A Hybrid Genetic based Approach for Real-time Reconfigurable
Scheduling of OS Tasks in Uniprocessor Embedded Systems
Ibrahim Gharbi
1,2
, Hamza Gharsellaoui
3,4,5
and Sadok Bouamama
1
1
National School of Computer Science, ENSI, University of Manouba, Manouba, Tunisia
2
Arar College of Technology, TVTC, Al Hudud ash Shamali, Kingdom of Saudi Arabia (KSA)
3
Higher School of Technology and Computer Science of Tunis (ENICarthage), Carthage University, Carthage, Tunisia
4
LISI-INSAT Laboratory, (INSAT), Carthage University, Carthage, Tunisia
5
Al-Jouf College of Technology, TVTC, Al-Jouf, Kingdom of Saudi Arabia (KSA)
Keywords:
Real-time Scheduling, Genetic Algorithms, Reconfigurable Embedded Systems, Hybridization.
Abstract:
This paper deals with the problem of scheduling uniprocessor real-time tasks by a hybrid genetic based
scheduling algorithm. Nevertheless, when such a scenario is applied to save the system at the occurrence
of hardware-software faults, or to improve its performance, some real-time properties can be violated at run-
time. We propose a hybrid genetic based scheduling approach that automatically checks the systems feasibility
after any reconfiguration scenario was applied on an embedded system. Indeed, if the system is unfeasible,
the proposed approach operates directly in a highly dynamic and unpredictable environment and improves a
rescheduling performance. This proposed approach which is based on a genetic algorithm (GA) combined
with a tabu search (TS) algorithm is implemented which can find an optimized scheduling strategy to resched-
ule the embedded system after any system disturbance was happened. We mean by a system disturbance any
automatic reconfiguration which is assumed to be applied at run-time: Addition-Removal of tasks or just mod-
ifications of their temporal parameters: WCET and/or deadlines. An example used as a benchmark is given,
and the experimental results demonstrate the effectiveness of proposed genetic based scheduling approach
over others such as a classical genetic algorithm approach.
1 INTRODUCTION
Today, real-time embedded systems are found in
many diverse application areas including; automo-
tive electronics, avionics, telecommunications, space
systems, medical imaging, and consumer electron-
ics. In all of these areas, there is rapid technologi-
cal progress. The software engineering principles for
embedded system should address specific constraints
such as hard timing constraints, limited memory and
power use, predefined hardware platform technology,
and hardware costs. On the other hand, the new gen-
erations of embedded control systems are addressing
new criteria such as flexibility and agility (HGS12).
For these reasons, there is a need to develop tools,
methodologiesin embeddedsoftware engineering and
dynamic reconfigurable embedded control systems as
an independent discipline. Each system is a subset
of n tasks. Each task is characterized by its worst
case execution times (WCETs) C
i
, an offset (release
time) a
i
, a period T
i
and a deadline D
i
. The general
goal of this paper is to be reassured that the system
is feasible and meets real-time constraints even if we
change its implementation and to correctly allow the
minimization of the response time of this system af-
ter any reconfiguration scenario (HGS12). To obtain
this optimization (minimization of response time), we
propose a hybrid genetic-based approach in which a
software agent is deployed to dynamically adapt the
system to its environment by applying reconfiguration
scenarios. A reconfiguration scenario means the ad-
dition, removal or update of tasks in order to save the
whole system on the occurrence of hardware/software
faults, or also to improve its performance when ran-
dom disturbances happen at run-time. Indeed, many
real-time systems rely on the EDF scheduling algo-
rithm in the case of uni-processor scheduling theory.
This algorithm has been shown to be optimal under
many different conditions. For example, for inde-
pendent, preemptive tasks, on a uni-processor, EDF
is optimal in the sense that if any algorithm can find
a schedule where all tasks meet their deadlines, then
385
Gharbi I., Gharsellaoui H. and Bouamama S..
A Hybrid Genetic based Approach for Real-time Reconfigurable Scheduling of OS Tasks in Uniprocessor Embedded Systems.
DOI: 10.5220/0005463903850390
In Proceedings of the 17th International Conference on Enterprise Information Systems (ICEIS-2015), pages 385-390
ISBN: 978-989-758-096-3
Copyright
c
 2015 SCITEPRESS (Science and Technology Publications, Lda.)

EDF can meet the deadlines (Der74).
