chapter 4 proposed hybrid intelligent approch for multiprocessor...
TRANSCRIPT
79
CHAPTER 4
PROPOSED HYBRID INTELLIGENT APPROCH FOR
MULTIPROCESSOR SCHEDULING
The present chapter proposes a hybrid intelligent approach
(IPSO-AIS) using Improved Particle Swarm Optimization (IPSO) with
Artificial Immune System (AIS) for multiprocessor task scheduling problem
with two cases, namely, static task scheduling and dynamic task scheduling
with and without load balancing.
4.1 INTRODUCTION
The drawback in the hybrid approach IPSO-SA is that it slightly
slower in convergence due to the slow speed of SA. To conquer the drawback
of the hybrid algorithm IPSO-SA’s slow convergence, the search is guided
towards untried regions in the solution space and also to improve further, an
immune based intelligent approach is thought of. An Immune system is a
naturally occurring event response system that can quickly adapt to the
changing situations. In AIS, the models of vaccination and receptor editing
are designed to improve the immune performance.
The combined effect of IPSO with AIS directs to a promising
result. The basic idea of combining IPSO with AIS is to combine the good
features of both the algorithms.
In the present chapter, a new hybrid intelligent algorithm is
developed using IPSO with AIS to achieve better solutions.
80
4.2 REVIEW OF LITERATURE
Luh and Liu (2004) developed a Reactive Immune Network (RIN)
for mobile robot learning navigation strategies within unknown environments.
In their approach, a modified virtual target method is integrated to solve the
local minima problem.
Wojtyla et al (2006) proposed an efficient method of extracting
knowledge when scheduling parallel programs onto processors using AIS.
The author’s proposed approach reorders the nodes of the program according
to the optimal execution order on one processor, which works in either
learning or production mode. In the learning mode an immune system to
optimize the allocation of the tasks to individual processors was used. In the
production mode the optimization module is not invoked, only the stored
allocations are used. The proposed approach gives similar results to the
optimization by a Genetic Algorithm (GA) but requires only a fraction of
function evaluations.
Fu et al (2007) proposed a hybrid artificial immune network which
uses the swarm learning of Particle Swarm Optimization to speed up the
convergence of the artificial immune system.
Yu (2008) developed an algorithm based on AIS to schedule for
heterogeneous computing environments. Empirical studies on bench mark
task graphs show that this algorithm significantly outperforms a deterministic
algorithm.
Lin and Ying (2012) developed an algorithm based on revised AIS
and Simulated Annealing effect (RAIS) to minimize the makespan of
blocking flow shop. The proposed algorithm is evaluated and compared with
81
well-known bench mark problems of taillard used. The results show that the
proposed algorithm outperforms the state-of-art algorithms on the same bench
mark problems.
Engin et al (2004) used a computational method based on the
principle of clonal selection and affinity maturation mechanism of the
immune response. The objective is to minimize the makespan of the
scheduling problem. The operating parameters of meta-heuristics play an
important role. He presents a genetric systematic procedure which is based on
a multi-step experimental design approach for determining the optimum
system parameters of AIS and tested with benchmark problems. Results infer
that an AIS based algorithm is effective to solve hybrid flow shop problems.
Lin et al (2011) presented a Hybrid Taguchi-Immune Algorithm
(HTIA) to deal with the unit commitment problem, which integrates Taguchi
method and traditional immune algorithm and provides a powerful global
exploration capability. The proposed algorithm shows better performance
compared with the other methods published before. The test results reveal that
the proposed method is feasible and robust and more effective.
Bagheri et al (2010) proposed an artificial based on integrated
approach to solve the flexible job shop scheduling to minimize makespan.
The algorithm uses several strategies for selecting the individuals for
reproduction and also different mutation operators used for reproducing new
individuals. The proposed approach is validated with benchmark problems,
and computational results show the quality.
Zandieh et al (2006) used an immune algorithm to tackle complex
problems of a Sequence Dependent Setup Times (SDST) and produce a
82
reasonable manufacturing schedule within an acceptable time. Results were
compared with the Random Key Genetic Algorithm (RKGA) presented
previously. From the results, it is known that IA outperformed RKGA.
King et al (2001) described an intelligent agent for task allocation
in a heterogeneous computing environment. It exploits some of the
functionalities in designing agent-based parallel control systems.
