VOL. X, NO.

X, XXXXXXXX ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

A REVIEW OF FIREFLY ALGORITHM

Xxxxx Xxxx 1, Xxxx X. Xxxxx 2, Xxxx. Xxxx Xxxxx 2

1
Microwave Research Group, Centre for Telecommunication Research and Innovation (CeTRI), Universiti
Teknikal Malaysia Melaka (UTeM), 76100 Durian Tunggal, Melaka, Malaysia
2
Department of Engineering Mechanics, Universiti Tun Hussein Onn Malaysia.
E-Mail: xxxxxx@xxxxxx.xxx

ABSTRACT
Firefly algorithm is one of the swarm intelligence that evolve fast for almost area of optimization and engineering
problems. Stand alone firefly algorithm already has managed to solve problems. For problems that have multi dimensional
and nonlinear problem, some modification or even hybridization with the other metaheuristic is advisable. This
modification and hybridization is to aim for help for the computational constrain and it will become more flexible and
more efficient.

Keywords: Firefly Algorithm Hybridization Modification

1
 VOL. X, NO. X, XXXXXXXX ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

INTRODUCTION characteristic. Firefly is an insect that mostly produces

The algorithm can be categorized into many short and rhythmic flashes that produced by a process of
classifications, one of its deterministic or stochastic. bioluminescence. The function of the flashing light is to
Deterministic algorithm is an algorithm that produce on attract partners (communication) or attract potential prey
the given input, and it will produce same result by and as a protective warning toward the predator. Thus, this
following the same computational steps. Deterministic the intensity of light is the factor of the other fireflies to
algorithm is quite efficient in finding local optima because move toward the other firefly.
of it will do local search. The light intensity is varied at the distance from
Stochastic can be defined in two group, heuristic the eyes of the beholder. It is safe to say that the light
and metaheuristic. Heuristic means ‘to find’ or ‘to intensity is decreased as the distance increase. The light
discover by trial and error. The quality solutions for a intensity also the influence of the air absorb by the
tough optimization problem can be found but, there is no surroundings, thus the intensity becomes less appealing as
guarantee that solution is the optimal solution. the distance increase (Yang, 2010). Firefly algorithm was
Metaheuristic is generally better than heuristic algorithm followed three idealize rules, 1) Fireflies are attracted
since the search process is randomization and local search toward each others regardless of gender. 2) The
(Yang, 2010). Even though metaheuristic only able to attractiveness of the fireflies is correlative with the
provide “acceptable” solution, but it leaves to tackle tough brightness of the fireflies, thus the less attractive firefly
and complex problem in engineering and science field in a will move forward to the more attractive firefly. 3) The
fair time (Talbi, 2009). There are two main components in brightness of fireflies is depend on the objective function
metaheuristic, diversification or exploration and (Yang, 2010).
intensification or exploitation (Yang, 2010), (Blum and Structure of firefly algorithm
Roli, 2003). In firefly algorithm, there are two important
Diversification or exploration is to explore the variables, which is the light intensity and attractiveness.
global search space thoroughly and able to generate Firefly is attracted toward the other firefly that has brighter
diverse solutions. Intensification or exploitation is to flash than itself. The attractiveness is depended with the
exploit the local information for the search process and light intensity.
generate current good solutions (Yang, 2010), (Yang and
Karamanoglu, 2013). Since diversification is about global
search, the advantage is it will increase the chance to gain
the optimality globally. The disadvantages are the process
will be slower and less convergence rate. While the
intensification is about local search, the optimization will
be converging quickly. The disadvantages are, it will may
lead to premature convergence and it will reduce the
chance to find global optimum (Yang and Karamanoglu,
2013). Seeing that diversion is better in global search with
diverse solutions, while intensification is fast in search
space thus will use less time. Balancing between these two
components with the selection of best solutions will be a
great help to find higher chance of diversity solutions, in
less time taken and global optimality is achievable (Yang
& Press 2010), (Blum and Roli, 2003), (Yang and
Karamanoglu, 2013).
Swarm intelligence in a system of collective Figure 1: Pseudo code for Firefly Algorithm (Yang, 2010).
agent that interact with the surrounding environment that
perform global pattern. This intelligence compose base for The light intensity thus attractiveness is inversely
the evaluation, comparing and imitation. Swarm proportional with the particular distance r from the light
intelligence system is act as in their coordinated without source. Thus the light and attractiveness is decrease as the
external disturbance. In years, the numbers of swarm base distance increase.
optimization is increased such as Particle Swarm
Optimization (PSO), Artificial Bee Colony optimization (1)
(ABC) and Firefly Algorithm (FA) for robot path planning
(Pal and Sharma, 2013).
I = light intensity,
FIREFLY ALGORITHM = light intensity at initial or original light
intensity,
Introduction of Firefly Algorithm = the light absorption coefficient
In the book of Yang (Yang, 2010), there is an
explanation of how the algorithm that follows the firefly r = distance between firefly i and j

