Вы находитесь на странице: 1из 20

ACSE Journal, Volume (6), Issue (1), January, 2006

Ant Colony Algorithms in Diverse Combinational


Optimization Problems -A Survey
K.Thangavel1, M.Karnan2*, P.Jeganathan2, ,A.Petha lakshmi3 R.Sivakumar4, G.Geetharamani5
1
Department of Computer science, Periyar University, Salem, India
2
Department of computer science, Gandhigram Rural Institute-Deemed University,
Gandhigram-624302, TamilNadu, India.
3
Department of Computer science, Mother Theresa University, Kodaikkanal, India
4
Department of R&D, Hindusthan college of Engg and Tech, Coimbatore,
5
Department of Computer science Bharathidasan University,Trichy,india

Abstract relatively short time. Algorithms of this type are


loosely called heuristics. They often use some
Ant Colony Optimization (ACO) metaheuristic is problem-specific knowledge to either build or
a recent population-based approach inspired by improve solutions.
the observation of real ants colony and based
upon their collective foraging behavior. In ACO, Recently, many researchers have focused
solutions of the problem are constructed within a their attention on a new class of algorithms, called
stochastic iterative process, by adding solution metaheuristics. A metaheuristic is a set of
components to partial solutions. Each individual algorithmic concepts that can be used to define
ant constructs a part of the solution using an heuristic methods applicable to a wide set of
artificial pheromone, which reflects its experience different problems. In other words, a
accumulated while solving the problem, and metaheuristic can be seen as a general-purpose
heuristic information dependent on the problem. heuristic method designed to guide an underlying
In this survey paper, it is intended to summarize problem specific heuristic toward promising
the methods of ant colony system used in various regions of the search space containing high-
types of applications. In particular, routing, quality solutions. A metaheuristic therefore a
assignment, scheduling, subset, machine learning general algorithmic framework, which can be
and network routing problems. applied to different optimization problems with
relatively few modifications to make them,
Keywords: combinatorial optimization, adapted to a specific problem. The use of
metaheuristics, ant colony system, pheromone. metaheuristics has significantly increased the
ability of finding very high-quality solutions to
1 Introduction hard, practically relevant combinatorial
optimization problems in a reasonable time. This
Combinatorial optimization problems are is particularly true for large and poorly
intriguing because they are often easy to state but understood problems.
very difficult to solve. Many of the problems
arising in applications are NP-hard, that is, it is Metaheuristics or general local search
strongly believed that they cannot be solved to methods are growing at a very fast, exploring
optimally within polynomially bounded rate: the most important conferences in
computation time. Hence, to practically solve operational research have one or many sessions
large instances one often has to use approximate entirely devoted to metaheuristics ant there are
methods, which return near optimal solutions in a also journals entirely devoted to them. In the
operations research literature, many elegant

7
ACSE Journal, Volume (6), Issue (1), January, 2006

mathematical models and solution methods have and coworkers [Dorigo et al. 1991a] and inspired
been developed to cope with real-world problems. by the foraging behavior of ant colonies.
Although some exact approaches, such as branch Individuals ants are simple insects with limited
and bound algorithms and dynamic programs, memory and capable of performing simple
have been proposed to solve the problems actions. However, the collective behavior of ants
optimally, the curse of dimensionality of provides intelligent solutions to problems such as
numerous combinatorial problems hinders the finding the shortest paths from the nest to a food
search for optimal solutions in the solution space. source. Ants foraging for food lay down
Seeking approximate solutions that are feasible or quantities of a volatile chemical substance named
good enough in an acceptable time therefore pheromone, marking their path that it follows.
shapes a viable alternative for the decision Ants smell pheromone and decide to follow the
makers. Following this orientation, several meta- path with a high probability and thereby reinforce
heuristics, such as genetic algorithms [Goldberg it with a further quantity of pheromone. The
2003, Holland 1975], tabu search [Glover 1990] probability that an ant chooses a path increases
and simulated annealing [Kirkpatrick et al. 1983], with the number of ants choosing the path at
have been proposed to deal with the previous times and with the strength of the
computationally intractable problems. Ant colony pheromone concentration laid on it.
optimization (ACO) is a new meta-heuristic
developed for composing approximate solutions. This concept has been applied to hard
The ant algorithm was first proposed by [Colorni combinatorial optimization problems by creating
et al. 1991] and has been receiving extensive a population of artificial ants that searches for
attention due to its successful applications to optimal solutions according to the problem’s
many combinatorial optimization problems constraints. Artificial ants are characterized as
[Dorigo et al. 1999a]. Like genetic algorithm and agents that imitate the behavior of real ants.
simulated annealing approaches, the ant However, it should be noted that an artificial ACS
algorithms also foster its solution strategy through has some differences in comparison with a real
use of nature metaphors. The ACO is based upon ACS, as follows
the behaviors of ants that they exhibit when ƒ Artificial ants have memory
looking for a path to the advantage of their ƒ They are not completely blind
colony. Unlike simulated annealing or tabu ƒ They live in an environment where time is
search, in which a single agent is deployed for a discrete.
single beam session, ACO and genetic algorithms On the other hand, an artificial ACS has several
use multiple agents, each of which has its characteristics adopted from real ACS:
individual decision made based upon collective ƒ Artificial ants have a probabilistic preference
memory or knowledge. Recently, the ACO for paths with a larger amount of pheromone
metaheuristic has been proposed to provide a ƒ Shorter paths tend to have larger rates of
unifying framework for most applications of ant growth in their amount of pheromone
algorithms [Dorigo and Di Caro 1999, Dorigo et ƒ The ants use an indirect communication
al. 1999b] to combinatorial optimization system based on the amount of pheromone
problems. Algorithms that actually are deposited in each path.
instantiations of the ACO metaheuristic will be
called ACO algorithms. This paper summarizes The key idea is that, when a given ant has to
various implementations of ACO algorithms in choose between two or more paths. The path that
different applications. was more frequently chosen by other ants in the
past will have a greater probability of being
The rest of the paper is organized as follows: chosen by the ant. Therefore, trails with greater
section 2 introduces the ant colony optimization amount of pheromone are synonyms of shorter
procedure, section 3 discusses several methods of paths. In essence, an ACS iteratively performs a
ACO technique in various applications, and loop containing two basic procedures, namely:
section 4 presents conclusion for this study on ant
colony optimization algorithms. ƒ a procedure specifying how the ants
construct/modify solutions of the problem
2 Ant Colony Optimization being solved.
ƒ a procedure to update the pheromone trails.
Ant Colony Optimization (ACO) is a population-
based approach first designed by Marco Dorigo

8
ACSE Journal, Volume (6), Issue (1), January, 2006

The construction/modification of a solution is Start searching for optimum value in the


performed in a probabilistic way. The probability search space, from a random position.
of adding a new item to the current partial Update the pheromone values of all the
solution is given by a function that depends on a ants using the following equation.
problem-dependent heuristic and on the amount Tnew ← (1 – ρ) * Told + ρ * T0 ,
of pheromone deposited by ants on this trail in the where Tnew is the updated pheromone value
past. The updates in the pheromone trail are Told is the old pheromone value
implemented as a function that depends on the Compute the solution for each ant.
rate o f pheromone evaporation and on the quality Locally Search for the optimum value.
of the produced solution. To realize an ACS one Compare the Local Maximum with the Global
must define: Maximum.
ƒ An appropriate representation of the problem, If Local_Maximum > Global_Maximum
which allows the ants to incrementally Global_Maximum= Local_Maximum
construct/modify solutions through the use of End
a probabilistic transition rule based on the
amount of pheromone in the trail and on a Update the pheromone of the ant, which
local heuristic: generates the Local Maximum using the
ƒ A heuristic function that measures the quality following equation: Tnew ← (1 – α) * Told + α *
of items that can be added to the current
∆Told, where α is a constant value.
partial solution.
Update the pheromone of the remaining
ƒ A method to enforce the construction of valid
ants using the following equation:
solutions, i.e. solutions that are legal in the
Tnew ← (1 – α) * Told, where α is a
real-world situation corresponding to the
constant value.
problem definition:
End
ƒ A rule for pheromone updating, which
specifies how to modify the pheromone trail.
ƒ A probabilistic rule of transition based on the This algorithm shows a typical procedure for
value of the heuristic function and on the implementation of an ACO algorithm. The main
contents of the pheromone trail. procedure of the ACO metaheuristic have three
components: (1) management of the ants’ activity,
Typical procedure for ant colony algorithm: (2) pheromone updating, and (3) daemon actions.
S ← Search space
N ← Number of Iterations
3 Applications Of Ant Colony
K ← Number of Ants Optimization Algorithms
T0 ← Initial pheromone value
ρ ← Constant value for pheromone update Nowadays numerous successful
Repeat for N times implementations of the ACO metaheuristic are
Initialize all the ants with initial available and they have been applied to many
pheromone value (T0 = 0.001). different combinatorial optimization problems.
These applications are summarized in Table-1 and
they are discussed in this section

Table 1. Current Applications of ACO algorithms listed according to problem types


Problem Type & Name Main References
Routing
Traveling Salesman Bullnheimer et al. 1999c, Cordon et al. 2000, Dorigo and Gambardella
1997a, Dorigo and Gambardella 1997b, Dorigo et al. 1991a, Dorigo et al.
1991b, Dorigo et al. 1996, Gambardella and Dorigo 1995, Stutzle and Hoos
1997, Stutzle and Hoos 2000, Randall 2002, Tsai et al. 2004
Vehicle Routing Bell and McMullen 2004, Bullnheimer et al. 1999a, Bullnheimer et al.
1999b, Gambardella et al. 1999, Reimann et al. 2002, Mazzeo and Loiseau
2004.
Sequential Ordering Gambardella and Dorigo 1997, Gambardella and Dorigo 2000.
Assignment