Sun and Yang (2008) presents a new hybrid optimization algorithm
which combines the strong global search ability of Artificial Immune System
(AIS) with a strong local search ability of External Optimization (EO)
algorithm. The algorithm tested with benchmark problems with makespan
criterion, and results show that the proposed method is effective.
Ebrahimi Moghaddam et al (2012) an Immune based Genetic
Algorithm (IGA) is proposed which reduces the search space of
Multiprocessor Task Scheduling Problem(MTSP), and effectively influences
the convergence speed of the optimization process and guarantees the validity
of the solutions by using crossover and mutation operators. Experimental
results showed that the proposed algorithm uses less number of iterations to
find the solution, when applied to the benchmark problems.
Kahraman et al (2009) proposed a new artificial immune system
algorithm to solve multi objective fuzzy flow shop scheduling problem, in
which fuzzy sets are used to model processing times and due dates. The
objective of the algorithm is to minimize the average tardiness and number of
tardy jobs. The effectiveness are tested by comparing it with genetic
algorithms. The outcome tells that the proposed algorithm is more effective.
83
Zelenka (2011) compared two approaches for Job Shop Scheduling
Problem (JSSP) in real manufacturing system. The first approach is based on
the mechanisms inspired by biological evolution and the immune system. The
second stochastic optimization algorithm is based on social simulation
models.
4.3 HYBRID INTELLIGENT ALGORITHM
The main objective of the hybridization is to integrate different
learning and adaptation techniques to overcome individual limitations and to
achieve synergetic effects through the combination of these techniques.
Biological immune systems can be viewed as a powerful
distributed information processing systems, capable of learning and self-
adaptation. AIS is rapidly emerging, which is inspired by theoretical
immunology and observed immune functions, principles and models.
The efficient mechanisms of immune system, including clonal
selection, learning ability, memory, robustness and flexibility make AISs
useful in many applications. AIS appear to offer powerful and robust
information processing capabilities for solving complex problems. Hence, a
hybrid intelligent algorithm IPSO-AIS is developed to improve the
performance of the multiprocessor scheduling problem.
4.3.1 Basic AIS-Based Scheduling Algorithm
The brief outline of the proposed algorithm based on AIS can be
described as follows (Ge et al 2008).
84
Step 1 : Initialize pop_size antibodies as an initial population, where
pop_size denotes the population size.
Step 2 : Select m antibodies from the population by the proportional
selection model and clone them to a clonal library.
Step 3 : Perform the mutation operation for each of the antibodies in the
clonal library.
Step 4 : Randomly select‘s’antibodies from the clonal library to perform
the operation of vaccination.
Step 5 : Replace the worst‘s’antibodies in the population by the
best‘s’antibodies from the clonal library.
Step 6 : Perform the operation of receptor editing if there is no
improvement of the highest affinity degree for a certain number
of generations.
Step 7 : Stop if the termination condition is satis ed; else, repeat Steps 2
to 7.
4.4 PROPOSED HYBRID INTELLIGENT ALGORITHM
(IPSO-AIS)
The steps involved in the proposed hybrid algorithm (Improved
PSO with Artificial Immune System is as follows,
Step 1 : Initialize Population size of the antibodies as PSA.
Step 2 : Initialize the number of particles N and its value may be generated
randomly. Initialize swarm with random positions and velocities.
85
Step 3 : Compute the finishing time for each and every particle using the
objective function and also find the “pbest “. If current fitness of
particle is better than “ pbest” the set “ pbest” to current value. If
“pbest” is better than “gbest then set “gbest” to current particle
fitness value.
Step 4 : Select particles individual “pworst” value, that means particle
moving away from the solution point.
Step 5 : Update the velocity and position of particle as per Equation (2.1)
and (2.2).
Step 6 : If best particle is not changed over a period of time,
a) Select ‘m’ antibodies out of the population PSA by the
proportional selection model and clone them to a colonal
library.
Step 7 : Perform the mutation operation for each of the antibodies in the
clonal library.
Step 8 : Randomly select ’s’ antibodies from the clonal library to perform
the operation of vaccination.
Step 9 : Replace the worst’s’ antibodies in the population by the best‘s’
antibodies from the clonal Library
Step 10 : Terminate the process if maximum number of iterations reached
or optimal value is obtained, else go to step 3.