According to (BV11), a hyper-period is defined as
HP = [ζ, 2∗LCM+ζ], where LCM is the well-known
Least Common Multiple of the tasks periods and ζ is
the largest task offset. This algorithm, in our origi-
nal paper assumes that sporadic and aperiodic tasks
span no more than one hyper-period of the periodic
tasks HP and the minimization of the response time
is evaluated for each reconfiguration scenario.
The organization of the paper is as follows. Sec-
tion 2 introduces the related work of the proposed
approach. In Section 3, we present the new genetic-
based approach for hybrid optimal scheduling theory.
Section 4 presents the performance study and dis-
cusses experimental results of the proposed approach
research. Section 5 summarizes the main results and
presents the conclusion of the proposed approach and
describes the intended future works.
2 BACKGROUND
We present related works dealing with reconfigura-
tions and real-time scheduling of embedded systems.
Companies building embedded real-time systems are
driven by a profit motive. To succeed, they aim to
meet the needs and desires of their customers by pro-
viding systems that are more capable, more flexible,
and more effective than their competition, and by
bringing these systems to market earlier. This desire
for technological progress has resulted in a rapid in-
crease in both software complexity and the processing
demands placed on the underlying hardware (BV11).
Nowadays, several interesting studies have been
published to develop reconfigurable embedded con-
trol systems. In (CAM05) Marian et al. propose
a static reconfiguration technique for the reuse of
tasks that implement a broad range of systems. The
work in (MNRE07) proposes a methodology based
on the human intervention to dynamically recon-
figure tasks of considered systems. In (ASV07),
an ontology-based agent is proposed by Vyatkin et
al. to perform system reconfigurations according to
user requirements and also the environment evolu-
tion. Window-constrained scheduling is proposed in
(WS99), which is based on an algorithm named dy-
namic window-constrained scheduling (DWCS). In
(PBC02), a window-constrained execution time can
be assumed for reconfigurable tasks in n among m
windows of jobs. In the following, we consider pe-
riodic, sporadic and aperiodic tasks. So, we note that
the optimal scheduling algorithm based on the EDF
principles and on the dynamic reconfiguration sce-
nario is that we propose in the current original work
in which we give solutions computed and presented
to respond to the systems requirements.
3 GENETIC ALGORITHMS
APPROACH
Based on the (BLS06) works, we will describe our
hybrid approach in order to achieve our objective. We
assume that our system is a mixture of a set of n peri-
odic, sporadic and aperiodic tasks noted ξ
n
. By con-
sidering that m new tasks are added to ξ
n
, our system
becomes ξ
n+m
which is composed of n old tasks and
m new tasks and some old/new tasks can miss their
deadlines. At this moment, a reconfiguration scenario
is automatically applied at real-time to adapt the sys-
tem to its environment. Or to have a feasible system
after each reconfiguration scenario, the difference be-
tween the deadline and the end of execution of each
task must be greater or equal to 0.
3.1 Model
In this section we will present some preliminaries
concepts and we will describe our contribution after.
ξ denotes a set of active sporadic and aperiodic tasks
σ
i
ordered by increasing deadline in a linked list.
a
i
denotes the arrival time of task σ
i
.
C
i
denotes the maximum computation time of task σ
i
.
c
i
denotes the dynamic computation time of task σ
i
,
i.e., the remaining worst case execution time needed
for the processor, at the current time, to complete task
σ
i,k
without interruption.
d
i
denotes the absolute deadline of task τ
i
.
D
i
denotes the relative deadline of task σ
i
. S
i
denotes
the first start time of task σ
i
. s
i
denotes the last start
time of task σ
i
.
f
i
denotes the estimated finishing time of task σ
i
.
L
i
denotes the laxity of task σ
i
.
R
i
denotes the residual time of task σ
i
. Baruah et al.
(SBS91) present a necessary and sufficient feasibility
test for synchronous systems with pseudo-polynomial
complexity. For this reason, we present the following
relationships among the parameters defined above:
d
i
= a
i
+ D
i
(1)
L
i
= d
i
- a
i
- C
i
(2)
R
i
= d
i
- f
i
(3)
f
1
= t + c
1
; f
i
= f
i−1
+ c
i
∀ i > 1 (4)
R
1
= d
1
- t - c
1
(5)
R
i
= R
i−1
+ (d
i
− d
i−1
) - c
i
. (6) (BS93)
For any other task σ
i
, with i > 1,
f
i
= f
i−1
+ c
i
(7) and, by equation (3), we have:
R
i
= d
i
- f
i
= d
i
- f
i−1
- c
i
= d
i
- (d
i−1
- R
i−1
) - c
i
=
R
i−1
+ (d
i
- d
i−1
) - c
i
ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems
386

3.2 Chromosomes
In our work, we encode the solution to the problem
into a chromosome to specify a GA and as conse-
quence, we define/implement the crossover operator
and the tabu search algorithm against the mutation op-
erator on the works of authors in (BLS06).