2
 VOL. X, NO. X, XXXXXXXX ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Modified firefly algorithm.

Attractiveness is proportionally to the light intensity seen In standart firefly algorithm, global best is the
by the another fireflies, thus attractiveness is β current best solution of firefly or known as the firefly that
hold the highest light intensity or attractiveness. This
(2) firefly will move randomly to find the next best firefly for
the next iteration. Since the firefly is move randomly, the
attractiveness will be lose in certain distance from the
= Attractiveness at r is 0
others fireflies. Thus this will lead to the loose of
The distance between two fireflies can define using performance in that particular iteration (Yang, 2010),
Cartesian distance (Tilahun and Ong, 2012), (Hassanzadeh and Meybodi,
2012).
In paper of (Tilahun and Ong, 2012) propose that,
(3) instead of the firefly move in random behavior, but in the
proper manner. The firefly is only move to the direction if
only there is improvement of brightness. The author
Firefly i is attracted toward the more attractive firefly j, the (Tilahun and Ong, 2012) propose modification is by
movemeunt is defined as generate unit vector, then to determine the movement of
firefly is by direction that leads toward the increases of
brightness firefly. If there is none; the current brightest
(4)
firefly will stay in the current position. As for (Wang et
al., 2012), in the paper also done some modification
In equation (4), the first term is for attraction, is the toward firefly algorithm for improvement. First is to add
Lévy flight for improvement in term of localized searching
limitation when the value is tend to zero or too large. If
for closer solution. In paper written by (Yu et al., 2013),
approaching zero ( , the attractiveness and Olamaei, Moradi and Kaboodi, 2013), (Farahani et al.,
brightness become constant, . In another word, a 2011) is to proposed to use adaptive formulation for
firefly can be seen in any position, easy to complete global randomization value . This is because when the value of
search. If the is nearing infinity or too large ( , the is large it is better for firefly to explore unknown place
attractiveness and brightness become decrease. The firefly while small value of will make firefly for local search.
movements become random. The implementation of In (Farahani et al., 2011) the author also propose the
firefly algorithm can be done in these two asymptotic movement of firefly that is change from random
behaviors. While the second the term is for randomization, movement to direct movement. This is happen when there
as is the randomize parameter. The can be replace by is no brighter firefly from the particular firefly, then firefly
ran -1/2 which is ran is random number generated from 0 will move randomly. To avaoid this, the author propose
to 1. that the random movement is change to direct movement
and the firefly will move toward the best solution in that
Variation of firefly algorithm iteration. This will make firefly in the better position for
Firefly algorithm is widely use to solve many next iteration and help to achive to near global best.
problems, such as solving the economic emissions load
dispatch problem (Apostolopoulos and Vlachos, 2011), Hybrid firefly algorithm
multilevel image thresholding selection (Horng & Jiang Solving problem with only one metaheuristic is
2010), finding optimal test sequence generation rather restrictive. This is because the single metaheuristic
(Srivatsava, Mallikarjun and Yang, 2013), solving is hard to reach the an optimal solution within reasonable
travelling salesman problem (Kumbharana and Pandey, time. Thus by combining with other metaheuristic or
2013), vector quantization for image compression (Horng, called hybridization will help with the high dimentional
2012) and object tracking (Gao et al., 2013). and nonlinear problems. Hybridize metaheuristic also can
Despite that firefly algorithm is widely used for provide a more efficient behavior and a higher flexibility
solving problem, standart firefly algorithm also have when dealing with real-world and large scale problems
some coming in term of trapping into several local optima (Abdullah et al., 2012), (El-Sawy, Zaki and Rizk-Allah,
when solving complex problem (Farook and Raju, 2013), ( 2012), (Rizk-Allah, Zaki and El-Sawy, 2013).
Yu, Yang and Su, 2013). There is disadvantages of using In this paper, the author (Farahani et al., 2012)
of single method only, this is because it will be overly propose of using Genetic Algorithm hybridize with Firefly
restrictive for high dimensional and nonlinear problem. Algorthim. This is because of the basic firefly algorithm
Thus, some modification and hybridization is suggested to weakness is in global optima. Thus, by using genetic
overcome the shortcoming of single method (Abdullah et algorithm which is more suitable for search globally, it
al., 2012). will find better solution. By completing this, firefly
algorithm will be used the solution for local search.