The flow chart for the hybrid algorithm is shown in Figure 4.1
86
Start
Initialize population size of Antibodies
Initialize the population Input number of processors, number of jobs and population size
D
Compute the objective function
Invoke Hybrid algorithm
If E < best ‘E’ (Pbest) so far
For each generation
For each particle
Search is terminatedoptimal solution
reached
B
Current value = new p best
Choose the minimum F of all particles as the g best
Calculate particle velocity using Equation (2.1)
A
No
Yes
Figure 4.1 Flowchart for the proposed hybrid intelligent approachIPSO-AIS
87
Calculate particle position using (2.2)
Update memory of each particle
If the best particle is notchanged over a period of time
Select ‘m’ antibodies from the population and clonethem to clonal library
Perform mutation operation to the antibodies
No
Perform vaccination operation on randomly selected‘s’ antibodies
C
B
Yes
A
Replace the worst antibodies by best antibodies
Figure 4.1 (Continued)
88
Perform receptor editing operation
If improvement inhighest affinity degree
Yes
No
Yes
No
End
End
If stopping conditionreached
Stop
B
D
C
Figure 4.1 (Continued)
89
4.5 SIMULATION PROCEUDRE
The details of the simulation carried out for implementing theproposed hybrid algorithm is given in the present section.
Benchmark datasets are taken from EricTailard’s site for dynamic
task scheduling. Two datasets are taken for simulation. Data set 1involves
50 tasks and 20 processors. Data set 2 involves 100 tasks with 20 processors.
The data for the static scheduling is randomly generated, such as 2 processors
with 20 tasks, 3 processors with 20 tasks, 3 processors with 40 tasks,
4 processors with 30 tasks, 4 processors with 50 tasks, 5 processors with
45 tasks and 5 processors with 60 tasks.
To demonstrate the effectiveness of the proposed hybrid algorithm,
the proposed approach is run with 30 independent trials with different values
of random seeds and control parameters. The optimal result is obtained for
following the parameter settings
Artificial Immune System
Number of generations : 200
Mutation rate : 0.1
Sampling rate : 0.1
Antibodies : Twice the number of tasks
Improved Particle Swarm Optimization
The initial solution is generated randomly
C1g, C1b and C2 : 2,2 and 2
Population size : Twice the number of tasks
(Salman et al 2002)
90
Wmin - Wmax : 0.5
Max. Iteration : 500
The proposed hybrid approach IPSO-ACO is developed using
MATLAB R2009 and executed in a PC with Intel core i3 processor with
3 GB RAM and 2.13 GHz speed.
4.6 STATIC SCHEDULING
The tasks considered are independent. Hence, the tasks can be
executed in any order in any processor. The objective function is the same as
specified in the Equations (2.4) to (2.9). The application of intelligent hybrid
algorithm (IPSO-AIS) for scheduling multiprocessor tasks is shown in the
present chapter.
4.6.1 Results and Discussion
The proposed hybrid approach IPSO-AIS is tested for static task
scheduling problem with the datasets specified in the simulation procedure
and the results achieved are shown in Table 4.1.
Table 4.1 Total finishing time and average waiting time using theproposed hybrid intelligent approach IPSO-AIS
No of Processors No of jobsProposed IPSO-AIS
AWT TFT2 20 22.16 52.643 20 38.65 48.373 40 34.26 61.204 30 23.92 65.474 50 25.96 67.835 45 27.56 64.965 60 30.19 69.01
91
The proposed hybrid approach IPSO-AIS is tested with various
randomly generated datasets. For the dataset 3 processors with 40 tasks,
IPSO-AIS produces total finishing time 61.20s and average waiting time
34.26s, for dataset 4 processors with 50 tasks 67.83s as total finishing time
and 25.96s as average waiting time and for dataset 5 processors with 60 tasks
30.19 s as average waiting time and 69.01s as total finishing time.
4.6.2 Performance Comparison
In order to validate the performance of the proposed hybrid
intelligent approach IPSO-AIS, comparisons have been made with the
approaches IPSO, IPSO-SA with the same datasets, and are reported in
Table 4.2. These results reveal that the proposed hybrid approach IPSO-AIS
is comparatively better than the other approaches.