To optimize the problem, we will optimize the ar-
rival times of all sporadic and aperiodic tasks of all
the genes composing the chromosome. A gene can
be depicted as a pair of character and integer values
(τ
i
, a
i, j
); that is a sporadic or aperiodic task symbol
and an arrival time. The chromosome’s size which is
the number of genes was defined as the total number
of executions of all sporadic and aperiodic tasks. The
length of the chromosome is l =
∑
h
i=1
l
H
min
i
m
, where
h is the number of sporadic and aperiodic tasks and
k
i
=
l
H
min
i
m
is the maximum number of executions for
sporadic/aperiodic task τ
i
over an hyper-period H. As
defined by (BLS06), to depict a non-existent arrival
time, we use a special value: -1. To handle the con-
straints of the genes of the chromosome, we need a
good design. These constraints are defined by two
consecutive arrival times for a particular event which
must have a difference of at least the minimum inter-
arrival time, and at most the maximum inter-arrival
time if it exists, else, it is set to H. Also, in order to
facilitate chromosome manipulations, all genes corre-
sponding to the same task τ
i
are grouped together and
ordered increasingly in a linked list Link
i
according to
a
i, j
. For example, given a set of 3 aperiodic tasks τ
1
(min
τ
1
= 6); τ
2
(min
τ
2
= 9) and τ
3
(min
τ
3
= 12). We
consider the following chromosome (τ
1
, -1) (τ
1
, 10)
(τ
1
, 16) (τ
2
, 11) (τ
2
, 20) (τ
3
, 12) (τ
3
, 25) (τ
3
, 37) as
a valid chromosome in a time interval (hyper-period)
H = [ζ, 2∗ LCM + ζ]; with ζ = 0 and lcm of (6, 9, 12)
= 36, i.e., H = [0,72]. In contrast, the following chro-
mosome (τ
1
, -1) (τ
1
, 10) (τ
1
, 16) (τ
2
, 11) (τ
2
, 20)
(τ
3
, 12) (τ
3
, 25) (τ
3
, 31) is not valid since the mini-
mum inter-arrival time of the third task is not satisfied
by the two last genes.
3.3 Initialization
We randomly generate the initial population at first
with respect to the constraints mentioned above. Af-
ter that, we find the hyper-period H based on the lcm
of all periods of tasks τ
i
. The authors in (BLS06)
used a value T as a maximum of the time inter-
val but we use in our work the hyper-period H. The
value of a
i, j
is randomly selected from a range deter-
mined by the arrival time of a
i, j−1
as well as min
i
and
max
i
. If max
i
is not specified, its value is set to H:
[a
i, j−1
+ min
i
, a
i, j−1
+ max
i
]. Also, if there is no pre-
vious gene (i.e., j = 1), or the previous gene depicts a
non-existent arrival time (i.e., a
i, j−1
= -1), the range
is [0, max
i
]. After that, the genes are ordered in an
increasing order and putted in a linked list (Link
i
).
3.4 Cross-over
Cross-over and mutation operators are the ways GAs
explore a solution space. Hence, they must be formu-
lated in such a way that they efficiently and exhaus-
tively explore the solution space (LE98).
On the other hand, task of formulating operators
is rather difficult as genetic operators must maintain
allowability. In other words, genetic operators must
be designed in such a way that if a constraint is not
violated by the parents, it will not be violated by the
children resulting from the operations (EEZ94).
In (BLS06), the authors adopt the cross-over and
mutation operators. In contrast, in our original work,
we will adopt cross-over operator and tabu search al-
gorithm on behalf of the mutation operator and we
will test this hybrid approach in order to prove its ef-
ficiency. Indeed, the crossover operator is special in
order to respect the constraint that every task which is
copied didn’t repeated another time in order to respect
the tasks unicity and creates two solutions or childs by
combining two parents (HHS14).
The principle of the crossover operator is to pass
traits from the parents to the two resulting childs (off-
springs) with variety methods. In (M.A), the authors
used the operator crossover called sexual crossover
or n-point crossover. The general idea in n-point
crossover adopted by (M.A) is that the two parent
chromosomes are aligned and cut into n+1 fragments
at the same places to identify the division points. Af-
ter that, two new childrens are created by altering the
genes of the parents. In (BLS06), the authors adopted
the method of creating two new children by inherit-
ing fragments from parents with a 50 % probability
where the division points of the parents depend on K
i
for each task τ
i
. For more details, see (BLS06).