9
ACSE Journal, Volume (6), Issue (1), January, 2006

Problem Type & Name Main References


Quadratic Assignment Maniezzo 1999, Maniezzo and Colorni 1999, Maniezzo et al. 1994,
Solimanpur et al. 2004, Solimanpur et al. 2005, Stutzle 1997, Stutzle and
Hoos 2000, Talbi et al. 2001.
Graph Costa and Hertz, 1997, Gamez and Puerta 2002, Gosavi et al. 2003,
Korosec et al. 2004, Li and Wu 2003, Shyu et al. 2005.
Generalized Assignment Eggers et al. 2003, Gutjahr 2002, Hoshyar et al. 2000, Lee Z.J. et al. 2002,
Lourenco and Serra 1998, Lourenco and Serra 2002, Song et al. 1999,
Stutzle and Hoos 2000, Xia et al. 2003, Yin 2003.
Frequency Assignment Maniezzo and Carbonaro 2000, Lim et al. 2005.
University Course Azimi 2005, Socha et al. 2002, Socha et al. 2003.
timetabling
Scheduling
Job shop Colorni et al. 1994, Lee Z.J. and Lee 2004, Sendova-Franks and Lentz
2002.
Open shop Li and Wu 2002, Pfahringer 1996, Ying and Liao 2004.
Flow shop Gajpal and Rajendran 2005, Shyu et al. 2004, Stutzle 1998, Tkindt et al.
2002.
Total Tardiness Bauer et al. 2000
Total weighted tardiness den Besten et al. 2000, Gagne et al. 2002, Merkle and Middendorf 2000,
Merkle and Middendorf 2003
Project Scheduling Annaluru et al. 2004, Jayaraman et al. 2000, Merkle et al. 2000, Merkle et
al. 2002, Samrout et al. 2005.
Group shop Blum 2002, Blum 2003a.
Subset
Multiple Knapsack Leguizamon and Michalewicz 1999.
Max Independent set Leguizamon and Michalewicz 2000.
Redundancy Allocation Liang and Smith 1999, Zhao et al. 2005.
Set Covering Hadji et al. 2000
Weighted Constraint Graph Cordone and Maffioli 2001, Reimann and Laumanns 2005
Tree Partition
Arc-weighted l-cardinality Bullnheimer et al. 1999b
tree
Maximum Clique Fenet and Solnon 2003
Machine Learning
Classification Rules Parpinelli et al. 2002, Vijayakumar et al. 2003, Zheng et al. 2003.
Bayesian Networks de Campos et al. 2002
Fuzzy Systems Bullnheimer et al. 1999c, Kuo et al. 2005
Network Routing
Connection-oriented Bean and Costa 2005, Bonabeau et al. 1998, Di Caro and Dorigo 1998b,
Network routing Schoonderwoerd et al. 1997, Su et al. 2005, White et al. 1998.
Connectionless Network Di Caro and Dorigo 1997, Di Caro and Dorigo 1998a, Di Caro and Dorigo
Routing 1998c, Heusse et al. 1998, Subramanian et al. 1997, Van der put 1998.
Optical Network Routing Varela and Sinclair 1999.
Image Processing
Detection of Thangavel et al.2005b.
Microcalcification in Digital
Mammograms
Other
Shortest common Michel and Middendorf 1998, Michel and Middendorf 1999
supersequence
Constraint Satisfaction Solnon2000, Solnon 2002
2D-HP Protein folding Shmygelska et al. 2002
Bin Packing Levine and Ducatelle 2003

10
ACSE Journal, Volume (6), Issue (1), January, 2006

Problem Type & Name Main References


General Abd-El-Barr et al. 2003, Baykasoglu et al. 2004, Chen and Meng 2004,
Chen et al. 2004, Dorigo et al. 2000, Ge et al. 2004, Karaboga et al. 2004,
Monmarche et al. 2000, Papadopoulos and Linke 2004, Taillard et al. 2001,
Tsai et al. 2004, Xiong and Wei 2002
Ants of the artificial colony are able to generate
successively shorter feasible tours by using
information accumulated in the form of a
3.1 Routing Problems pheromone trail deposited on the edges of the
TSP graph. Computer simulations demonstrate
This section considers routing problems, that is, that the artificial ant colony is capable of
problems in which one or more agents have to generating good solutions to both symmetric and
visit a predefined set of locations and whose asymmetric instances of the TSP. The method is
objective function depends on the ordering in an example, like simulated annealing, neural
which the locations are visited. networks and evolutionary computation, of the
successful use of a natural metaphor to design an
Bell and McMullen (2004) applied the optimization algorithm. There are many ways in
meta-heuristic method of ant colony optimization which ACS can be improved so that the number
(ACO) to an established set of vehicle routing of tours needed to reach a comparable
problems (VRP). The procedure simulates the performance level can diminish, making its
decision-making processes of ant colonies as they application to larger problem instances feasible.
forage for food and is similar to other adaptive
learning and artificial intelligence techniques such The most appropriate parallelization
as Tabu Search, Simulated Annealing and Genetic technique is ultimately dependent upon the nature
Algorithms. Modifications are made to the ACO of the problem being solved. Problems like the
algorithm used to solve the traditional traveling TSP in which each of the solution elements may
salesman problem in order to allow the search of be easy to compute; yet each solution contains
the multiple routes of the VRP. Experimentation many such elements. Randall (2002) evaluated
shows that the algorithm is successful in finding the Parallel Ants scheme in which ants construct
solutions within 1% of known optimal solutions tours in parallel. A master ant is used to
and the use of multiple ant colonies is found to coordinate the activities of the colony. This
provide a comparatively competitive solution scheme is conceptually simple and suitable for the
technique especially for larger problems. popular MPI model on MIMD architectures. The
Additionally, the size of the candidate lists used results showed that acceptable speedup and
within the algorithm is a significant factor in efficiency could be achieved for larger problems.
finding improved solutions, and the However, one of the disadvantages to this scheme
computational times for the algorithm compare is the large amount of communication required to
favorably with other solution methods. maintain the pheromone matrix.

The Vehicle Routing Problem (VRP) C.F. Tsai et al. (2004) presented a new
requires the determination of an optimal set of metaheuristic approach called ACOMAC
routes for a set of vehicles to serve a set of algorithm for solving the traveling salesman
customers. Mazzeo and Loiseau (2004) dealt here problem (TSP) by introducing multiple ant clans’
with the Capacitated Vehicle Routing Problem concept from parallel genetic algorithm to search
(CVRP) where there is a maximum weight or solution space that using various islands to avoid
volume that each vehicle can load. We developed local minima and thus yield global minimum or
an Ant Colony algorithm (ACO) for the CVRP near global minimum to the traveling salesman
based on the metaheuristic technique introduced problem. In addition, we present two methods
by Colorni, Dorigo and Maniezzo. And presented called the multiple nearest neighbor (NN) and the
preliminary results that show that ant algorithms dual nearest neighbor (DNN) to ACOMAC to
are competitive with other metaheuristics for improve large TSPs. According to the simulation
solving CVRP. results, adding the NN or DNN approach to
ACOMAC or ACS, as initial solutions, also
Dorigo and Gambardella (1997a) significantly enhances the performance of
described an artificial ant colony capable of ACOMAC and ACS in solving the traveling
solving the traveling salesman problem (TSP).

11
ACSE Journal, Volume (6), Issue (1), January, 2006

salesman problem. Meanwhile, using


ACOMAC+DNN with TSP can yield better Gosavi et al. (2003) supplied more
solutions than the other stated approaches. results for an extended ant colony algorithm,
Moreover, ACOMAC or ACOMAC+ NN, which can he used to compute a minimum cost
utilizing five ant clans with a total of 20 ants, is Steiner tree from a graph. In each pass of the
demonstrated to yield better solutions. proposed algorithm, ants are placed at the
Furthermore, ACOMAC with a local weighting terminal nodes of the Steiner tree to he computed.
(w) set to 0.6 can yield better solutions in terms of They are then allowed to move towards one
length. It is observed that a larger number of ant another, along the edges of the graph, until they
clans in DNN, NN or the multiple ant clans’ merge into a single entity. In this process, the
method will yield better solutions to the traveling paths taken by the ants define a Steiner tree.
salesman problem. However, more ant clans will Edges receive reinforcement in the form of
also increase the time required to solve the pheromone deposits along the paths taken by the
problem. Accordingly, the use of five clans with ant that is based on the quality of the Steiner tree
the proposed algorithm is recommended to solve produced. Pheromones eventually accrue most
the traveling salesman problem. along better edges. As a result, after several
iterations, a good Steiner tree can be extracted
3.2 Assignment Problems from the deposit. The algorithm can easily he
used in several practical applications.
The task in assignment problems is to assign a set
of items (objects, activities, etc.) to a given In the existed theorems, the results have
number of resources (locations, agents, etc.) been sharpened for a specific ACO algorithm to
subject to some constraints. Assignments can be the strength of the well-known convergence
represented as a mapping from a set I to a set J, property of the Simulated Annealing
and the objective function to minimize of the metaheuristic. As in Simulated Annealing, it turns
assignments done. out that a convergence guarantee can be obtained
by a suitable speed of “cooling” (i.e., reduction of
Eggers et al. (2003) addressed a new the influence of randomness): The geometric
optimization problem that we call the Keyboard pheromone decrement on not reinforced arcs
Arrangement Problem (KAP). It consists of effected by constant evaporation factor is too fast
finding the best possible arrangement of letters on and leads (in general) to premature convergence
a keyboard. Based on the generic framework of to sub-optimal solutions. On the other hand,
Ant Colony optimization, an algorithm is introducing a fixed lower pheromone bound stops
developed and applied to this problem. The cooling at some point and leads to random-
special feature of the algorithm is that the search-like behavior without convergence. In
evaluation and the construction of a solution by between lies a compromise of allowing
an ant can be performed simultaneously. Hence, it pheromone trails to tend to zero, but slower than
gives an important advantage compared to geometrically. This can be achieved either by
metaheuristics based on local search, for which decreasing evaporation factors, or else by
the evaluation of too many solutions would be “slowly” decreasing lower pheromone bounds. In
detrimental. a certain window of the cooling speed, get
convergence to the optimal solution with
Gamez and Puerta (2002) proposed a probability one. Future theoretical research
novel way of dealing with the graph triangulation should deal with the speed of convergence, and
problem. It is centered on the application of ant address the question of the expected running time
colony optimization, in which heuristic until the optimum is reached. Some interesting
knowledge is used during the construction of results concerning the last-mentioned question
potential solutions. The use of this knowledge has have recently been found for the field of
improved the results obtained by other evolutionary Algorithms. Also the Ant Colony
techniques, both in accuracy and efficiency. With Optimization approach might be accessible to
respect to the genetic algorithms used in order to such investigations [Gutjahr 2002].
test the ACO algorithms, it is clear that the
proposal of entering good solutions in the initial The excellent performance of recently
population helps to speed up the search process. proposed turbo codes depends strongly on their
As further work, apply new ACO approaches to interleaver pattern. Exhaustive searching for a
the problem studied here. good interleaving pattern is a very difficult task,