Table 4.2 Comparison of job finishing time and average waiting timeusing IPSO, IPSO-SA and the proposed IPSO-AIS
No ofProcessors
No ofjobs
IPSO IPSO-SAProposedIPSO-AIS
AWT TFT AWT TFT AWT TFT2 20 29.12 57.34 25.61 54.23 22.16 52.643 20 45.00 54.01 40.91 50.62 38.65 48.373 40 41.03 69.04 38.45 65.40 34.26 61.204 30 29.74 70.97 26.51 66.29 23.92 65.474 50 30.06 70.62 28.34 68.01 25.96 67.835 45 33.65 68.04 30.12 66.43 27.56 64.965 60 36.56 72.31 32.76 69.13 30.19 69.01
For the dataset 3 processors with 40 tasks, IPSO produces average
waiting time 41.03s, hybrid approach IPSO-SA produces as 38.45s and the
proposed hybrid intelligent approach IPSO-AIS produces as 34.26s. For the
92
same dataset, total finishing time produced by IPSO is 69.04s, by hybrid
approach IPSO-SA is 65.40s and by the proposed intelligent approach is
61.20s. For the dataset 5 processors with 60 tasks, IPSO produces average
waiting time as 36.56s and total finishing time as 72.31s, IPSO-SA produces
average waiting time as 32.76s and total finishing time as 69.13s and the
proposed hybrid intelligent algorithm IPSO-AIS produces 30.19s as average
waiting time and total finishing time as 69.01s. It is empirically proved that
the proposed hybrid approach IPSO-AIS simultaneously reduces both Total
finishing time and average waiting time.
Thus, based on the results, it is inferred that the proposed hybrid
intelligent algorithm IPSO-AIS produces better results than the conventional
methodologies LPT, SPT, GA, standard PSO, and hybrid approach IPSO-SA.
The variations found in the total finishing time and average waiting
time using different approaches namely, using IPSO, IPSO-SA and IPSO-AIS
are shown from Figures 4.2 to 4.8.
Figure 4.2 Total Finishing Time and Average waiting time for 2processors with 20 jobs using IPSO, IPSO-SA and IPSO-AIS
93
Figure 4.3 Total Finishing Time and Average waiting time for 3processors with 20 jobs using IPSO, IPSO-SA and IPSO-AIS
Figure 4.4 Total Finishing Time and Average waiting time for 3processors with 40 jobs using IPSO, IPSO-SA and IPSO-AIS
94
Figure 4.5 Total Finishing Time and Average waiting time for 4processors with 30 jobs using IPSO, IPSO-SA and IPSO-AIS
Figure 4.6 Total Finishing Time and Average waiting time for 4processors with 50 jobs using IPSO, IPSO-SA and IPSO-AIS
95
Figure 4.7 Total finishing time and average waiting time for 5processors with 45 jobs using IPSO, IPSO-SA and IPSO-AIS
Figure 4.8 Total finishing time and average waiting time for 5processors with 60 jobs using IPSO, IPSO-SA and IPSO-AIS
96
Thus, the results reveal that the proposed IPSO-AIS produce an
improvement in the performance, when compared to the standard PSO and
hybrid approach IPSO-SA.
4.7 DYNAMIC TASK SCHEDULING WITHOUT LOAD
BALANCING
In the dynamic task scheduling problem, reducing the total
completion time of processors is a major issue. Hence, to minimize the
makespan of the entire schedule, the objective function is represented in
Equations (2.10) to (2.12).
4.7.1 Results and Discussion
The obtained results have been tabulated and shown in Table 4.3,
which represents the cost and convergence time comparison of IPSO, IPSO-
SA and Intelligent Hybrid Algorithm. The results reveal that the IPSO-AIS
performs better than the other algorithms.
Table 4.3 Best cost, worst cost, average cost and convergence timeusing IPSO, IPSO-SA and the proposed hybrid intelligentapproach IPSO-AIS for dynamic task scheduling withoutload balancing
Method IPSO IPSO-SAProposedIPSO-AIS
Number of tasks 50 100 50 100 50 100
Best Cost 2374 4527 2156 4376 2136 4309
Worst Cost 3136 5213 2901 4908 2886 4856
Average Cost 2755 4870 2528.5 4624 2511 4582.5
Convergence Time 4.0521 5.7112 4.2156 5.8428 4.9124 7.4682
97
Comparisons have been made based on the algorithms with IPSO,
IPSO-SA and the proposed hybrid intelligent approach IPSO-AIS on the best,
average and worst cost achieved for dynamic task scheduling. For dataset 1,
IPSO achieves the best cost as 2374, IPSO-SA achieves the best cost as 2156
and the proposed hybrid intelligent approach IPSO-AIS achieves the best cost
as 2136. For dataset 2, IPSO produces best cost 4527, IPSO-SA produces the
best cost as 4376 and the proposed hybrid intelligent approach IPSO-AIS
produces the best cost as 4309. The proposed hybrid approach IPSO-AIS
produces better results compared with other algorithms, but the convergence
time for the proposed hybrid algorithm IPSO-AIS is higher (app 1.16 times)
than IPSO-SA, because of the extra calculation involved in the immunization.
