© 2016 IJSRSET | Volume 2 | Issue 1 | Print ISSN: 2395-1990 | Online ISSN : 2394-4099

Themed Section: Engineering and Technology

Exploring Genetic Algorithms to Solve QoS

Dr. Gaytri Devi, Rajesh Dawar
GVM Institute of Technology and Management, DCRUST , Sonipat

ABSTRACT
Today, there has been an increasing demand for real-time multimedia applications on Internet that can provide
transmission of audio, video and tele-conferencing. Many of these applications have multiple quality-of-service
(QoS) guarantees in terms of delay, bandwidth, transmission success ratio etc. To fulfil this demand, QoS-based
network framework has been adopted. One of the key issues in the QoS framework is how to determine a feasible
path that satisfies QoS requirements. In general, the problem of finding a path satisfying multiple constraints is NP-
complete and finding an exact solution can be difficult. Various heuristics and approximation algorithms have been
proposed in literature to solve this problem . However , the QoS path calculated by these algorithms may turn out to
be invalid over the time because of dynamic nature of network conditions. Using Genetic algorithm to solve Multi
constrained QoS problem, may be proved a better solution. Genetic approach makes the algorithm more robust in
the case that the frequency of state information change in the network is higher than the rate of state information
received at the node. In order to solve the QoS constrained routing efficiently, the scheme of routing based on a
genetic algorithm (GA) has been discussed in this paper.
Keywords:- QoS, Genetic algorithm, Chromosome, Routing.

I. INTRODUCTION non-QoS routing in terms of the accuracy of the global

state. Inaccuracy can lead QoS to failure.
There is a need to develop routing algorithm which
consider more than one QoS parameter such as One of the solution may be using Genetic algorithm .
bandwidth, delay, cost and loss probability while The main advantage of using genetic algorithms is that
finding the path to support high-speed and real-time the Genetic algorithm is insensitive to variations in
services .The basic problem of QoS routing is to find a network topologies with respect to route optimality and
path satisfying multiple constraints. It is concerned with convergence speed. GA is a kind of parallel
identifying the path that will consider multiple optimization algorithm that simulates the evolution
parameters like bandwidth, delay, cost, hop count etc. process of a creature, and is suitable to find optimal
instead of one. solution in a large and complicated search space.
Genetic algorithm is stochastic, not deterministic. It can
The problem of QoS routing to find the path satisfying automatically obtain and accumulate the knowledge
multiple constraints is intractable. Various heuristics about the research space, so as to control the search
and approximation algorithms have been proposed in process adaptively, until the optimal solution is finally
literature to solve QoS problem .However, in found [1]. GA does not need the solutions of the
implementation of multimedia network , the status problem to be continuous so that it can work well for
information available for making routing decision the QoS routing problem. Further, it is quite simple and
changes dynamically . The heuristic and approximate robust, and very easy to be implemented . A genetic
routing algorithm does not provide adequate algorithm works on a population of possible solutions,
performance with imprecise state information. The while other heuristic methods use a single solution in
effectiveness and performance of the routing algorithms their iterations. Thus GA is more effective and efficient
can significantly be degraded by this outdated than conventional heuristic and approximate algorithms
information. QoS routing is much more sensitive than in multimedia networks.