12
ACSE Journal, Volume (6), Issue (1), January, 2006

growing in the order of N!, where N is the own-force assets. The general idea of the
interleaver size. Hoshyar et al. (2000) presented a proposed algorithm is to combine the advantages
modified version of the ant colony system (ACS) of ACO, the ability to cooperatively explore the
algorithm for finding good interleaver patterns search space and to avoid premature convergence,
used in turbo codes. By defining an approximate and that of immune system (IS), the ability to
evaluation method for turbo codes with specified quickly find good solutions within a small region
interleaver patterns, a fitness function for use in of the search space. From our simulation for those
the modified ACS algorithm is obtained. The WTA problems, the proposed algorithm indeed is
simulation results show gains up to 0.7dB with very efficient.
sizes of 50 and 100 for the interleavers found by
the ACS algorithm compared with the randomly Li and Wu (2003) proposed an
generated interleavers. accelerated ant colony algorithm (AACA) for
complex nonlinear system optimization problems
Many real-world engineering problems based on a new objective-function-based heuristic
can be expressed in terms of partial differential pheromone assignment method. This approach
equations and solved by using the finite-element uses a bit encoding technique to map complex
method, which is usually parallelised, i.e. the nonlinear system optimization problems to an
mesh is divided among several processors. To oriented graph search problem, and enhance its
achieve high parallel efficiency it is important computational efficiency with a searching
that the mesh is partitioned in such a way that strategy adaptable to the specific topological
workloads are well balanced and interprocessor structures of solution spaces. Combining the
communication is minimized. Korosec et al. proposed method with its inherent prominent
(2004) presented an enhancement of a technique capability in dealing with discrete-space
that uses a nature-inspired metaheuristic approach optimization problems, it can be applied to solve
to achieve higher-quality partitions. The so-called various types of complex nonlinear system
multilevel ant-colony algorithm, which is a optimization problems in practice, such as
relatively new metaheuristic search technique for flexible process production planning,
solving optimization problems, was applied and communication network routing, and power
studied, and the possible parallelization of this system global control and scheduling. A practical
algorithm is discussed. The multilevel ant-colony problem of complicated hybrid production
algorithm performed very well and is superior to scheduling in a flexible drinking water process
classical k-METIS and Chaco algorithms; it is has been solved by AACA, demonstrating its
even comparable with the combined strong capability of solving complex system
evolutionary/multilevel scheme used in the optimization problems with interacted
JOSTLE evolutionary algorithm and returned continuous-time and discrete-event variables.
solutions that are better than the currently
available solutions in the Graph Partitioning Lim et al. (2005) developed a hybrid
Archive. There are many possibilities for heuristic to solve the bandwidth minimization
improving this algorithm. One possibility is in the problem that used hill climbing guided by ACO.
mapping of the graph onto the grid: with a proper The method was tested extensively and shown to
mapping convergence the results can be be comparable to the newly developed
improved. The use of a load-balancing method approaches. It shows that a good local search
between levels would also be a very promising combined with a high-level global search
way to go. The next possibility is in determining mechanism is very efficient for the bandwidth
which and how many vertices from the cell will minimization problem. In the ACO component,
be picked and with what probability. And, finally, two constructive methods are used alternately, an
we could change the way the pheromone is idea that can be explored further, for example, by
evaporated, deposited and restored. tuning these to find the best performance. Yet,
another interesting direction of research would be
Lee Z.J. et al. (2002) proposed an the parallel implementation of the ACO with hill
immunity-based ant colony optimization (ACO) climbing.
algorithm for solving weapon–target assignment
(WTA) problems. The WTA problem, known as a Maniezzo and Carbonar (2000)
NP-complete problem, is to find a proper considered the problem in assigning frequencies
assignment of weapons to targets with the to radio links between base stations and mobile
objective of minimizing the expected damage of transmitters in order to minimize the global

13
ACSE Journal, Volume (6), Issue (1), January, 2006

interference over a given region. This problem is models the distance between the locations is
NP-hard and few results have been reported on known a priori but it is sequence dependent for
techniques for solving it to optimality. We have the problem considered in this paper. An ant
applied to this problem an ANTS metaheuristic, algorithm is developed to solve the proposed
that is, an approach following the ant colony mathematical model. A technique is presented to
optimization paradigm. Computational results, efficiently implement the proposed ant algorithm.
obtained on a number of standard problem Experimentally derived rules to set the values of
instances, testify the effectiveness of the proposed parameters in different problems are presented.
approach. The performance of the proposed ant algorithm is
tested over a number of problems selected from
The problem of cell-to-switch the literature and is compared to many other
assignment is essential to the development of PCS algorithms existing in the literature. The
or global communication services. Shyu et al. experimental results reveal that the proposed ant
(2005) developed an ACO algorithm to solve this algorithm is effective and efficient for single row
problem. The problem is modeled in a form of the machine layout problems.
matching problem in a weighted directed bipartite
graph so that the artificial ants can construct their Talbi et al. (2001) proposed a powerful
paths on the graph. Such an abstract structure and robust algorithm for the QAP, which is based
might be helpful for ACO designers to model on ant colonies. Compared with previous ant
their own problems. Experiments are conducted systems for the QAP (HAS-QAP algorithm), we
to capture the behavior of ants in problem have refined the ants’ cooperation mechanism,
optimization and empirically study the both in the pheromone matrix update phase and in
performances of ACO and other approaches with the exploitation/diversification phase by using a
large-scale problem instances. frequency matrix. The search process of each ant
has also been reinforced with a local search
The inter-cell layout problem addressed procedure based on TS. Results show a noticeable
in this paper is an important issue in the increase in performance compared to HAS-QAP
successful implementation of CM. The problem is and also to parallel independent TS, thus
mathematically modeled as a QAP. One of the demonstrating the complementary gains brought
important data required for the layout of by the combined use of a powerful local search,
manufacturing cells is the flow of materials across ants-like cooperation and parallelism.
different cells. In earlier attempts, it is mostly
assumed a priori known data. Solimanpur et al. Xia et al. (2003) proposed an
(2004) introduced the sequence of operations and appropriately adapted ant colony system
the production volume of parts as two major embedded with a simple improved LBG
factors affecting the intercellular flow of algorithm for vector quantization codebook
materials. A mathematical model is proposed for design. The emphasis is put on the design of the
calculating the material flow among the cells. An probability transfer function and the tabu list in
ant algorithm is developed to solve the formulated the ant colony algorithm, the utilization of the
problem. The performance of the proposed next nearest neighborhood in the LBG algorithm,
algorithm is compared with other heuristics as well as the update of the pheromone in both
developed for facility layout problem as well as local and global sense. Experimental results show
many other ant algorithms recently developed for that the new algorithm outperforms other well-
QAP. The experimental results show that the known codebook design algorithms, and
proposed ant algorithm performs significantly particularly, the improvement of PSNR exceeds
better than the facility layout algorithms. 2dB compared with the conventional LBG
algorithm.
Solimanpur et al. (2005) dealt with the
single row machine layout problem in which the Yin (2003) proposed a new polygonal
size of machines and the clearance between the approximation method using a global search
machines are assumed different. The problem is heuristic called the ant colony search (ACS)
formulated as 0-1 non-linear mathematical algorithm. Its major components, namely, graph
models. It is shown that the formulated 0-1 non- representation, initial ant distribution, node
linear models are more intractable than the transition rule, and pheromone updating rule have
traditional QAP formulation of facility layout been investigated and adapted to the underlying
problem. This is due to the fact that in the QAP problem. The important properties of the

14
ACSE Journal, Volume (6), Issue (1), January, 2006

proposed method were thoroughly examined Simulated Annealing (SA), Tabu Search
through empirical studies. Inspired by research on (TS), Genetic Algorithm (GA), and Ant Colony
genetic algorithms, the elitist strategy and the System (ACS) are four of the main algorithms for
hybrid strategy are proposed. The performances solving challenging problems of intelligent
of the proposed methods were compared to those systems. In this paper, Azimi (2005) applied
of genetic-based and taboo search-based methods. these four techniques and three novel hybrid
The numerical results found are very combinations of them to a classical Examination
encouraging. Timetabling problem (ETP), an NP complete
problem. The novel hybrid algorithms consist of a
The difficulty of solving combined heat
Sequential TS–ACS, a Hybrid ACS/TS, and a
and power (CHP) economic dispatch lies in the
constraints imposed by the multi-objectives. The Sequential ACS–TS algorithms. These various
mutual dependencies of heat–power capacity hybrid combinations are then tested on 10
make it hard to find a feasible region, not to different scenarios of the classical ETP. Statistical
mention the optimum. Song et al. (1999) comparative analysis conclude that all of the three
presented a novel ant colony search algorithm proposed novel techniques are significantly better
(ACSA) approach for this problem. The main than each of their non-hybrid competitors, and
characteristics of the ACSA are positive
furthermore the Sequential ACS–TS provides the
feedback, distributed computation and the use of
a constructive greedy heuristic. Positive feedback superior solution of all. In the Hybrid ACS/TS,
accounts for the rapid discovery of good using TS is time-consuming for each ant and
solutions, distributed computation avoids therefore single iteration of ACS takes a
premature convergence, and the greedy heuristic substantially longer duration, and so for a
helps to find acceptable solutions in the early complete run of it. Therefore only a few iterations
stages of the search process. However, despite the of the main algorithm can be applied in the
attraction of the ACSA’s potential search ability, indicated points. The Sequential ACS–TS method
there are still some difficulties, such as the at first produces a good solution by using ACS
handling of constraints and premature but in final iterations of ACS, solutions are
convergence. This paper proposed to couple the converged, because of a lots of pheromone in
ACSA with other search techniques to improve its good regions and a poor amount of it in bad ones.
performance. The numerical results reported are In that time ACS causes poor improvement in the
encouraging. final solution and it is useless for us to continue it.
Therefore we interrupt ACS before its
To show that Ant Colony Optimization convergence and then use TS to search the
algorithms could be good alternatives to existing neighborhood region. TS search the neighborhood
algorithms for hard combinatorial optimization of solution and can find the minimum point in the
problems, recent research in this area has mainly close points. So, TS is chosen because; it worked
focused on the development of algorithmic better than other metaheuristic methods in cost
variants, which achieve better performance than reduction in the same time duration.
Ant System. Stutzle and Hoos (2000) presented
MAX–MIN Ant System (MMAS), an Ant Colony 3.3 Scheduling Problems
Optimization algorithm derived from Ant System.
MMAS differs from Ant System in several Scheduling is concerned with the allocation of
important aspects, whose usefulness we scarce resources to tasks over time. Scheduling
demonstrate by means of an experimental study. problems are central to production and
Additionally, we relate one of the characteristics manufacturing industries, but also arise in a
specific to MMAS — that of using a greedier variety of other settings. For all the machine-
search than Ant System—to results from the scheduling models considered in the following it
search space analysis of the combinatorial holds that (1) the processing times of all jobs and
optimization problems attacked in this paper. The operations are fixed and known beforehand and
computational results on the Traveling Salesman (2) the processing of jobs and operations cannot
Problem and the Quadratic Assignment Problem be interrupted.
show that MMAS is currently among the best
performing algorithms for these problems. Sorting can be an important mechanism
for the transfer of information from one level of
biological organization to another. Sendova-