In our work, we adopt another method in order
to have a better solution from generation to another.
Indeed, for each pair of parents (chromosomes), two
crossover integers i
1
and i
2
are randomly generated
with the condition 1≤ i
1
, i
2
≺ h where h is the num-
ber of genes composing the parent chromosome, be-
ing scheduled. In the random position i
1
, the first chil-
dren (child1) inherits i
2
genes from parent
1
(P
1
) and
the second children (child2) inherits i
2
genes from
parent
2
(P
2
). Then, child1 inherits the remaining
genes of P
2
and similarly, child2 inherits the remain-
ing genes of P
1
. Finally, the genes must be ordered
AHybridGeneticbasedApproachforReal-timeReconfigurableSchedulingofOSTasksinUniprocessorEmbedded
Systems
387

in the two new offsprings (childs) in order to avoid
violating constraints.
Now, let us consider the same example of three
aperiodic tasks used by (BLS06) to more explain our
crossover operator principle. Tasks are τ
1
(min
τ
1
=
100), τ
2
(min
τ
2
= 150) and τ
3
(min
τ
3
= 200) like de-
scribed in table 1. In this example, each parent was
composed of five genes. The random running of the
two integers i
1
and i
2
was done and we have i
1
= 2
and i
2
= 3 with the condition 1≤ i
1
, i
2
≺ 5. Then, at
the position 2, the crossover operator copy 3 genes (2,
3, 4) from the parent 1 for child1 and and 3 genes (2,
3, 4) from the parent 2 for child2. Then, child1 inher-
its the first and the fifth genes from parent 2, and the
second children child2 inherits the first and the fifth
genes from parent 1. Table 1 shows an example of a
chromosome composed of five genes.
Table 1: Crossover Operator.
Chromosome
Name
Task τ
1
Task τ
2
Task τ
3
Parent
1
(P
1
) (τ
1
, 25) (τ
1
,
150)
(τ
2
, -1) (τ
2
,
150)
(τ
3
, 0)
Parent
2
(P
2
) (τ
1
, 5) (τ
1
,
200)
(τ
2
, 50) (τ
2
,
200)
(τ
3
, 55)
Children
1
(Child
1
)
(τ
1
, 5) (τ
1
,
150)
(τ
2
, -1) (τ
2
,
150)
(τ
3
, 55)
Children
2
(Child
2
)
(τ
1
, 25) (τ
1
,
200)
(τ
2
, 50) (τ
2
,
200)
(τ
3
, 0)
3.5 Tabu Search Algorithm
In GAs, there is a mutation operator which is occurred
when a gene is randomly chosen and mutated, and the
resulting chromosome is evaluated for its new fitness
function value. In (BLS06), the authors define a mu-
tation operator as an operator that mutates genes in
a chromosome by altering their arrival times. This
work is limited in the sense that every mutated gene
must be inserted in the best place and insertions occur
from the left to right along the executions of a task
and if no suitable insertion location is found, then no
task execution can be added among the already ex-
isting task execution and this gene must be rejected.
This result is strong in time and space and especially
in worst case when many genes are rejected.
In our work, we will replace the mutation opera-
tor by the tabu search (TS) algorithm like below: The
tabu search solves the problem and returns the best
solution found. Indeed, this algorithm replaced the
mutation operator and evaluated in randomly gener-
ated chromosomes and returned a good results which
proved our approach. To more explain this main
idea of this TS algorithm, let us consider a chromo-
some represented by a set of n genes. We adopt the
change of one gene by another of the same task as an
(filled up) elementary transformation and we evalu-
ated the Fitness-Function (fit) for the chromosomeun-
der study after each alteration of the selected gene. In
fact, we select a gene of the chromosome and we will
be interested by the genes before and after it. For each
gene of its neighbors (gene before and gene after), we
alter its arrival time. A new arrival time a
′
i, j
is chosen
for it from the range [a
i, j−1
+ min
i
, a
i, j−1
+ max
i
] by
the dichotomic algorithm to reduce the search space
and we evaluate the fitness function after each tabu it-
eration in the hyper-period. The same procedure was
executed for the gene before if the selected gene is not
the first one on the chromosome and for the gene after
if the selected gene is not the last one on the chromo-
some also. After completion of this procedure, the
new gene leads to the best fitness function in the cur-
rent chromosome was chosen to replace the previous
one and the worst individual characterized by the least
fitness in the current chromosome will be replaced by
the offspring (child) and this gene of the best solution
is added to the tabu list.