IJSRSET173390 | 11 Jan 2016 | Accepted: 30 Jan 2016 | January-February-2016 [(2)1: 638- 643] 638
 This paper explores the approach of solving QoS is called feasible. On the other hand optimization refers
problem based on genetic algorithm. The layout of the to path having minimum or maximum value for a
paper is as follows-Section 2 gives brief description of constraint e.g. Choose the path that has minimum
QoS . Section 3 presents the process of genetic delay among all the paths. This path is called optimal
algorithm. Section 4 discuss the genetic algorithm for path.
routing problem and also presents the related work of
genetic algorithm in context of QoS . Section 5 The further QoS issues have been discussed in[2][3].
concludes the paper.
III. Genetic Algorithm
II. Quality of Service
Genetic Algorithm (GA) is an optimization algorithm
The fundamental problem of routing in a network that that is based on natural selection, inspired by the theory
provides QoS guarantee is to find a path between of genetics.
specified source and destination node pair that
simultaneously satisfies multiple QoS parameters. It starts with an initial set of random solutions called
population. Population evolve through successive
Quality of Service(QoS) puts some restrictions in the iterations to obtain new and improved generation.
form of certain constraints on the path. These During each iteration of the algorithm, each solution in
constraints may be desired bandwidth, delay, variation the population are tested to see whether they give a
in delay experienced by receiver(jitter),packet loss that valid solution. This testing operation is nothing but the
can be tolerated, number of hops, cost of links etc. fitness functions to achieve the objective. A new
generation is formed by selection according to the
These parameters are represented in the form of metrics. fitness values. The generation undergo one or more
One metric for each constraint is to be specified like genetic operations such as crossover and mutation.
bandwidth metric, jitter (variation in delay) metric, Again fitness function is applied to test its fitness. This
delay metric, number of hops metric, packet loss ratio process continues until either the solution is found or a
etc. from one node to all other nodes in the network. certain termination condition is met.
Metric for a complete path with respect to each
parameter is determined by the composition rules of The general process of GA as described by Goldberg [1]
metrics. The three basic rules are [15].- is as follows-
(i) Additive Metric: The value of the constraint over
the entire path is the addition of all links constituting 1. Initialization of initial population- The initial
path. For Example- delay, hop count, cost or jitter. population is normally created randomly or based on
ii) Multiplicative Metric: Using this metric, the value some heuristics which is assumed to give solutions.
for the complete path is multiplication of metric value Each individual solution in the population is called a
of all its edges. chromosome. A chromosome is encoded form of
Examples are – reliability (1-lossratio) and error free solution that consists of several genes.
Transmission (probability) 2. Evaluation of population-Evaluate the fitness of
Multiplicative metric can be converted into additive by each chromosome on the population. The fitness is
taking logarithm. usually the value of the objective function in the
iii) Concave Metric: In this metric, either min edge optimization problem being solved.
value or max edge value is taken as constraint value for 3. Reproduction and Selection - In this process,
a path among all the edges of that path. For Example- chromosomes are selected to be put into new generation.
Bandwidth A selection scheme is utilized to choose the
chromosomes. The selection scheme could be
For a complete path, the constraints may be required proportionate selection, ranking selection, tournament
either as a constrained form or in a optimization form. selection etc, According to their fitness values, some of
In constrained form, some condition is put on constraint chromosomes are reproduced and some are eliminated
value e.g. Choose that path only which has delay less in such a way that Chromosomes with higher quality i.e.
than or equal to 60 ms. The path obeying the condition higher fitness value have higher probabilities of being

International Journal of Scientific Research in Science, Engineering and Technology (ijsrset.com)