3
 VOL. X, NO. X, XXXXXXXX ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

There is hybridization done after completing Some modification done by (Kumbharana and
between two or more algorithms paralleled. In the paper Pandey, 2013) in his paper to adapt TSP with FFA which
by (El-Sawy, Zaki and Rizk-Allah, 2012), the Ant Colony are initial solution, distance function and movement. In
Optimization and Firefly algorithm is done parallel and the initial solution, the pseudo code by Yang describe that the
result is sorted base on the constrain violation by fireflies is scattered over the search space. But with good
descending order of the feasibility rule. The rules are solution among random solutions, it will make immediate
grouped based on, the feasible solution preferable compare advance toward better solution. By using this method, it
to others infeasible solution, or if there are two feasible will make faster to find local optima. As for distance
solution, which is have better objective function is function, instead of using Cartesian distance, the author
preferable, and last is if there is none feasible solution, uses Hamming’s distance and numbers of swapping
which is have small constrain of violation is preferable. required being same with the next solution. For the
In the hybrid firefly algorithm introduce by movement, the firefly is represented in form of
(Abdullah et al., 2012) is combination between firefly permutation matrices, and then inverse it to preserve the
algorithm and differential evolution. The population of previous path formed.
firefly will be produce into two group, one is in the with Firefly algorithm also has been use in the
the potential fitness values, and will proceed to calculate research of vehicle routing problem. Vehicle routing is a
the distance between solution using Euclidean distance problem to find the set of route for the fleet of vehicles to
and attractiveness. The others sub population which is serve the number of stops. The number of stops is
contains of less significant fitness value will used deterministic, when there is no demand exceed the
evolutionary operation of differential evolution. This will capacity of the vehicle, it is the standards vehicle routing
produce offspring solution, if these offspring solutions problem. The VRP can be extend into many constrains
have better fitness values, it will be replace the original such as time window, mix pick up (backhauls) and
solution. If not, the original solution will be remains till deliveries (linehauls) (Breedam, 2001).
next iteration. There are many variation of vehicle routing
problems, one of it is Vehicle routing problem with time
Application of firefly algorithm windows (VRPTW). VRPTW is aim for to find route to
Firefly algorithm is vastly used for solving the all vehicle to such that all the customer will be serve
engineering problem and optimization. Even only standart within the respective time windows. In practice, the
firefly apply for solving problem can produce superior customers and the carrier companies have difference
result, but some modification or hybridization is done for concerns. For the customers, is to receive the deliveries on
convenient of application to apply firefly algorithm to time, while the carries company is to delivers the goods to
solving multi dimentional and nonlinear problem. the different customers efficiently while be able to cut cost
and save the time (Tas et al., 2013). This paper of (Pan et
Routing problem. al., 2013), the authors adopt the VRPTW coding from
Firefly algorithm is adapt from the behavior of CVRP. The modification of infeasible solution is done, so
firefly that attracted to the more attractive firefly (Yang, that only the effective solution that have the requirement
2010). Thus, by using the characteristic of firefly, of value will be kept.
researcher apply firefly algorithm to fint the shortest path Finding path by using firefly algorithm also is not
or route for vehicle, bus or salesman. Below is example of restricted for road vehicle, firefly algorithm also had
application of firefly algorithm to find the shortest or applied for finding route of uninhabited combat air vehicle
optimum route for the problem. There are many variable (UCAV). There are constrain of UCAV that should not be
and conditions that the researcher to deal in order to solve over looked. The route planning for UCAV should include
the problems. the terrain, data, threat information, fuel and time (Wang
Travelling salesman problem is to find the et al., 2012). Thus in the paper, the author (Wang et al.,
shortest distance in tour city that visited once. The 2012) do improvement of FA algorithm, so that it able to
salesman have to find the shortest distant that can cover all search the most optimal route with multi constrains.
the stops, visited once and return to the starting point. In
paper by (Jati, Manurung and Suyanto, 2013), the author Image processing.
use evolutionary discrete firefly algorithm (EDFA) with Firefly algorithm in image processing wold also
some modification from the original. In the original is not new, there are bunch of research base on image
EDFA, the firefly did not have direction of movement. processing using firefly algorithm such as multilevel
Thus, movement in new EDFA is adapt from the evolution image thresholding selection (Horng and Jiang, 2010),
strategies (ES), which is used the inversion mutation. In active contour model for medical image segmentation
the paper, tha author also discuss about discrete firefly (Sahoo and Chandra, 2013) and vector quantization using
algorithm (DFA). The movement of firefly in DFA is the firefly algorithm for image compression (Horng,
different from EDFA. The total number of firefly in EDFA 2012).
is constant per generation. While the total number of The example of how to include a figure in one
firefly in DFA is changing base on total number of firefly column is shown in Figure-1.
with best objective function.