The best cost obtained using the proposed hybrid intelligent
approach IPSO-AIS for dataset 1 and dataset2 are shown in Figures 4.9 and
4.10.
Figure 4.9 Best costs for 50 tasks and 20 processors using IPSO,IPSO-SA and IPSO-AIS
98
Figure 4.10 Best costs for 100 tasks 20 processors using IPSO, IPSO-SAand IPSO-AIS
The proposed hybrid intelligent approach IPSO-AIS performs
better than with the IPSO and the hybrid algorithm IPSO-SA.
4.7.2 Performance Comparison
The performance of the proposed hybrid approach IPSO-AIS is
compared with the previously proposed ((Visalakshi and Sivanandam 2009)
hybrid PSO algorithms PSO-HC and PSO-SA for the same datasets.
Table 4.4 Performance comparison of various PSO based hybridapproaches
MethodPSO-HC
(Visalakshi andSivanandam 2009)
PSO-SA(Visalakshi and
Sivanandam 2009)
ProposedIPSO-AIS
Number oftasks 50 100 50 100 50 100
Best cost 2322 4621 2186 4496 2136 4309Worst cost 2994 5449 2916 4948 2886 4856
Average cost 2658 5035 2551 4722 2511 4582.5Convergence
time in seconds 4.9636 7.3588 6.4311 8.7349 4.9124 7.4682
99
For the dataset 1, PSO-HC produces the best cost as 2322, PSO-SA
produces the best cost as 2186 and the proposed hybrid approach IPSO-AIS
produces the best cost as 2136. For dataset 2, PSO-HC produces the best cost
as 4621, PSO-SA produces the best cost as 4496 and the proposed hybrid
approach IPSO-AIS produces the best cost as 4309. The proposed hybrid
intelligent approach IPSO-AIS performs better when compared with the other
previously proposed hybrid methods of PSO-HC and PSO-SA.
Thus, the result concludes that the proposed hybrid intelligent
approach IPSO-AIS performs better than the other hybrid approaches PSO-
HC and PSO-SA.
4.8 DYNAMIC TASK SCHEDULING WITH LOAD BALANCING
In order to improve the performance and utilization of
multiprocessor system, load balancing of tasks have to be considered.Therefore, the concept of load balancing is dealt, in which the objective
function is the same as represented in the Equations (2.13) to (2.16).
4.8.1 Results and Discussion
Table 4.5 illustrates the best cost, worst cost, average cost and
convergence time for, IPSO, hybrid algorithm IPSO-SA and the proposed
hybrid intelligent algorithm IPSO-AIS.
For dataset 1, the best cost achieved using IPSO is 12.0042, IPSO-
SA produces the best cost as 12.9961and the proposed hybrid intelligent
algorithm IPSO-AIS produces the best cost as 13.0014. For dataset 2, the best
cost produced by IPSO is 21.4291, IPSO-SA produces the best cost as
22.0223 and the proposed hybrid intelligent approach IPSO-AIS produces the
best cost as 22.132. The average cost is also improved in the proposed hybrid
intelligent algorithm IPSO-AIS. The convergence time for the proposed
100
IPSO-AIS method is 6.2154s for dataset 1 and 8.3992s for dataset 2, which is
higher than the hybrid algorithm IPSO-SA.
Table 4.5 Best cost, worst cost, average cost and convergence time usingIPSO, IPSO-SA and the proposed hybrid intelligent approachIPSO-AIS for dynamic task scheduling with load balancing
Method IPSO IPSO-SAProposedIPSO-AIS
Number of tasks 50 100 50 100 50 100
Best Cost 12.0042 21.4291 12.9961 22.0223 13.0014 22.132
Worst Cost 10.9820 19.2103 11.4832 20.9313 11.4881 20.9474
Average Cost 11.4931 20.3197 12.2396 21.4768 12.2448 21.5397
Convergence Timein seconds 5.1176 6.9064 5.1284 6.9205 6.2154 8.3992
The best cost obtained using the intelligent hybrid algorithm
IPSO-AIS for data set 1 and data set2 are shown in Figures 4.11 and 4.12.