639
 selected and chromosomes having small values die off. For instance ,in context of network routing problem i.e.
Some of the parent chromosomes are included in the to find the best path in the network , a possible
new generation so as to keep the population size chromosome would be a string consisting of the various
constant. nodes that constitute the paths between source and
destination . In this, the source and the destination nodes
4. Perform crossover. This process allows two are sure to participate in every generation. Other nodes
chromosomes to exchange information and produce two or the genes become a part of the chromosome if they
new chromosomes. Firstly, the parent chromosomes are find an optimal path between the source and destination.
selected randomly from the generation. Then the The fitness function can be the criteria chosen to
Crossover operator combines sub parts of two parent evaluate the path according to the objective of the
chromosomes and produces off spring that contains problem . For example - If the objective is to minimize
some parts of both the parent. These new chromosomes the route delay between source and destination, then the
are returned to the generation. fitness function will compute the route delays of all
paths in the generation and return the minimum delay
5. Perform mutation- path between source to destination. In case of QoS
This is performed to the new chromosomes produced routing problem, the evaluation function will check
by crossover. This operation is the random change of more than one QoS parameters.
some elements /gene in a chromosome. Each gene will
be considered for mutation with a certain probability. Using Genetic algorithm to solve QoS routing problem
Steps 2 to 5 are repeated until a chromosome is found is not a new concept. Many of the researchers have
for the objective function or the termination condition is successfully implemented this concept. There are
met. various algorithms presented in literature to resolve QoS
routing problem with genetic algorithm. These
The performance of GA is based on efficient algorithms are varying in terms of encoding , evaluation
representation, evaluation of fitness function and other of fitness function, size of population, rate of crossover
parameters like size of population, rate of crossover, & mutation , length of chromosome ,selection
mutation and the strength of selection. Genetic mechanism and the number/type of QoS parameters
algorithms are capable to discover optimal or near considered . Some of the work has been explored
optimal solution depending on the selection function. here-
Thus, a genetic algorithm has five basic requirements
for its process as follows- Chang [4] presents a genetic algorithmic approach to the
shortest path routing problem. Variable-length
1. An encoding method, that is a genetic representation chromosomes have been employed. This algorithm
of solutions to the problem i.e. chromosome. displays a much better quality of solution and a much
2. A way to create an initial population of chromosome. higher rate of convergence than other shortest path
3. An evaluation function to test the solutions in terms algorithms. This paper also develops a population-sizing
of their fitness. equation that facilitates a solution with desired quality
4. A selection mechanism for the proposed GA.
5. The genetic operators - crossover and mutation that
helps in reproduction. Inagaki [5] proposed an algorithm that can find not only
shortest route but also semi- shortest routes. This can
IV. Genetic algorithm for QoS routing search for semi-optimal routes by comparing each
problem possible solution in the search process using the feature
of a genetic algorithm. The algorithm employs fixed
GA is fit for network model to find the optimal path. In (deterministic) length chromosomes. The chromosomes
order to use Genetic Algorithms for networking in the algorithm are sequences of integers and each gene
applications, the chromosome consists of the network represents a node ID that is selected randomly from the
parameters as the genes of the chromosomes. set of nodes connected with the node.

International Journal of Scientific Research in Science, Engineering and Technology (ijsrset.com)

640
 In [6],Munetomo proposed a GA-based routing been proposed to deal with the
algorithm i.e. Genetic Routing algorithm with first problem, and the gene structure algorithm (GSA)
Migrations (GRM) which employs genetic operator to to generate the GS. Based on the GS chromosome
generate alternate paths that can be quickly used in the initialization and mutation operator are given to solve
case of link failures. The GRM employs a source the last two problems. In the process of evolution, there
routing approach in order to effectively utilize the is no need to check and repair the illegal chromosomes.
similarity among routes in a routing table. In the The mutation operator based on the GS is the unique
proposed algorithm, the algorithm chromosome is genetic operator used in the evolution to search all
encoded as a list of node IDs that are on the path from potential solutions, which simplifies the genetic
the source node to the destination node. Since different operation. Consequently, the proposed GA can search
paths can have different number of nodes, the the solution space in a very effective and efficient
chromosomes are of variable length. This algorithm manner.
employs crossover, mutation and migration genetic
operators in generating the next generation of solutions. [10] proposes a dynamic routing method based on a
As an initial route, a default route is generated by using Genetic Algorithm. A new string structure and genetic
the Dijkstra’s shortest path algorithm based on a hop operations suitable for network problems is proposed,
count metric. Alternative routes in a routing table are and an optimization method that uses past solutions as
generated by applying genetic operators such as the initial data for new searches is also proposed. These
mutation and crossover. The genetic operators are techniques dramatically improve the efficiency and
invoked at a specified probability after every evaluation convergence speed of the Genetic Algorithm.
of weight values. Migration is performed to exchange
routes among nodes. After migration, the migrated In [11], the genes are put in a chromosome in the same
string is modified by deletion/addition operators. order the nodes form the communication route, so the
chromosomes have different size. If genetic operations
In [7], a GA based routing algorithm which uses two are chosen randomly, the new offsprings of a population
QoS parameters for routing. The proposed method is may be unsuitable individual populations. As a result, a
called ARGAQ method. In ARGAQ method two communication route between two adjacent nodes may
parameters: the Delay Time (DT) and Transmission not exist and some complicated genetic operations
Success Rate (TSR) are considered as QoS parameters. should be carried out in order to find a new
The DT means the time it takes a packet to go from one communication route. Also, because the individuals of a
node to another one. The TSR is the rate of packets population have different sizes, the crossover operations
transmitted correctly without losses. The packet loss are complicated. Moreover, the number of searching
may be caused by the overflow in the queue or from the iterations has been decreased by using the past routing
network trouble. The ARGAQ method is a source-based patterns as a part of the initial data for new searches.
routing mechanism. ARGAQ provides only an The proposed method is therefore useful for dynamic
approximate approach to QoS routing as it computes routing control during traffic fluctuations.
path subject to single mixed metric T, which is the ratio
of Delay Time and Transmission Success Ratio. In [12], Barolli proposed an adaptive routing method
In [8], a multi-purpose optimization method for QoS based on GA (ARGA) subject to one additive metric
routing based on GA has been proposed. The proposed (i.e., TD). In this method, the network is expressed as a
method is a source-based routing method and has a tree network and the tree junctions represent the genes.
flexible and adaptive behaviour. The chromosomes in the algorithm have the fixed length.
[9] presents a genetic algorithm to solve the MCP Genetic operators (i.e., crossover and mutation) are
problem subject to transmission delay and transmission carried out on the tree junctions.
success ratio. Three key design problems are
investigated for this new algorithm, i.e., how to encode [13] proposed a robust GA-based QoS routing
the problem in genetic representation, how to avoid the algorithm for the general k-constrained MCP routing
illegal chromosomes in the process of population problem. The proposed algorithm is also able to produce
initialization and genetic operation, and how to design multiple feasible paths and this enables the algorithm to
effective genetic operator. Gene structure (GS) has become more robust in the case that the actual state