15
ACSE Journal, Volume (6), Issue (1), January, 2006

Franks and Lentz (2002) studied the algorithm- permutation flowshop scheduling problem under
underlying worker sorting in Leptothorax ant study.
colonies. Worker sorting is related to task
allocation and therefore to the adaptive Jayaraman et al. (2000) presented a new
advantages associated with an efficient system for co-operative search approach, the ant colony
the division of labour in ant colonies. This paper optimization paradigm, for the optimal design of
considered four spatially explicit individual-based batch chemical processes and illustrates it by
models founded on two-dimensional correlated solving (1) the combinatorial optimization
random walk. This work aimed to establish problem of multiproduct batch scheduling and (2)
whether sorting at the level of the worker the continuous function optimization problem for
population could occur with minimal assumptions the design of multiproduct batch plant with single
about the behavioral algorithm of individual product campaigns and horizon constraints. The
workers. The behavior of an individual worker in ant algorithm is simple to implement and results
the models could be summarized by the rule of the case studies show its ability to provide
‘‘move if you can, turn always’’. All four models speedy and accurate solutions.
attained a level of worker sortedness that was
compatible with results from experiments on The resource allocation problem is to
Leptothorax ant colonies. It is found that the allocate resources to activities so that the cost
presence of a sorting pivot, such as the nest wall becomes as optimal as possible. Lee Z.J. and Lee
or an attraction force towards the centre of the C.Y. (2004) proposed a hybrid search algorithm
worker population, was crucial for sorting. And with heuristics for resource allocation problem
make a distinction between such pivots and encountered in practice. The proposed algorithm
templates and discuss the biological implications has both the advantages of genetic algorithm
of their difference. (GA) and ant colony optimization (ACO) that can
explore the search space and exploit the best
Annaluru et al. (2004) proposed an ant solution. In this implementation, both GA and
colony algorithm to determine the optimal ACO are well designed for the resource allocation
locations and ratings of capacitors in distribution problem. Furthermore, heuristics are used to
network far reactive power compensation. The ameliorate the search performance for resource
approach is multilevel. Two separate tables of allocation problem. Simulation results were
pheromones maintained by the algorithm. Ants reported and the proposed algorithm indeed have
generate solution stochastically, based on these admirable performance for tested problems.
pheromone tables. The pheromone tables are Because of the complementary properties of
updated periodically, so that pheromones accrue genetic algorithm and ant colony optimization,
more along better solutions. Results obtained by the hybrid approach outperforms other existing
the proposed algorithm have been compared with algorithms.
earlier schemes. This approach can be applied to
other practical problems requiring complex Li and Wu (2002) proposed a nested ant
decision making as well. colony algorithm to tackle the hybrid production
scheduling problems. A novel objective-function-
Gajpal and Rajendran (2005) The based heuristic pheromone assignment method is
problem of scheduling in permutation flowshops developed for pheromone update, ensuring that
with the objective of minimizing the completion- the amount of pheromone in path segments is
time variance of jobs is considered and solved by positively proportional to the optimality of the
making use of ant-colony optimization (ACO) corresponding solution candidates; the bit
algorithms. The objective is to minimize the encoding information is also utilized for
completion-time variance of jobs. Two existing pheromone update to avoid inefficient searches.
ant-colony algorithms and the proposed ant- Computer simulation results manifest its
colony algorithm have been compared with an predominant performance in such kind of
existing heuristic for scheduling with the complex optimization problem with hybrid data
objective of minimizing the completion-time structures.
variance of jobs. It is found that the proposed ant-
colony algorithm gives promising and better Samrout et al. (2005) elaborated a new
results, on an average, as compared to those effective technique to assess quasi-optimal
solutions given by the existing ant-colony maintenance periods. Three algorithms have
algorithms and the existing heuristic for the replaced genetic algorithms in their search with

16
ACSE Journal, Volume (6), Issue (1), January, 2006

good results. Subsequently, point out that ant constraints, the solution is to find a MST for each
colony algorithms were very performing cluster, which can be done exactly in polynomial
compared to genetic algorithms at least for the time. Our ACO exploits these two characteristics
studied system under the same conditions. of the CMST by a solution construction originally
Moreover, obtained results encourage us to study developed for the CVRP. Given the CVRP
larger and more complex systems using ACO solution, then apply an implementation of Prim’ s
(other reliability properties, system structures, algorithm to each cluster to obtain a feasible
cost functions). This conclusion is reinforced by CMST solution. Results from a comprehensive
the fact that similar effects have been obtained by computational study indicate the efficiency and
which established a comparison between GA and effectiveness of the proposed approach.
ACO performances.
Zhao et al. (2005) developed the
Shyu et al. (2004) proposed an multiobjective Ant Colony System (ACS) meta-
application of the ACO to a two-machine heuristic to provide a solution for the reliability
flowshop-scheduling problem. In the flowshop, optimization problems of series-parallel system
no intermediate storage is available between two and has been demonstrated its application to the
machines and each operation demands a setup reliability design of gearbox. The problems
time on the machines. The problem seeks to involve the selection of components with multiple
compose a schedule that minimizes the total choice and redundancy levels that produce
completion time. First, present a transformation maximum benefits, and are subject to the cost and
of the scheduling problem into a graph-based weight constraints at the system level. These are
model. An ACO algorithm is then developed with very common and realistic problems involving
several specific features incorporated. A series of conceptual design of engineering system and
computational experiments is conducted by reliability engineering. It is becoming
comparing the proposed algorithm with previous increasingly important to develop efficient
heuristic algorithms. Numerical results evince solutions to these problems because many
that the ACO algorithm exhibits impressive mechanical and electrical systems are becoming
performances with small error ratios. more complex, even as development schedules
Incorporating different styles of convergence, get shorter and reliability requirements become
pheromone updating rules and colony relationship very stringent. The multiobjective Ant Colony
are worthy of further research. Another direction System algorithm offers distinct advantages to
of potential interest may be in the study of non- these problems compared to alternative
graph-based models. optimization methods, and can be applied to a
more diverse problem domain with respect to the
3.4 Subset Problems type or size of the problems. Through the
combination of probabilistic search, multi-
In subset problems, a solution to the problem objective formulation of local moves and the
under consideration is represented as a subset of dynamic penalty method, the multiobjective
the set of available items subject tot problem- ACSRAP, which performs very well on the
specific constraints. Obviously, many problems redundancy apportionment problems (RAP) of the
not considered in this section could be interpreted series-parallel k-out-of-n: G subsystem and
as subset problems. reliability design of gear box, allows us to obtain
an optimal design solution very frequently and
The problem of connecting a set of client more quickly than with other heuristic
nodes with known demands to a root node approaches.
through a minimum cost tree network, subject to Tkindt et al. (2002) considered the 2-
capacity constraints on all links is known as the machine flowshop-scheduling problem with the
capacitated minimum spanning tree (CMST) objective of minimizing both the total completion
problem. As the problem is NP-hard, Reimann time and the make span criteria. The latter is
and Laumanns (2005) proposed a hybrid ant assumed to be optimized prior to the former. In
colony optimization (ACO) algorithm to tackle it view of the NP-hardness of the problem an Ant
heuristically. The algorithm exploits two Colony Optimization approach is proposed to
important problem characteristics: (i) the CMST solve it. The heuristic also uses feature of
problem is closely related to the capacitated Simulated Annealing search and local search
vehicle routing problem (CVRP), and (ii) given a algorithms. Computational experiments show its
clustering of client nodes that satisfies capacity effectiveness compared to existing heuristics. The

17
ACSE Journal, Volume (6), Issue (1), January, 2006

extension to the total completion time problem is


also studied. Vijayakumar et al. (2003) presented a
cutting optimization model for multipass turning
Ying (2004) explored the potential of operations and an ACO based metaheuristic has
ACS heuristic for scheduling problems. Since been applied to solve the machining optimization
ACS is a versatile and robust heuristic, the problem. The results of the proposed approach are
proposed algorithm of this study can be applied to compared with results of simulated annealing and
different scheduling problems and other genetic algorithm. The ACO algorithm can obtain
combinatorial optimization problems with minor a near-optimal solution in an extremely large
modifications. The effectiveness of ACS for solution space within a reasonable computation
solving the n/m/P/Cmax problem is demonstrated. time. The effectiveness of the ACO algorithm has
It clearly suggests such a meta-heuristic to be been proved through an example. The ACO
well worth exploring in the context of solving algorithm developed in this project has some
different scheduling problems. There are several initiative and encouraging features: (a). The
possible extensions to this study in the future proposed ACO-based metaheuristic generates a
research. First, except for the Palmer’s method, superior solution than the simulated annealing
different dispatching rules may be tested. Second, and genetic algorithm based approaches; (b) The
the ACS provides a variety of options and ACO algorithm is completely generalized and
parameter settings that is worth fully examining. problem independent so that it can be easily
Third, the developed ACS algorithm may be modified to optimize this turning operation under
extended to other manufacturing environments various economic criteria, and numerous practical
such as job shop and open shop, which are constraints; and (c) The algorithm can also be
currently investigated by the authors. Finally, one extended to other machining problems, such as
may continue the research in other ACO methods milling operations and threading operations.
for this sequencing problem.
Zheng et al. (2003) shown how to apply
3.5 Machine Learning ant colony behaviour simulation algorithm to
generating the optimal mask used for texture
Many problems in the field of machine learning classification, and present the hybrid-learning
can be cast as optimization problems and be model based on ant colony algorithms, genetic
solved by either classic exact algorithms or by algorithms and simplex algorithm. Experimental
heuristic approached. Examples are the generation results are given to illustrate the capability of
of classification rules, data clustering, regression, proposed method. Proper selection of control
and several others. ACO algorithms have been parameters for an application of ACS is still an
applied to a few such problems so far, opening open issue. In this work we selected parameters
the application of this algorithmic technique to a manually. Further experiments with the adaptive
new range of important problems. parameter settings of the ACS are necessary. In
addition, the ability of proposed approach is
Kuo et al. (2005) proposed a hybrid limited by training samples, and combination with
Case-Based Reasoning (CBR) system with the other texture models and detection methods is not
integration of fuzzy sets theory and Ant System only helpful, but sometimes is also necessary.
based Clustering Algorithm (ASCA) in order to
enhance the accuracy and speed in case matching. 3.6 Network Routing Problems
The cases in the case base are fuzzified in
advance, and then grouped into several clusters by The network routing problem (NRP) is the
their own similarity with fuzzified ASCA. When problem of finding minimum cost paths among all
a new case occurs, the system will find the closest pairs of nodes in the network. It should be noted
group for the new case. Then the new case is that if the costs are fixed, then the NRP is reduced
matched using the fuzzy matching technique only to a set of minimum cost path problems, each or
by cases in the closest group. Through these two which can be solved efficiently via a polynomial
steps, if the number of cases is very large for the time algorithm.
case base, the searching time will be dramatically
saved. In the practical application, there is a Bean and Costa (2005) introduced an
diagnostic system for vehicle maintaining and analytic modeling approach to the study of a
repairing, and the results show a dramatic novel class of adaptive network routing
increase in searching efficiency. algorithm, which is inspired by the emergent