4
 VOL. X, NO. X, XXXXXXXX ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Figure-1. Variant of slump with laterite content of laterized concrete.

2 1 4.5
In paper of (Horng, 2012), had used firefly 3 4 2.9
algorithm to solve vector quantization by hybridize the CONCLUSIONS
firefly algorithm with Linde–Buzo–Gray, (FFA-LBG).
LBG is a scalar quantization that used for input vectors for Firefly algorithm is considering new algorithm in
determining the codebook. In FFA-LBG, LBG is put at the the swarm intelligence family. Despite that, the usage of
first part of the initial solutions. The author conclude that the firefly algorithm in the various types of problem shows
FFA-LBG method is achievable same par with the PSO- that the anticipation from the researcher to use this
LBG and QPSO-LBG with the less the computation time. algorithm. This algorithm already proves that it is superior
In image processing, there is one segment for compared to the previous introduce swarm intelligence
image tracking using video sequence. Object tracking is from the research done before. Even though the firefly
one of the research that is vastly studied into real world algorithm has proven to be superior compared to the
application such as surveillance, vision based control and previous swarm intelligence, some modification can be
robotics. On the other hand, video sequencing for object done to improve the local search a well as global search to
tracking still a challenging topic for researches due to ensure the solution obtains is the optimum and not
large amount of data used and high specification for real premature solution.
time computation (Gao et al., 2013). The authors state that, Firefly algorithm also suitable be used for the
to overcome problems in the object tracking the high dimensional and nonlinear problems. The downside
researcher used the varieties of methods and algorithms. of it is the single that the single metaheuristic is hard to
One of the optimization algorithm that is doing well is reach the an optimal solution within a reasonable time.
meanshift. From this research, the authors conclude that Thus, by combining the metaheuristic will help to
the firefly algorithm is indeed superior compare to particle overcome the shortcoming of the single metaheuristic
filter, particle swarm optimization and meanshift in algorithm.
tracking object. The result is compare in four different In the future work, the researcher should tackle
method and shows that firefly algorithm out perform in firefly in more various types of problem such as find
term of speed and accuracy. optimum route for new build trains rails route which is
Table-1 shows an example of a table. If the table have multiple constrains such as to preserve the nature
is large, your can put the table in a single column format. maximum as possible and multi tracking for object
tracking
Table-1. This is the title here.
REFERENCE
x y Result Abdullah, A., Deris, S., Mohamad, M. and Hashim, S.
0 0 2.3 (2012). A new hybrid firefly algorithm for complex and

5
 VOL. X, NO. X, XXXXXXXX ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

nonlinear problem. Distributed Computing and Artificial Computing (UIC/ATC), 2010 7th International
Intelligence, Springer Berlin Heidelberg, pp.673--680. Conference on. IEEE, pp.58--63.

Apostolopoulos, T. and Vlachos, A. (2010). Application of  Jati, G., Manurung, R. and Suyanto, (2013). Discrete
the firefly algorithm for solving the economic emissions Firefly Algorithm for Traveling Salesman
load dispatch problem. International Journal of Problem. Swarm Intelligence and Bio-Inspired
Combinatorics, 2011. Computation, pp.295--312.