Let us consider the following example, we assume
that the gene 2 of the task τ
1
(child1) is randomly se-
lected in the tabu search algorithm.
The chromosome is: (τ
1
, 11) (τ
1
, 61) (τ
1
, 120) (τ
2
,
3) (τ
2
, 90) (τ
3
, 0) and the selected gene for the tabu
search procedure is the second gene of the first task,
i.e., (τ
1
, 61). The research space is composed by the
first and the third genes, i.e., (τ
1
, 11) and (τ
1
, 120).
For the gene (τ
1
, 11), the range is [0 + 50, 0 + 100] =
[50, 100] (min = 50 and max = 100).
For the gene (τ
1
, 120), the range is [61+ 50, 61+
100] = [111, 161]. Now, we will explore the first solu-
tion space represented by the first range and the sec-
ond one by the dichotomic research method in order
to get the best value that optimizes the fitness func-
tion. The new value a
′
i, j
will replace the previous ex-
ecution time on the gene (τ
1
, 61). After calculate the
new value is equal to 70 and we have the new gene
(τ
1
, 70) and the chromosome becomes (τ
1
, 11) (τ
1
,
70) (τ
1
, 120) (τ
2
, 3) (τ
2
, 90) (τ
3
, 0). This procedure
is repeated until the changes leads to a valid chromo-
some.
3.6 Fitness Function
In our work based on a genetic algorithm, the eval-
uation of a model is obligatory. This evaluation is
represented by a fitness function (objective function)
that we noted fit. Indeed, fit(ch(τ)) =
∑
K
τ
j=1
R
τ
, where
R
τ
= d
τ, j
− f
τ, j
. By this equation we have fit(ch(τ)) =
∑
K
τ
j=1
d
τ, j
− f
τ, j
for target task τ. By considering the
difference between the deadline of an execution and
ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems
388

the execution’s actual completion, i.e., d
τ, j
− f
τ, j
. We
are thus interested in all periodic, sporadic and aperi-
odic tasks in a task set ξ
h
=
∑
h
i=1
τ
i
which is composed
by h tasks among m + n tasks. So, we have fit(ξ
h
) =
∑
h
i=1
fit(ch(τ
i
)) =
∑
h
i=1
∑
K
τ
j=1
(d
τ
i, j
− f
τ
i, j
). Here our
system will be feasible if R
τ
= d
τ, j
− f
τ, j
≥ 0 and the
solution is optimal if this fitness function value con-
verges on zero.
4 EXPERIMENTATION RESULTS
AND DISCUSSIONS
This section is intended for presenting the simula-
tions results, the discussions and assumptions about
the presented work. The advantages and disadvan-
tages about them are discussed.
4.1 Simulations Results
In a real-worldoptimization problem, there often exist
multiple benchmarks to be compared with test results.
In this section, we will restraint to the benchmarks
presented by (DZ13). We present results obtained
with a real-coded genetic algorithm with a population
size of 20 and a generation number of 50 in order to
check reliability of a solution and feasibility of a task
set model. The proposed GA uses the crossover op-
erator and the TS algorithm presented in the above
section. First, the proposed GA begins by the param-
eters of tasks presented in the benchmarks and then
iterates to converge on a small particular fitness func-
tion. When this fitness function value converges on
zero, we terminate the research and we consider this
solution as optimal. The following table 2 shows the
simulations results presented by our approach.
Table 2: Simulations Results.
Approach
Name
Worst Average Best
Hybrid Ap-
proach
1.020*10
−8
1.017*10
−8
1.010*10
−8
Classical
Approach
1.028*10
−8
1.090*10
−8
1.021*10
−8
4.2 Discussion
We perform 1000 runs with different task sets of the
benchmark values. The first row in table 2 shows the
performance of the proposed hybrid GA. We observe
that in all 1000 runs, the GA is able to find a solution
near the optimal feasible solution (10
−8
). The sec-
ond row in table 2 shows the results of the classical
GA, i.e., crossover and mutation operators presented
by (BLS06).
Now, as we show in table 2, if we want to com-
pare the first hybrid method and the second classical
method, we observe the performance of our proposed
solution.
We conclude that, our proposed GA can find the
correct reliable near optimal solution compared with
the classical GA. In contrast, the classical GA is faster
than our approach due to lateness obtained by the TS
algorithm when the number of populations is so high.