18
ACSE Journal, Volume (6), Issue (1), January, 2006

problem-solving behaviors observed in biological for the detection of microcalcification in digital


ant colonies. This class of algorithm utilizes ‘‘ant- mammogram [Thangavel et al.2005b]. Where, the
mammogram image is segmented using MRF-
like’’ agents that traverse the network and ACSGA method to extract the suspicious region.
collectively construct routing policies. Previous The experimental results shows that the proposed
studies have focused exclusively on simulation ant algorithm performs significantly better than
experiments, which indicate that such algorithms other segmentation techniques and other heuristic
perform well in response to real-time changes in algorithms respectively. This initial study, while
traffic demands and network conditions. The promising, shows its potential application in the
analytic model presented in this paper permits field of image processing.
useful insights into certain fundamental aspects of
ant-based algorithms. In particular, the work 3.7 Other Problems
presented in this paper motivates our proposal of
a number of modifications to the basic design of This section presents some additional applications
ant-based routing algorithms, which result in of ACO that do not fit in the previous sections but
improved performance with respect to which present some particularly interesting
equilibrium performance measures. A useful features not present in other ACO
direction in which to extend this model would be implementations.
to consider more complex traffic models.
Evolutionary computation presents a
Su et al. (2005) introduced an ant colony new paradigm shift in hardware design and
search algorithm (ACSA) to solve the optimal synthesis. According to this paradigm, deriving
network reconfiguration problem for power loss inspiration from biological organisms pursues
reduction. The network reconfiguration problem hardware design. The new paradigm is expected
of one three-feeder distribution system from the to radically change the synthesis procedures in a
literature and one practical distribution network of way that can help discovering novel designs and
Taiwan Power Company (TPC) are, respectively, or more efficient circuits. Abd-El-Barr et al.
solved using the proposed ACSA method, the (2003) presented a multi-objective optimization
genetic algorithm (GA), and the simulated of logic circuits based on a modified Ant Colony
annealing (SA). Numerical results show that the (ACO) algorithm. The performance of the
proposed method is better than the other two proposed algorithm is evaluated using a set of
methods. randomly generated circuits. The results obtained
using the proposed algorithm is compared to
The best way of illustrating how the those obtained using existing ACO-based
ACO metaheuristic operates is by describing how techniques. It is shown that the designed circuits
it has been applied to combinatorial problems. using the proposed algorithm outperform those of
This is done with a full and detailed description of the existing techniques.
most of the current applications of ACO. This
review work is limited to a brief description of the The main characteristic of today’s
main points to consider when applying ACO manufacturing environments is volatility. Under a
algorithms to a few applications of problems volatile environment, demand is not stable. It
representative of important classes of changes from one production period to another.
optimization problems. To operate efficiently under such environments,
the facilities must be adaptive to changing
3.7 Image Processing production requirements. From a layout point of
view, this situation requires the solution of the
Thangavel et al. (2005a) summarized and dynamic layout problem (DLP). DLP is a
compared the methods of automatic detection of computationally complex combinatorial
microcalcifications in digitized mammograms optimization problem for which optimal solutions
used in various stages of the Computer Aided can only be found for small size problems. It is
Detection systems (CAD). It is seen by the survey known that classical optimization procedures are
that the ACS is not used widely to medical image not adequate for this problem. Therefore, several
processing. And the author developed a novel heuristics including taboo search, simulated
technique, Markov Random Field method hybrid annealing and genetic algorithm are applied to
with Ant Colony System, Genetic Algorithm and this problem to find a good solution. Baykasoglu
Backpropagation Network (MRF–ACSGA-BPN) et al. (2004) proposed an ant colony algorithm

19
ACSE Journal, Volume (6), Issue (1), January, 2006

with a simple but effective data structure and the observation of some real ant colony behavior
solution generation mechanism to solve the DLP exploiting the stigmergic communication
by considering the budget constraints. The paradigm. In ant algorithms stigmergic
proposed ant colony algorithm is applied to test communication is implemented by means of a
problems from the available literature that results stigmergic variable which takes different forms in
are known. In these tests the proposed algorithm the different applications: artificial pheromone
found not all of the best solutions but competitive trail in shortest path problems, level of nest
solutions. energy in puck-foraging, level of customer
demand in the mailmen example, puck
The random time changing behavior is distribution in robotic clustering, and the
very popular in speech signal. In order to correct distribution of objects in the lower-dimensional
it the warping method is often used in speech space in exploratory data analysis. Ant algorithms
signal processing which based on template exhibit a number of interesting properties like
matching. Ant colony algorithm is a novel flexibility, robustness, decentralization and self-
random optimized algorithm; it had shown many organization, which make them well suited for the
promising properties in solving complicated solution of problems that are distributed in nature,
optimization problems. Applying the thought of dynamically changing, and require built-in fault-
ant colony algorithm to speech signal processing, tolerance.
Chen and Meng (2004) presented a new dynamic
time programming based on ant colony algorithm. Dreo and Siarry (2004) designed a
It uses both the global and local characters of heterarchical algorithm called “Continuous
speech signal. The theoretic analyses and Interacting Ant Colony” (CIAC) for the
simulation experiments all certify the new optimization of multiminima continuous
algorithm feasibility. The matching results of the functions. CIAC uses two communication
new method can show more accurate similarity channels showing the properties of trail and direct
between speech signals than the Dynamic Time communications. CIAC presents interesting
Warping (DTW) method. emergent properties as it was shown through
some analytical test functions. We believe that
Chen et al. (2004) presented an artificial two characteristics of CIAC are especially
Ants Sleeping Model (ASM) and an ant algorithm interesting. First, the use of the heterarchical
for cluster analysis (A4C), based on the principle concept through the implementation of
of cellular automata. Inspired by the behaviors of communication channels making a link between
gregarious ant colonies, the ant agent to represent search processes. Secondly, the idea that ant
data object is used. In ASM, each ant has two colony algorithms are more efficient on
states: sleeping state and active state. The ant’s dynamical problems, probably even on
state is controlled by a function of the ant’s continuous spaces. These ideas should supply a
fitness to the environment it locates and a framework to achieve simple and efficient
probability for the ants becoming active. The state optimization algorithms in a self-organized and
of an ant is determined only by its local flexible system, as needed in dynamical
information. By moving dynamically, the ants problems. In view of these issues, further work
form different subgroups adaptively, and hence will try to take advantage of the CIAC potential to
the data objects they represent are clustered. design a high-level heterarchical framework
Experimental results show that the A4C algorithm aimed at continuous dynamical optimization.
on ASM is significantly better than other Ge et al. (2004) proposed a hybrid
clustering methods in terms of both speed and optimization technique for global optimization of
quality. It is adaptive, robust and efficient, continuous multi-extreme functions. The scheme
achieving high autonomy, simplicity and incorporates a deterministic searching algorithm,
efficiency. Based on the ASM model, effective Powell method into the ant colony algorithm.
formulae for computing the fitness and activating This approach updates the probability and
probability of agents is proposed. In A4C, the diversity of the ant distribution in the all-
ants group can form into high-quality clusters by continuous space and improves the reliability and
making simple moves according to little local efficiency of the research by using the heuristic
neighborhood information. search method in global area and the direct-search
in local area. Experiment results show the
Dorigo et al. (2000) defined an ant effectiveness and the applicability of the proposed
algorithm to be a multi-agent system inspired by algorithm.

20
ACSE Journal, Volume (6), Issue (1), January, 2006

liquid–liquid extraction processes using stochastic


In order to transform and analyze signals optimization algorithms. The main objective is to
that have been sampled from analogue sources, comparatively study stochastic optimization
digital signal processing (DSP) algorithms are algorithms in view of speed and robustness when
employed. The advantages of DSP are based on searching conventional as well as complex
the fact that the performance of the applied process representations for optimal solutions. The
algorithm is always predictable. There is no work employs a generic representation framework
dependence on the tolerances of electrical that accounts for simple as well as complex
components as in analogue systems. DSP extraction process arrangements. The resulting
algorithms can be reasonably described as a superstructure formulations encompass all the
digital filter. Digital filters can be broadly divided potential configurations and interconnections that
into two-sub classes: finite impulse-response may exist for liquid–liquid extraction systems.
filters and infinite impulse-response (IIR) filters. The task of optimizing the superstructure is
Because the error surface of IIR filters is assigned to stochastic optimization algorithms in
generally multi-modal, global optimization the form of simulated annealing (SA), genetic
techniques are required in order to avoid local algorithms (GA) and ant colony optimization
minima and design efficient digital IIR filters. (ACO). The proposed model is applied to
Karaboga et al. (2004) proposed a new method different separation problems involving
based on the ant colony optimization algorithm quaternary mixtures. The liquid phase non-
with global optimization abilities for digital IIR idealities are modeled both by highly non-linear
filter design. ACO algorithm is a global correlations and standard thermodynamic
optimization technique having certain advantages. methods. All three algorithms have been found to
In this work, firstly a modified version of TACO address the problem with success in terms of the
algorithm, which had been particularly introduced quality of the solution found.
for continuous optimization, has been described.
Secondly, a new method based on the modified C.F. Tsai et al. (2004) proposed a novel
TACO algorithm has been presented for digital algorithm called Ant Colony Optimization with
IIR filter design. Simulation studies show that the Different Favor (ACODF) for data clustering. The
proposed method is accurate and has a fast ACODF algorithm has the following three
convergence rate, and the results obtained important desirable strategies: (a) using ACO
demonstrate that the new method based on TACO with different favor to solve the clustering
can be efficiently used for digital IIR filter design. problem, (b) adopting simulated annealing
concept for ants to decreasingly visit cities to get
Monmarche et al. (2000) presented a local optimal solutions, (c) utilizing tournament
new optimization algorithm based on a model of selection strategy to choose a path. ACODF
the foraging behavior of a population of primitive method is compared with the FSOM+K-means
ants (Pachycondyla apicalis). These ants are approach and GKA. Through experiments, it is
characterized by a relatively simple but efficient shown that ACODF efficiently finds accurate
strategy for prey search in which individuals hunt clusters in large high dimensional datasets.
alone and try to cover a given area around their
nest. The ant colony search behavior consists of a Xiong and Wei (2002) introduced a
set of parallel local searches on hunting sites with coding and selectional operation based on genetic
sensitivity to successful sites. Also, their nest is algorithm through recognizing the standard ant
periodically moved. Accordingly, the proposed colony algorithm, and made better the pheromone
algorithm performs parallel random searches in for general function optimization. The algorithm
the neighborhood of points called hunting sites. was proved efficiently by several function
Hunting sites are created in the neighborhood of a solutions. In a broad sense, all searching or
point called nest. At constant intervals of time the optimization is all ‘equation’ optimization to
nest is moved, which corresponds to a restart objectical equation. The equation referred to here
operator that re-initializes the parallel searches. referred in particular to mathematical equation.
This proposed algorithm is applied, called API, to Traditional optimization for equation, such as,
numerical optimization problems with Newton etc. had relative faster convergence,
encouraging results. higher precision, and widespread application. But
traditional numerical optimization was quite
Papadopoulos and Linke (2004) affected by objectical equation characters, and
addressed the synthesis and optimal design of also, it’s searching started from a beginning, the