International Journal of Scientific Research in Science, Engineering and Technology (ijsrset.com)

641
 information in the network has changed since the last 2. A "fitness" is calculated for each chromosome .The
state information update received by the router. If the fitness is a value of the objective function for the
first feasible path found turns out to be no longer network configuration. The objective function may be
feasible, there are other alternative paths that can be any function that can check two or more QoS
tried without having to execute a new routing requirements. If the quality of one routing path is better
computation. This will increase the success of finding a than others according to the requirement then its fitness
feasible path that can fulfil a particular QoS function would have a better value.
requirements. Variable length chromosomes have been 3. According to their fitness values, some of paths are
used in this algorithm. Each chromosome consists of a reproduced and some are eliminated.
sequence of nodes that are in the path from sender to 4. The crossover and mutation operators are performed
receiver. on paths to generate new paths. For example- select
two possible paths and perform crossover to generate a
In [14],Multi-constraint QoS Unicast Routing Using new sequence of paths . This may lead to invalid paths
Genetic Algorithm (MURUGA), which incorporates i,e. the paths having loops . So some mechanism should
multiple constraints required by multimedia applications be provisioned to avoid illegal paths.
to find the feasible path satisfying the constraint This process will continue until a optimal path is found
requirement has been proposed. In this paper, a study of or termination condition is met.
the existing GA based QoS routing algorithms has been
done. The drawback in all these existing algorithms has V. CONCLUSION
been the consideration of single or single mixed
constraint while computing the feasible path. The traditional routing method is inadequate for the
MURUGA satisfies multiple additive constraints for successful transmission of networked multimedia
finding a feasible route. MURUGA is a source based applications as these applications require quality
QoS routing algorithm and uses a tree structure for gene parameters to be present in the network like strict delay
coding. The chromosome is initialized based on the & bandwidth guarantee etc. To support these
chromosome structure. The population initialization applications, it is important for the network to provide
based on the chromosome structure can generate Quality-of Service (QoS) guarantee. One of the key
chromosomes without loops. This algorithm adopts a technology to provide QoS guarantees is QoS routing.
heuristic nonlinear fitness function which uses the upper QoS routing means to find a path that satisfy multiple
bound of the QoS constraints to evaluate chromosomes. constraints of QoS like delay, jitter, bandwidth, cost,
This fitness function makes the routing algorithm select hopcount etc.
a feasible route quickly. However, crossover may
generate chromosomes with routing loops. Checking It is well known that Multi-constrained problem is NP-
and repairing function are needed to deal with illegal complete. Heuristic and approximate algorithms for
chromosomes (chromosomes with loops). Multi-constrained problem are not effective in dynamic
network environment for real-time applications when
Based on these works, the generalized approach of the state information of the network is out of date.
executing routing problem through genetic algorithm Genetic algorithm are very good at solving
can be outlined as- optimization problems . Since solving an NPcomplete
QoS routing problem is similar to solving an
1. For initial population, generate the possible paths optimization problem, therefore it is possible to apply
from source to destination. The chromosomes that genetic algorithm to this problem. This paper has
represent paths from a source to destination selected explored this possibility i.e. how to solve a QoS routing
randomly among all the valid routes. These valid routes problem with genetic algorithm. Our future work will
can be collected by route table or calculated by some focus on to develop a QoS routing algorithm based on
method. genetic approach.
The representation of path in chromosome can be done
by either a suitable encoding in the form of string or
directly in the sequence of nodes.