Figure 4.11 Best costs for 50 tasks and 20 processors using IPSO,IPSO-SA and IPSO-AIS
101
Figure 4.12 Best costs for 100 tasks and 20 processors using IPSO,IPSO-SA and IPSO-AIS
Thus, the results conclude that the proposed hybrid intelligent
approach IPSO-AIS performs well when compared to the standard PSO, IPSO
and IPSO-SA for the dynamic task scheduling problem with load balancing
concept. However, the time taken for convergence is slightly (app 1.2 times)
higher than IPSO-SA.
4.8.2 Performance Comparison
The performance of the proposed IPSO is compared with the
previously proposed (Visalakshi and Sivanandam 2009) hybrid PSO
algorithms PSO-HC and PSO-SA for the same datasets.
For the dataset 1, PSO-HC produces the best cost as 12.008, PSO-
SA produces the best cost as 12.982 and the proposed hybrid approach IPSO-
AIS produces best cost as 13.0014. For dataset 2, PSO-HC produces best cost
as 21.114, PSO-SA produces 21.998 as the best cost and the proposed hybrid
approach IPSO-AIS produces 22.132 as the best cost. The proposed hybrid
102
intelligent approach IPSO-AIS perform well when compared with the other
previously (Visalakshi and Sivanandam 2009) proposed hybrid methods PSO-
HC and PSO-SA.
Table 4.6 Performance comparison of various PSO based hybridapproaches
MethodPSO-HC
(Visalakshi andSivanandam 2009)
PSO-SA(Visalakshi and
Sivanandam 2009)
ProposedIPSO-AIS
Number of tasks 50 100 50 100 50 100
Best cost 12.008 21.114 12.982 21.998 13.0014 22.132
Worst cost 9.885 19.392 11.476 20.926 11.4881 20.9474
Average cost 10.9465 20.253 12.229 21.462 12.2448 21.5397
Convergencetime in seconds 6.2172 8.4994 7.6559 10.6415 6.2154 8.3992
Thus, the comparison reveals that the proposed hybrid approach
IPSO-AIS achieves better results than the other approaches.
4.9 CONCLUSION
The chapter four has thus dealt with the application of IPSO-AIS to
solve different types of multiprocessor task scheduling with two cases,
namely, static independent task scheduling and dynamic scheduling with and
without load balancing.
The proposed hybrid intelligent approach IPSO-AIS is tested with a
static task scheduling problem to reduce both the total finishing time and
average waiting time. For the dataset 5 processors with 60 tasks, IPSO
produces an average waiting time of 36.56s and the total finishing time of
72.31s, IPSO-SA produces average waiting time of 32.76s and the total
103
finishing time of 69.13s. The proposed hybrid intelligent algorithm IPSO-AIS
produces 30.19s as average waiting time and the total finishing time as 69.01s
for the same dataset. Thus, IPSO-AIS reduce simultaneously both the total
finishing time and average waiting time.
The proposed hybrid intelligent approach IPSO-AIS is applied to
dynamic task scheduling without load balancing problem. For dataset 2, IPSO
produces the best cost as 4527, IPSO-SA produces the best cost as 4376 and
the proposed hybrid intelligent approach IPSO-AIS produces 4309.
The proposed hybrid intelligent approach IPSO-AIS is applied to
dynamic task scheduling with load balancing problem. For dataset 2, the best
cost produced by IPSO is 21.4291, IPSO-SA produces the best cost as
22.0223 and the proposed hybrid approach IPSO-AIS produces the best cost
as 22.132.
The proposed hybrid intelligent approach IPSO-AIS is compared
with the other hybrid approaches which were earlier proposed, namely,
PSO-HC and PSO-SA. The results infer that the proposed hybrid intelligent
approach IPSO-AIS improves the performance of the scheduling.
The proposed hybrid intelligent approach reduces the makespan for
both static and dynamic task scheduling problems, but there is slight increase
in the convergence time (app 1.15 times) when compared with IPSO-SA for
dynamic task scheduling. Hence, other hybrid technologies need to be tried so
that the convergence time is better than the methodologies tried out. Hence,
new hybrid algorithms are proposed in the subsequent chapters to further
refine the cost and the convergence time achieved, which is the main
objective of task scheduling. The next chapter deals with the hybrid
algorithm, Improved Particle Swarm Optimization with Ant Colony
Optimization.