21
ACSE Journal, Volume (6), Issue (1), January, 2006

beginning’s good and bad would directly [4] BAUER A., BULLNHEIMER B., HARTL R.F., and
STRAUSS C. Minimizing total tardiness on a single
influence the optimization result. These factor machine using ant colony optimization. Central
would result in the traditional way was uneasy to European Journal for Operations Research and
gain global optimization. Changed the ant colony Economics. 8, 2 (2000), 125-141.
algorithm into directional graph, general equation [5] BAYKASOGLU A., DERELI T., and SABUNCU I. An
ant colony algorithm for solving budget constrained and
optimization also proved effective, for those unconstrained dynamic facility layout problems. Omega.
unsuccessive, undifferential, local crowded (2004). (Article in Press)
extreme value, the experiment made known it had [6] BEAN N., COSTA A. An analytic modeling approach
better ability. for network routing algorithms that use ant-like mobile
agents. Computer Networks. (2005). (Article in Press).
[7] BELL J.E., and MCMULLEN P.R. Ant colony
4. Conclusion optimization techniques for the vehicle routing problem.
Advanced Engineering Informatics. 18 (2004), 41-48.
[8] BLUM C. ACO applied to group shop scheduling: A
Today, several hundred papers have been written case study on intensification and diversification. In M.
on the applications of ACO. It is a true DORIGO, G. DI CARO, M. SAMPLES (Eds.),
metaheuristic, with dozens of application areas. Proceedings of ANTS 2002 – From Ant Colonies to
While both the performance of ACO algorithms Artificial Ants: Third International workshop on Ant
Algorithms, Berlin, Springer-Verlag. 2463 (2002), 14-
and theoretical understanding of their working 27.
have significantly increased, as discussed in the [9] BLUM C. An ant colony optimization algorithm to
previous sections, there are several areas in which tackle shop-scheduling problems. Technical Report
until now only preliminary steps have been taken TR/IRIDIA/2003-1, IRIDIA, University Lebre de
Bruxelles, Brussels. (2003a).
and where much more research will have to be [10] BLUM C., BLESA M.J., Metaheuristics for the edge-
done. weighted k-cardinality tree problem. Technical report
LSI-03-1-R, Department de Llenguatges I Sistemes
One of these research areas in the Inforamtics, Universitat Politecnica de Catalunya,
Barcelona, Spain. 2003b.
extension of ACO algorithms to more complex [11] BONABEAU E., HENAUX F., GUERIN S., SNYERS
optimization problems that include (1) dynamic D., KUNTZ P., and THERAULAZ G. Routing in
problems, in which the instance data, such as telecommunication networks with “smart” ant-like
objective function values, decision parameters, or agents. In S. ALBAYRAK & F. GARIJO (Eds.),
constraints, may change while solving the Proceedings of IATA’98, Second International Workshop
on Intelligent Agents for Telecommunication
problem; (2) stochastic problems, in which one Applications, Berlin, Springer-Verlag. 1437 (1998), 60-
has only probabilistic information about objective 72.
function values, decision variables values, or [12] BULLNHEIMER B., HARTL R.F., and STRAUSS C.
Applying the Ant System to the vehicle routing problem.
constraint boundaries, due to uncertainty, noise, In S. VOSS, S. MARTELLO, IH. OSMAN, C.
approximation, or other factors; and (3) multiple ROUCAIROL (Eds.), Meta-Heuristics: Advances and
objective problems, in which a multiple objective Trends in Local Search Paradigms for Optimization,
function evaluates competing criteria of solution Dordrecht, Nethrelands, Kluwer Academic Publishers.
(1999a) 285-296.
quality. [13] BULLNHEIMER B., HARTL R.F., and STRAUSS C.
Active research directions in ACO An improved ant system algorithm for the vehicle
include also the effective parallelization of ACO routing problem. Annals of Operations Research. 89
algorithms and, on a more theoretical level, the (1999b) 319-328.
[14] BULLNHEIMER B., HARTL R.F., and STRAUSS C. A
understanding and characterization of the new rank-based version of the Ant System: A
behavior of the ACO algorithms while solving a computational study. Central European Journal for
problem. Operations Research and Economics. 7, 1 (1999c), 25-
38.
[15] CASILLAS J., CORDON O., and HERRERA F.
References Learning cooperative fuzzy linguistic rules using ant
colony algorithms. Technical report DECSAI-00-01-19,
Department of Computer Science and Artificial
[1] ABD-EL-BARR M., SAIT SM., SARIF B., and AI-
Intelligence, University of Granada, Granada, Spain.
SAIARI U. A Modified Ant Colony Algorithm for
(2000).
Evolutionary Design of Digital Circuits. IEEE. (2003),
[16] CHEN H.H., and MENG Q.C., Dynamic Time
708-715.
Programming Based On Ant Colony Algorithm.
[2] ANNALURU R., DAS S., and PAHWA A. Multi-Level
Proceedings of the Third International Conference on
Ant Colony Algorithm for Optimal Placement of
Machine Learning and Cybernetics, Shanghai. (August
Capacitors in Distribution Systems, IEEE. (2004), 1932-
2004), 3557-3562.
1937.
[17] CHEN L., XU X., CHEN Y., and HE P. A Novel Ant
[3] AZIMI Z.N. Hybrid heuristics for Examination
Clustering Algorithm Based on Cellular Automata.
Timetabling problem. App. Mathematics and
Proceedings of the IEEE/WIC/ACM International
Computation. 163 (2005), 705-733.

22
ACSE Journal, Volume (6), Issue (1), January, 2006

Conference on Intelligent Agent Technology (IAT’04). [32] DORIGO M., and DI CARO G. GAMBARDELLA,
(2004). L.M. Ant algorithms for discrete optimization. Artificial
[18] COLORNI A., DORIGO M., MANIEZZO V., and Life. 5 (1999b), 137–172.
TRUBIAN M. Ant System for job-shop scheduling. [33] DORIGO M., DI CARO G., and GAMBARDELLA
JORBEL Belgian Journal of Operations Research, L.M. Ant algorithms for distributed discrete
Statistics and Computer Science. 34, 1 (1994), 39-53. optimization. Artificial Life. 5 (1999), 137–172.
[19] COLORNI A., DORIGO M., and MANIEZZO V., In [34] DORIGO M., and GAMBARDELLA L.M. Ant Colonies
F.J. VARELA, P. BOURGINE. Distributed optimization for the traveling salesman problem. BioSystems. 43, 2
by ant colonies. Proceedings of the First European (1997), 73-81.
Conference on Artificial Life, Paris. (1991). [35] DORIGO M., and GAMBARDELLA L.M. Ant Colony
[20] CORDON 0., DE VIANA I.F., HERRERA F., and System: A cooperative learning approach to the traveling
MORENO L. A new ACO model integrating salesman problem. IEEE Transactions on Evolutionary
evolutionary computation concepts: The best-worst Ant Computation. 1, 1 (1997), 53-66.
System. In M. DORIGO, M. MIDDENDORF, T. [36] DORIGO M., MANIEZZO V., and COLORNI A.
STUTZLE (Eds.). Abstract Proceedings of ANTS 2000- Positive feed back as a search strategy. Technical Report
From Ant Colonies to Artificial Ants: Second Intl. (16), Politecnico di Milano, Italy. (1991).
Workshop on Ant Algorithms, Brussels, IRIDIA, [37] DORIGO M., MANIEZZO V., and COLORNI A. The
Universite Libre de Bruxelles. (2000), 22-29. Ant System: An autocatalytic optimizing process.
[21] CORDONE R., and MAFFIOLI F. Coloured Ant System Technical report 91-016 revised, Diparmento di
ant local search to design local telecommunication Elettronica, Politecnico di Milano, Milan. (1991).
networks. In E.J.W. BOERS, J. GOTTLIEB, P.L. [38] DORIGO M., MANIEZZO V., and COLORNI A. Ant
LANZI, R.E. SMITH, S. CARGNONI, E. HART, G.R. System: Optimization by a colony of cooperating agents.
RAIDL, H. TIHINK (Eds.). Applications of Evolutionary IEEE Transactions on Systems, Man, and Cybernetics-
Computing: Proceedings of Evo Workshops 2001, Part B, 26, 1 (1996) 29-41.
Berlin, Springer-Verlag. 2037 (2001), 60-69. [39] DREO J, and SIARRY P, Continuous interacting ant
[22] COSTA D., and HERTZ A. Ants can color graphs. Jour. colony algorithm based on dense heterarchy, Future
of the Operational Research Society, 48 (1997), 295- Generation Computer Systems. 20 (2004), 841-856.
305. [40] EGGERS J., FEILLET D., KEHL S., WAGNER M.O.,
[23] DE CAMPOS L.M., GAMEZ J.A. and PUERTA J.M. and YANNOU B. Optimization of the keyboard
Learning Bayesian networks by ant colony optimization: arrangement problem using an Ant Colony algorithm,
Searching in the space of orderings. Mathware and Soft European Jour. of Operational Research, 148 (2003),
Computing. 9, 2-3 (2002), 251-268. 672-686.
[24] DEN BESTEN M.L., STUTZLE T. and DORIGO M. [41] FENET S., and SOLNON C. Searching for maximum
Ant Colony optimization for the total weighted tardiness cliques with ant colony optimization. In G.R. RAIDL,
problem. In M. SCHOENAUER, K. DEB, G. J.A. MAYER, M. MIDDENDORF, S. CAGNONI, J.J.R.
RUDOLPH, X. YAO, E. LUTTON, J.J. MERELO, H.P. CARDALDA, D.E. CORNE, J. GOTTLIEB, A.
SCHWEFEL (Eds.), Proceedings of PPSN-VI, Sixth GUILLOT, E. HART, C.G. JOHNSON, E.
International Conference on Parallel Problem Solving MARCHIORI (Eds.), Applications of Evolutionary
from Nature, Berlin, Springer-Verlag. 1917 (2000), 611- Computing, Proceedings of Evo Workshops 2003,
620. Berlin, Springer-Verlag. 2611 (2003), 236-245.
[25] DI CARO G., and DORIGO M. AntNet: A mobile [42] GAGNE C., PRICE W.L., and GRAVEL M. Comparing
agents approach to adaptive routing. Technical report an ACO algorithm with other heuristics for the single
IRIDIA/97-12, IRIDIA, Universite Libre de Bruxelles, machine-scheduling problem with sequence-dependent
Brissels. (1997). setup times, Journal of the Operational Research
[26] DI CARO G., and DORIGO M. AntNet: Distributed Society. 53 (2002), 895-906.
stigmergetic control for communications networks. [43] GAJPAL Y., and RAJENDRAN C, An ant-colony
Journal of Artificial Intelligence Research. 9 (1998a), optimization algorithm for minimizing the completion-
317-365. time variance of jobs in flowshops. Int. J. Production
[27] DI CARO G., and DORIGO M., Extending AntNet for Economics. (2005) (Article in Press).
best-effort quality-of-service routing, Unpublished [44] GAMBARDELLA L.M., and DORIGO M. Ant-Q: A
presentation at ANTS’98-From Ant Colonies to reinforcement learning approach to the traveling
Artificial Ants: First International Workshop on Ant salesman problem. In A. PRIEDITIS, S. RUSSELL
Colony Optimization, Brussels. (1998b) (Eds.), Proceedings of the Twelfth International
[28] DI CARO G., and DORIGO M. Two ant colony Conference on Machine Learning (ML-95), Palo Alto,
algorithms for best-effort routing in datagram networks. CA, Morgan Kaufmann. (1995), 252-260.
In Y. PAN, SG. AKL, K. LI (Eds.), Proceedings of the [45] GAMBARDELLA L.M., and DORIGO M. HAS-SOP:
Tenth IASTED International Conference on Parallel and A hybrid Ant System for the sequential ordering
Distributed Computing and Systems (PDCS’98), problem. Technical report IDSIA-11-97, IDSIA, Lugano,
Anaheim, CA, IASTED/ACTA Press. (1998c), 541-546. Switzerland. (1997).
[29] DORIGO M. Optimization, Learning and Natural [46] GAMBARDELLA L.M., and DORIGO M. Ant Colony
Algorithms (in Italian). PhD Thesis, Diparmento di System hybridized with a new local search for the
Elettronica, Politecnico di Milano, Milan. (1992). sequential ordering problem. INFORMS Journal on
[30] DORIGO M., BONABEAU E., and GUY Computing. 12, 3 (2000) 237-255.
THERAULAZ. Ant algorithms and stigmergy, Future [47] GAMBARDELLA L.M., TAILLARD E.D., and
Generation Computer Systems.16 (2000), 851-871. AGAZZI G. MACS-VRPTW: A multiple ant colony
[31] DORIGO M., and DI CARO G. The ant colony system for vehicle routing problems with time windows,
optimization meta-heuristic, In: D. CORNE, M. In D. CORNE, M. DORIGO, F. GLOVER (Eds.), New
DORIGO, F. GLOVER (Eds.), New Ideas in Ideas in Optimization, London, McGraw Hill. (1999),
Optimization, McGraw-Hill, London, (1999a), 11–32. 63-76.