International Journal of Scientific Research in Science, Engineering and Technology (ijsrset.com)

642
 VI. REFERENCES Transactions of the Institute of Electrical
Engineers of Japan C 119 (5) (1999) 624-631.
[1]. D. E. Goldberg, Genetic Algorithms in Search, [13]. S. Yussof,O.H.see, "A Robust GA-based QoS
Optimization and Machine Learning. Addison Routing Algorithm for Solving Multi-constrained
Wesley, 1989. Path Problem" JOURNAL OF COMPUTERS,
[2]. S.Upadhaya, G.Devi, "Exploring Issues for QoS VOL. 5, NO. 9, SEPTEMBER 2010
Based Routing Algorithms" published in [14]. R. Leela, N. Thanulekshmi, S. Selvakumar,
International Journal on Computer Science and "Multi-constraint QoS unicast routing using
Engineering (E-ISSN : 0975-3397 ) Vol. 02, No. genetic algorithm (MURUGA)", Applied Soft
05, 1792-1795,2010. Computing 11 (2) (2011) 1753-1761
[3]. S.Upadhaya, G.Devi, "Characterization of QoS [15]. Xiao, X. & Ni, L. M.(1999). "Internet QoS: A big
Based Routing Algorithms", International Journal picture", IEEE Network, vol.13, no. 2, pp. 8-18,
of Computer Science & Emerging Technologies March-April 1999.
133 Volume 1, Issue 3, October 2010
[4]. Chang Wook Ahn, R.S. Ramakrishna, "A genetic
algorithm for shortest path routing problem and
the sizing of population Evolutionary Computing,
vol. 6, pp. 566 - 579, December 2002
[5]. J. Inagaki, M. Haseyama, and H. Kitajima, "A
genetic algorithm for determining multiple routes
and its applications," in Proc. IEEE Int.
Symp.Circuits and Systems, 1999, pp. 137-140.
[6]. M. Munetomo, Y. Takai, Y. Sato, "A migration
scheme for the genetic adaptive routing
algorrithm", IEEE International Conference on
SystMan and Cybernatics, vol. 3, pp. 2774 - 2779,
October 1998.
[7]. L. Barolli, A. Koyama, H. Sawada, T. Suganuma,
N. Shiratori, ‘A New QoS Routing Approach for
Multimedia Applications based on Genetic
Algorithms’, IEEE CW (2002) 289-295
[8]. L. Barolli, A. Koyama, K. Matsumoto, B.O.
Apduhan, ‘A GA-based Multi-purpose
Optimization Algorithm for QoS Routing’, AINA
(1) (2004) 23-28.
[9]. T.Lu, J. Zhu," A genetic algorithm for finding a
path subject to two constraints"Applied Soft
Computing 13 (2013) 891-898.
[10]. N. Shimamoto, A. Hiramatsu, and K. Yamasaki,
"A dynamic routing control based on a genetic
algorithm," in Proc. IEEE Int. Conf. Neural
Networks, 1993, pp. 1123-1128.
[11]. M.Munetomo, Y.Takai, and Y.Sato, "An
Adaptive Routing Algorithm with Load Balancing
by a Genetic Algorithm", Trans. of IPSJ,Vol.39,
No.2, pp.219-227, 1998.
[12]. L. Barolli, A. Koyama, S. Motegi, S. Yokoyama,"
Performance evaluation of a genetic algorithm
based routing method for high-speed network",

International Journal of Scientific Research in Science, Engineering and Technology (ijsrset.com)

643