 1999.
[12] D. Angus and T. Hendtlass, Dynamic ant colony optimization
Applied Intelligence vol. 23 pp.33-38 2005.
[13] S. Nahar and F. Hashim, Modelling and analysis of an efficient
traffic network using ant colony optimization algorithm Third
International Conference on Computational Intelligence, Communication Systems and Networks, 2011, pp. 32-36.
[14] P. Bedi, N. Mediratta, S. Dhand, and A. Singhal, Avoiding traffic
jam using ant colony optimization-A novel approach International
Conference on Computational Intelligence and Multimedia Applications, 2007, pp. 61-67.
[15] R. Montemanni, L. M. Gambardella, A. E. Rizzoli, and A. V.
Donati, Ant Colony System for a Dynamic Vehicle Routing
Problem Journal of Combinatorial Optimization, vol. 10, 2005,
pp. 327-343.
[16] Darren M. Chitty and Marcel L. Hernandez, A Hybrid Ant Colony
Optimization Technique for Dynamic Vehicle Routing, Lecture
Notes in Computer Science, vol. 3102, 2004, pp. 48-59.
[17] L. Joon-Woo, K. Jeong-Jung, and L. Ju-Jang, Improved Ant
Colony Optimization algorithm by path crossover for optimal
path planning Industrial Electronics, 2009. ISIE 2009. IEEE
International Symposium, 2009, pp.1996-2000.

Copyright (c) IARIA, 2015.

ISBN: 978-1-61208-417-6

[18] S. Ok, W. Seo, and J. Ahn, An Ant Colony Optimization Approach for the Preference-based Shortest Path Search Journal of
the Chinese Institute of Engineers, Vol 34, Issue 2,2011, pp.181196.
[19] M. Boryczka and W. Bura, Ant Colony Optimization for the
Multi-criteria Vehicle Navigation Problem Studies in Computational Intelligence, Volume 456, 2013, pp. 29-53.
[20] M. Glabowski, B. Musznicki, P. Nowak and P. Zwierzykowski,
Shortest Path Problem Solving Based on Ant Colony Optimization Metaheuristic Image Processing and Communication, Volume
17, Issue 1-2, 2013, pp.7-17.
[21] L. Andrei, Analysis of Ant Colony Optimization for Dynamic
Shortest Path Problems, 2012.
[22] Y. Jiang, W. Wang and Y. Zhao, Solving the Shortest Path Problem
in Vehicle Navigation Systems by Ant Proceedings of 7th WSEAS
International Conference on Signal Processing,Computational Geometry and Artificial Vision,2007,pp. 188-192.
[23] A. M. Ghazy, F. EL-Licy, and H. A. Hefny, Threshold Based
AntNet Algorithm for Dynamic Traffic Routing of Road Networks, Egyptian Informatics Journal, Volume 13, Issue 2,2012,
pp. 111-121.
[24] S. Kurihara , Traffic Congestion forecating Algorithm Based on
Pheromone Communication Model, Ant Colony Optimization Techniques and Applications, Dr. Helio Barbosa (Ed.), ISBN 97895351-1001-9, 2013.
[25] B. Tatomir, L.J.M Rothkrantz, and A.C Suson, Travel Time Prediction for Dynamic Routing Using Ant Based Control, presented
at Conference on Winter Simulation (WSC), Austin, TX, 2009, pp.
1069 - 1078.
[26] J.J Kponyo, Y. Kuang, Z. Li, Real Time Status Collection and
Vehicular Traffic Control Using Ant Colony Optimization, International Conference on Computational Problem-Solving (ICCP),
2012, pp. 69 - 72.
[27] J. Kponyo, Y. Kuang, and E. Zhang, Dynamic Travel Path Optimization System Using Ant Colony Optimization, Proceedings
of the 2014 UKSim-AMSS 16th International Conference on
Computer Modelling and Simulation. IEEE Computer Society,
2014, pp.141-147.
[28] J Kponyo, Y Kuang, E Zhang, K Domenic, VANET clusteron-Demand Minimum Spanning Tree (MST) Prim Clustering
Algorithm, in 2013 International Conference on Computational
Problem-solving (ICCP), Jiuzhai, 26-28, October 2013,pp. 101104.

30