23
ACSE Journal, Volume (6), Issue (1), January, 2006

[48] GAMBARDELLA L.M., TAILLARD E.D., and X. YAO, A. ZALZALA (Eds.), Proceedings of the 1999
DORIGO M. Ant Colonies for the quadratic assignment Congress on Evolutionary Computation (CEC’99),
problem. Journal of the Operational Research Society. Piscataway, NJ, IEEE Press. (1999), 1459-1464.
50, 2 (1999), 167-176. [67] LEGUIZAMON G., and MICHALEWICZ Z. Ant
[49] GAMEZ J.A., and PUERTA J.M. Searching for the best Systems for subset problems. Unpublished manuscript.
elimination sequence in Bayesian networks by using ant (2000).
colony optimization. Pattern Recognition Letters. 23 [68] LEVINE J., and DUCATELLE F. Ant colony
(2002), 261-277. optimization and local search for bin packing and cutting
[50] GE Y., MENG Q.C., YAN C.J., and XU J. A Hybrid Ant stock problems. Journal of the Operational Research
Colony Algorithm for Global Optimization of Society, to appear. (2003).
Continuous Multi-Extreme Functions. Proceedings of [69] LI Y., and WU T.J. A Nested Ant Colony Algorithm for
the Third International Conference on Machine Hybrid Production Scheduling. Proceedings of the
Learning and Cybernetics, Shanghai. (August 2004), American Control Conference Anchorage, AK. (May
2427-2432. 2002), 1123 - 1128.
[51] GLOVER F. Tabu search: a tutorial. Interfaces, 20 [70] LI Y., WU T.J., and HILL D.J. An Accelerated Ant
(1990), 74–94. Colony Algorithm for Complex Nonlinear System
[52] GOLDBERG D.E. Genetic Algorithms in Search, Optimization. Proceedings of the 2003 IEEE
Optimization, and Learning, NY, Addison Wesley. International Symposium on intelligent Control Houston,
(1989). Texas. (October 2003), 709-713.
[53] GOSAVI S., DAS S., VAZE S., SINGH G., and [71] LIANG Y.C., and SMITH A.E. An Ant System
RUEHLER E. Obtaining Subtrees from Graphs: An Ant approach to redundancy allocation. In P.J. ANGELINE,
Colony Approach, IEEE. (2003), 160-166. Z. MICHALEWICZ, M. SCHOENAUER, X. YAO, A.
[54] GUTJAHR W.J. ACO algorithms with guaranteed ZALZALA (Eds.), Proceedings of the 1999 Congress on
convergence to the optimal solution. Information Evolutionary Computation (CEC’99), Piscataway, NJ,
Processing Letters. 82 (2002), 145-153. IEEE Press. (1999), 1478-1484.
[55] HADJI R., RAHOUAL M., TALBI E., and BACHELET [72] LIM A., LIN J., RODRIGUES B., and XIAO F. Ant
V. Ant colonies for the set covering problem. In M. colony optimization with hill climbing for the bandwidth
DORIGO, M. MIDDENDORF, T. STUTZLE (Eds.), minimization problem. Applied Soft Computing. (2005).
Abstract Proceedings of ANTS 2000-From Ant Colonies (Article in Press)
to Artificial Ants: Second International Workshop on Ant [73] LOURENCO H., and SERRA D. Adaptive approach
Algorithms, Brussels, Universite Libre de Bruxelles. heuristics for the generalized assignment problem.
(2000). Universitat Popeu Fabra, Department of Economics and
[56] HEUSSE M., SNYERS D., GUERIN S., and KUNTZ P. Management, Barcelona, Spain. Technical report No.
Adaptive agent-driven routing and load balancing in 304, (1998).
communication networks. Advances in Complex [74] LOURENCO H., and SERRA D. Adaptive search
Systems, 1, 2 (1998), 237-254. heuristics for the generalized assignment problem.
[57] HOLLAND J.H. Adaptation in Natural and Artificial Mathware and Soft Computing. 9, 2-3 (2002), 209-234.
Systems. Ann Arbor: The University of Michigan. [75] MANIEZZO V. Exact and approximate non-
(1975). deterministic tree-search procedures for the quadratic
[58] HOSHYAR R., JAMALI SH, and LOCUS C. Ant assignment problem. INFORMS Journal on Computing.
colony algorithm for finding good interleaving pattern in 11, 4 (1999), 358-369.
turbo codes, IEE PIWC –Commun. 147, 5 (2000). [76] MANIEZZO V., and CARBONARO A. An ANTS
[59] JAYARAMAN V.K., KULKARNI B.D., KARALE S., heuristic for the frequency assignment problem. Future
and SHELOKAR P, Ant colony framework for optimal Generation Computer Systems. 16, 8 (2000), 927-935.
design and scheduling of batch plants. Computers and [77] MANIEZZO V., and COLORNI A. The Ant System
Chemical Engineering. 24 (2000), 24: 1901-1912. applied to the quadratic assignment problem. IEEE
[60] KARABOGA N., KALINLI A., and KARABOGA D. Transactions on Data and Knowledge Engineering. 11, 5
Designing digital IIR filters using ant colony (1999), 769-778.
optimization algorithm. Engineering Applications of [78] MANIEZZO V., COLORNI A., and DORIGO M. The
Artificial Intelligence. 17 (2004), 301-309. Ant System applied to the quadratic assignment problem.
[61] KIRKPATRICK S., GELATT D., and VECCHI M.P. IRIDIA, Universite Libre de Bruxelles, Brussels.
Optimization by simulated annealing. 220 (1983), 671– Technical report IRIDIA/94-28, (1994).
680. [79] MAZZEO S., and LOISEAU I. An Ant Colony
[62] KOROSEC P., SILC J. and ROBIC B. Solving the Algorithm for the Capacitated Vehicle Routing.
mesh-partitioning problem with an ant-colony algorithm. Electronic Notes in Discrete Mathematics. 18 (2004),
Parallel Computing. 30 (2004), 785-801. 181-186.
[63] KUO R.J., KUO Y.P., and CHEN K.Y. Developing a [80] MERKLE D., and MIDDENDORF M. An ant algorithm
diagnostic system through integration of fuzzy case- with a new pheromone evaluation rule for total tardiness
based reasoning and fuzzy ant colony system. Expert problems. In S. CAGONI, R. POLI, G.D. SMITH, D.
Systems with Applications. 28 (2005), 783-797. CORNE, M. OATES, E. HART, P.L. LANZI, E.J.
[64] LEE Z.J., and LEE C.Y. A hybrid search algorithm with WILLIAM, Y. LI, B. PAECHTER, T.C. FOGARTY
heuristics for resource allocation problem. Information (Eds.), Real-World Applications of Evolutionary
Sciences. (2004). Computing, Berlin, Springer-Verlag. 1803 (2000), 287-
[65] LEE Z.J., and LEE C.Y., SU S.F. An immunity-based 296.
ant colony optimization algorithm for solving weapon [81] MERKLE D., and MIDDENDORF M. Ant colony
target assignment problem. Applied Soft Computing. 2 optimization with global pheromone evaluation for
(2002), 39-47. scheduling a single machine. Applied Intelligence. 18, 1
[66] LEGUIZAMON G., and MICHALEWICZ Z. A new (2003), 105-111.
version of Ant System for subset problems. In PJ. [82] MERKLE D., and MIDDENDORF M., SCHMECK H.
ANGELINE, Z. MICHALEWICZ, M.SCHOENAUER, Ant colony optimization for resource-constrained project