Apostolopoulos, T. and Vlachos, A. (2011). Application of Horn, M. (2012). Vector quantization using the firefly
the Firefly Algorithm for Solving the Economic Emissions algorithm for image compression. Expert Systems with
Load Dispatch Problem. International Journal of Applications, 39(1), pp.1078--1091.
Combinatorics, 2011, pp.1-23.
Kumbharana, S. and Pandey, G. (2013). Solving
Blum, C. and Roli, A. (2003). Metaheuristics in Travelling Salesman Problem using Firefly Algorithm.
combinatorial optimization: Overview and conceptual International Journal for Research in Science &
comparison. ACM Computing Surveys (CSUR), 35(3), Advanced Technologies, 2(2), pp.53--57.
pp.268--308.
Merz, P. and Freisleben, B. (2001). Memetic algorithms
Breedam, A. (2001). Comparing descent heuristics and for the traveling salesman problem. Complex Systems,
metaheuristics for the vehicle routing problem. Computers 13(4), pp.297--346.
& Operations Research, 28(4), pp.289--315.
Olamaei, J., Moradi, M. and Kaboodi, T. (2013). A new
El-Sawy, A., Zaki, E. and Rizk-Allah, R. (2012). A Novel adaptive modified Firefly Algorithm to solve optimal
Hybrid Ant Colony Optimization and Firefly Algorithm capacitor placement problem. pp.1--6.
for Solving Constrained Engineering Design Problems.
Journal of Natural Sciences and Mathematics, 6(1), pp.1- Pal, N. and Sharma, S. (2013). Robot Path Planning using
22. Swarm Intelligence: A Survey. METHODS, 83(12).
Pan, F., Ye, C., Wang, K. and Cao, J. (2013). Research on
Farahani, S., Abshouri, A., Nasiri, B. and Meybodi, M. the Vehicle Routing Problem with Time Windows Using
(2011). A Gaussian firefly algorithm. International Firefly Algorithm. JCP, 8(9), pp.2256--2261.
Journal of Machine Learning and Computing, 1(5),
pp.448--453. Rizk-Allah, R., Zaki, E. and El-Sawy, A. (2013).
Hybridizing ant colony optimization with firefly algorithm
Farahani, S., Abshouri, A., Nasiri, B. and Meybodi, M. for unconstrained optimization problems. Applied
(2012). Some hybrid models to improve firefly algorithm Mathematics and Computation, 224(3), pp.473--483.
performance. Int J Artif Intell, 8(S12), pp.97--117.
Sahoo, A. and Chandra, S. (2013). L'evy-flight firefly
Farook, S. and Raju, P. (2013). Evolutionary Hybrid algorithm based active contour model for medical image
Genetic-Firefly Algorithm for Global Optimization. segmentation. Contemporary Computing (IC3), 2013 Sixth
IJCEM International Journal of Computational International Conference on. IEEE, pp.159--162.
Engineering & Management, 16(3), pp.37-45.
Srivatsava, P., Mallikarjun, B. and Yang, X. (2013).
Gandomi, A., Yang, X. and Alavi, A. (2011). Mixed Optimal test sequence generation using firefly algorithm.
variable structural optimization using firefly algorithm. Swarm and Evolutionary Computation, 8, pp.44--53.
Computers & Structures, 89(23), pp.2325--2336.
Talbi, E. (2009). Metaheuristics. 1st ed. Hoboken, N.J.:
Gao, M., He, X., Luo, D., Jiang, J. and Teng, Q. (2013). John Wiley & Sons.
Object tracking using firefly algorithm. IET Computer
Vision, 7(4), pp.227--237. Tas, D., Dellaert, N., Van Woensel, T. and De Kok, T.
(2013). Vehicle routing problem with stochastic travel
Hassanzadeh, T. and Meybodi, M. (2012). A new hybrid times including soft time windows and service costs.
algorithm based on Firefly Algorithm and cellular learning Computers & Operations Research, 40(1), pp.214--224.
automata. Electrical Engineering, (ICEE) 2012, 20th
Iranian Conference on. IEEE, pp.628--633.

Horng, M. and Jiang, T. (2010). Multilevel image

thresholding selection based on the firefly algorithm.
Ubiquitous Intelligence & Computing and 7th
International Conference on Autonomic & Trusted

6
 VOL. X, NO. X, XXXXXXXX ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Tilahun, S. and Ong, H. (2012). Modified firefly

algorithm. Journal of Applied Mathematics, pp.1-12.

Wang, G., Guo, L., Duan, H., Liu, L. and Wang, H.

(2012). A Modified Firefly Algorithm for UCAV Path
Planning. International Journal of Hybrid Information
Technology, 5(3), pp.123-144.

Yang, X. and Karamanoglu, M. (2013). Swarm

Intelligence and Bio-Inspired Computation. Swarm
Intelligence and Bio-Inspired Computation, pp.3-23.

Yang, X. (2009). Firefly algorithms for multimodal

optimization. Stochastic algorithms: foundations and
applications. Springer Berlin Heidelberg, pp.169--178.

Yang, X. (2010). Nature-Inspired Metaheuristic

Algorithms. 2nd ed. Frome: Luniver Press.

Yu, S., Yang, S. and Su, S. (2013). Self-Adaptive Step

Firefly Algorithm. Journal of Applied Mathematics, pp.1-
8.