24
ACSE Journal, Volume (6), Issue (1), January, 2006

scheduling. In D. WHITLEY, E. CANTU-PAZ, L. [97] SHYU S.J., LIN B.M.T., and HSIAO T.S. Ant colony
SPECTOR, I. PARMEE, H.G. BEYER (Eds.), optimization for the cell assignment problem in PCS
Proceedings of the Genetic and Evolutionary networks. Computers & Operations Research. (2005).
Computation Conference (GECCO-2000), San (Article in Press)
Francisco, Morgan Kaufmann. (2000), 893-900. [98] SHYU S.J., LIN B.M.T., and YIN P.Y. Application of
[83] MERKLE D., and MIDDENDORF M., SCHMECK H. ant colony optimization for no-wait flowshop scheduling
Ant colony optimization for resource-constrained project problem to minimize the total completion time.
scheduling. IEEE Transactions on Evolutionary Computers & Industrial Eng. 47 (2004), 181-193.
Computation. 6, 4 (2002), 333-346. [99] SOCHA K., KNOWLES J., and SAMPLES M. A Max-
[84] MICHEL R., and MIDDENDORF M. An island model Min Ant System for the university course timetabling
based Ant System with look-ahead for the shortest problem. In M. DORIGO, G. DI CARO, M. SAMPLES
supersequence problem. In A.E. EIBEN, T. BACK, M. (Eds.), ANTS 2002-From Ant Colonies to Artificial Ants:
SCHOENAUER, H.P. SCHWEFEL (Eds.), Proceedings Third International Workshop on Ant Algorithms, Berlin,
of PPSN-V, Fifth International Conference on Parallel Springer-Verlag. 2463 (2002), 1-13.
Problem Solving from Nature, Berlin, Springer-Verlag. [100] SOCHA K., SAMPLES M., and MANFRIN M. Ant
1498 (1998), 692-701. algorithms for the university course timetabling problem
[85] MICHEL R., and MIDDENDORF M. An ACO with regard to the state-of-the-art. In G.R. RAIDL, JA.
algorithm for the shortest supersequence problem, In D. MEYER, M. MIDDENDORF, S. CAGONI, J.J.R.
CORNE, M. DORIGO, F. GLOVER (Eds.), New Ideas CARDALDA, D.W. CORNE, J. GOTTLIEB, A.
in Optimization, London, McGraw Hill. (1999), 51-61. GUILLOT, E. HART, C.G. JOHNSON, E.
[86] MONMARCHE N., VENTURINI G., and SLIMANE MARCHIORI (Eds.), Appliations of Evolutionary
M. On how Pachycondyla apicalis ants suggest a new Computing, Proceedings of Evo Workshops 2003,
search algorithm. Future Generation Computer Systems. Berlin, Springer-Verlag, 2611 (2003), 334-345.
16, (2000), 937-946. [101] SOLIMANPUR M., VRAT P., and SHANKAR R. Ant
[87] PAPADOPOULOS A.I., and LINKE P. On the synthesis colony optimization algorithm to the inter-cell layout
and optimization of liquid-liquid extraction processes problem in cellular manufacturing. European Journal of
using stochastic search methods. Computers and Operational Research. 157 (2004), 592-606.
Chemical Engineering. 28 (2004), 2391-2406. [102] SOLIMANPUR M., VRAT P., and SHANKAR R. An
[88] PARPINELLI R.S., LOPES H.S., and FREITAS A.A. ant algorithm for the single row layout problem in
Data mining with an ant colony optimization algorithm. flexible manufacturing systems. Computers &
IEEE Transactions on Evolutionary Computation. 6, 4 Operations Research. 32 (2005), 583-598.
(2002), 321-332. [103] SOLNON C. Solving permutation constraint satisfaction
[89] PFAHRINGER B., Multi-agent search for open shop problems with artificial ants. In W. HORN (ed.),
scheduling: Adapting the Ant-Q formalism. Austrian Proceedings of the 14th European Conference on
Research Insititute for Artificial Intelligence, Vienna. Artificial Intelligence, Armsterdam, IOS Press. (2000),
Technical report TR-96-09, (1996). 118-122.
[90] RANDALL M. A Parallel Implementation of Ant [104] SOLNON C. Ants can solve constraint satisfaction
Colony Optimization. Journal of Parallel and problems. IEEE Trans. on Evolutionary Computation. 6,
Distributed Computing. 62, (2002), 1421-1432. 4 (2002), 347-357.
[91] REIMANN M., and LAUMANNS M. Savings based ant [105] SONG Y.H., CHOU C.S., and STONHAM T.J.
colony optimization for the capacitated minimum Combined heat and power economic dispatch by
spanning tree problem. Computers & Operations improved ant colony search algorithm. Electric Power
Research. (2005). (Article in Press) Systems Research. 52, (1999), 115-121.
[92] REIMANN M., STUMMER M., and DOERNER K. A [106] STUTZLE T. Max-Min Ant System for the quadratic
savings based Ant System for the vehicle routing assignment problem. FG Informatik, TU Darmstadt,
problem. In W.B. LANGDON, E. CANTU-PAZ, K. Germany, 1997.
MATHIAN, R. ROY, D. DAVIS, R. POLI, K. [107] STUTZLE T. An ant approach to the flow shop problem.
BALAKRISHNAN, V. HONAVAR, G. RUDOLPH, H. In Proceedings of the Sixth European Congress on
WEGENER, L. BULL, MA. POTTER, AC. SCHULTZ, Intelligent Techniques & Soft Computing (EUFIT’98),
JF. MILLER, E. BURKE, N. JONOSKA (Eds.), Aachen, Germany, Verlag Mainz, Wissenschaftsverlag, 3
Proceedings of the Genetic and Evolutionary (1998), 1560-1564.
Computation Conference (GECCO-2002), San [108] STUTZLE T., and HOOS H.H. The Max-Min Ant
Francisco, Morgan Kaufmann. (2002), 1317-1325. System and local search for the traveling salesman
[93] SAMROUT M., YALAOUI F., CHATELET E., and problem. In T. BACK, Z. MICHALEWICZ, X. YAO
CHEBBO N. New methods to minimize the preventive (Eds.), PROCEEDINGS OF THE 1997 IEEE
maintenance cost of series-parallel systems using ant INTERNATIONAL CONFERENCE ON
colony optimization. Reliability Engineering and System EVOLUTIONARY COMPUTATION (ICEC’97),
Safety. 89 (2005), 346-354. PISCATAWAY, NJ, IEEE PRESS. (1997), 309-314.
[94] SCHOONDERWOERD R., HOLLAND O., and [109] STUTZLE T., and HOOS H.H. Max-Min Ant System.
BRUTEN J. Ant based load balancing in Future Generation Computer Systems. 16, 8 (2000), 889-
telecommunications networks. Adaptive Behavior. 5, 2 914.
(1997), 169-207. [110] SU C.T., CHANG C.F., and CHIOU J.P. Distribution
[95] SENDOVA-FRANKS A.B., and LENTZ J. Random network reconfiguration for loss reduction by ant colony
Walk Models of Worker Sorting in Ant Colonies. J. search algorithm. Electric Power Systems Research.
theor. Biol. 217 (2002), 255-274. (2005). (Article in Press)
[96] SHMYGELSKA A., AGUIRE-HERNANDEZ R., and [111] SUBRAMANIAN D., DRUSCHEL P., and CHEN J.
HOOS H.H. An ant colony optimization algorithm for Ants and reinforcement learning: A case study in routing
the 2D HP protein-folding problem. In M. DORIGO, G. in dynamic networks. In Proceedings of the 15th Intl.
DI CARO, M. SAMPLES (Eds.), ANTS 2002, Berlin, Joint Conference on Artificial Intelligence, San
Springer-Verlag. 2463 (2002), 40-52. Francisco, Morgan Kaufmann. (1997), 832-838.

25
ACSE Journal, Volume (6), Issue (1), January, 2006

[112] TAILLARD E.D., GAMBARDELLA L.M., [128] ZHENG H., WONG A., and Nahavandi S., Hybrid Ant
GENDREAU M., and POTVIN J.Y. Adaptive memory Colony Algorithm for Texture Classification. IEEE.
programming: A unified view of metaheuristics. (2003), 2648-2652.
European Journal of Operational Research. 135, (2001),
135: 1-16. Corresponding author: Prof. M.KARNAN
[113] TALBI E.G., ROUX O., FONLUPT C., and
ROBILLARD D. Parallel Ant Colonies for the quadratic 153,Palayam,Palani,Tamilnadu,India
assignment problem. Future Generation Computer
Systems. 17 (2001), 441-449.
[114] THANGAVEL K., KARNAN M., SIVA KUMAR R.,
and KAJA MOHIDEEN A. Automatic Detection of
Microcalcification in Mammograms– A Review.
International Journal on Graphics Vision and Image
Processing. 5 (2005a) 31-61.
[115] THANGAVEL K., KARNAN M., SIVA KUMAR R.,
and KAJA MOHIDEEN A. Ant Colony System for
Segmentation and Classification of Microcalcification in
Mammograms. International Journal on Artificial
Intelligence and Machine Learning. 8 (2005b).
[116] TKINDT V., MONMARCHE N., TERCINET F., and
LAUUGT D. An Ant Colony Optimization algorithm to
solve a 2-Machine bi-criteria flowshop-scheduling
problem. European Journal of Operational Research.
142 (2002), 250-257.
[117] TSAI C.F., TSAI C.W., and TSENG C.C. A new hybrid
heuristic approach for solving large traveling salesman
problem. Information Sciences. 164 (2004), 67-81.
[118] TSAI C.F., TSAI C.W., WU H.C., AND YANG T.
ACODF: A novel data clustering approach for data
mining in large databases. The Journal of Systems and
Software. 73, (2004), 133-145.
[119] VAN DER PUT R. Routing in the faxfactory using
mobile agents. Technical Report R&D-SV-98-276,
(1998).
[120] VARELA G.N., and SINCLAIR M.C. Ant colony
optimization for virtual-wavelength-path routing and
wavelength allocation. In P.J. ANGELINE, Z.
MICHALEWICZ, M. SCHOENAUER, X. YAO, A.
ZALZALA (Eds.), Proceedings of the 1999 Congress on
Evolutionary Computation (CEC’99), Piscataway, NJ,
IEEE Press, (1999), 1809-1816.
[121] VIJAYAKUMAR K., PRABHAHARAN G., ASOKAN
P., and SARAVANAN R. Optimization of multi-pass
turning operations using ant colony system.
International Journal of Machine Tools & Manufacture.
43, (2003), 1633-1639.
[122] WHITE T., PAGUREK B., and OPPACHER F.
Connection management using adaptive mobile agents,
In H.R. ARABNIA (Ed.), Proceedings of the
International Conference on Parallel and Distributed
Processing Techniques and Applications (PDPTA’98),
Las Vegas, NV, CSREA Press. (1998), 802-809.
[123] XIA L., XUEHUI L., and JIHONG Z. Codebook Design
By A Hybridization Of Ant Colony With Improved LBG
Algorithm. IEEE Int. Conf. Neural Networks & Signal
Processing, Nanjing, China. (2003), 14-17.
[124] XIONG W.Q., and WEI P. A Kind of Ant Colony
Algorithm for Function Optimization. Proceedings of the
First International Conference on Machine Learning
and Cybernetics, Beijing. (2002), 4-5.
[125] YIN P.Y. Ant colony search algorithms for optimal
polygonal approximation of plane curves. Pattern
Recognition. 36 (2003), 1783 - 1797.
[126] YING K.C., and LIAO C.J. An ant colony system for
permutation flow-shop sequencing. Computers &
Operations Research. 31 (2004), 791-801.
[127] ZHAO J.H., LIU Z., and DAO M.T., Reliability
optimization using multi-objective ant colony system
approaches. Reliability Engineering and System Safety.
(2005), 1-11.

26

Вам также может понравиться