Finally, we can say that genetic based algorithms
can be computationally expensive in time but effi-
cient, important and practice to difficult feasibility-
based optimization problems.
5 CONCLUSION AND FUTURE
WORK
In this paper we have proposed a real-time hybrid ge-
netic based scheduling approach for solving the em-
bedded systems feasibility problem. A specific fitness
function fit is designed to guide the search. Computa-
tional results show that our approach is very promis-
ing. The approach was tested and the results indicate
that performance of our original algorithm is better
than others. Comparison with classical genetic ap-
proach demonstrates the effectivenessof the proposed
rescheduling approach. Since we were unable to find
a counter example for which the approach fails, we
conjecture that it always finds an optimal solutions.
We plan to further strengthen this work to include
distributed embedded systems problems.
ACKNOWLEDGEMENTS
The authors thank anonymous reviewers whose com-
ments clarified and enriched the contents of this paper.
REFERENCES
Y. Al-Safi and V. Vyatkin. An ontology-based reconfigu-
ration agent for intelligent mechatronic systems. Int.
Conf. Hol. Multi-Agent Syst. Manuf., vol. 4659, pp.
114-126, Regensburg, Germany, 4th edition, 2007.
Lionel C. Briand, Yvan Labiche, and Marwa Shousha. Us-
ing genetic algorithms for early schedulability anal-
ysis and stress testing in real-time systems. Genetic
Programming and Evolvable Machines, 7(2):145–
170, 2006.
AHybridGeneticbasedApproachforReal-timeReconfigurableSchedulingofOSTasksinUniprocessorEmbedded
Systems
389

G. Buttazzo and J. Stankovic. RED: Robust Earliest Dead-
line Scheduling. Int. Workshop On Responsive Com-
puting Systems, Austin, 3rd edition, 1993.
Ballester R. Ripoll L. Brocal V., BalbastreP. Task period se-
lection to minimize hyperperiod, Emerging Technolo-
gies & Factory Automation (ETFA), IEEE conference
on, pp.1-4, 2011. doi: 10.1109/ETFA.2011.6059178,
Toulouse, France, 16th edition, 2011.
K. Sierszecki C. Angelov and N. Marian. Design models
for reusable and reconfigurable state machines. L.T.
Yang et al., Eds., Proc. of Embedded Ubiquitous Com-
put., 2005.
M. Dertouzos. Control Robotics: The Procedural Con-
trol of Physical Processes. Proceedings of the IFIP
Congress, 1974.
Stephen C. H. Leung Qingshan Chen Defu Zhang, Li-
jun Wei. A binary search heuristic algorithm based
on randomized local search for the rectangular strip
packing problem. INFORMS Journal on Computing,
25(2):332–345, 2013.
Eiben A. E., Raue P. E., and Ruttkay Z. Solving con-
straint satisfaction problems using genetic algorithms.
In IEEE World Conference on Evolutionary Comput-
ing, pages 542–547, 1994.
M.Khalgui H. Gharsellaoui and S.BenAhmed. Feasible Au-
tomatic Reconfigurations of Real-Time OS Tasks. IGI-
Global Knowledge, USA, isbn13: 9781466602946
edition, 2012.
Gharsellaoui H., Hasni H., and Ben Ahmed S. Real-time re-
configurable scheduling using genetic algorithms. In
Genetic and Evolutionary Computation Conference,
GECCO, Vancouver, BC, Canada, 2014.
Haupt R. L. and Haupt S. E. Practical Genetic Algorithms.
Wiley-Interscience, 1998.
M.A.Pawlowsky. Crossover operators. In Practical Hand-
book of Genetic Algorithms Applications.
T. Strasser A. Zoitl O. Hummer M. N. Rooker, C. Sub-
der and G. Ebenhofer. Zero downtime reconfiguration
of distributed automation systems: The CEDAC ap-
proach. Int. Conf. Indust. Appl. Holonic Multi-Agent
Syst., Regensburg, 3rd edition, 2007.
I. Ripoll P. Balbastre and A. Crespo. Schedulability analysis
of window-constrained execution time tasks for real-
time control. 14th Euromicro Conf. Real- Time Syst.,
14th edition, 2002.
B. Mishra A. Raghunathan L. Rosier S. Baruah, G. Koren
and D. Shasha. On-line Scheduling in the Presence of
Overload. IEEE Symposium on Foundations of Com-
puter Science, San Juan, Puerto Rico, 1991.
R. West and K. Schwan. Dynamic window-constrained
scheduling for multimedia applications. IEEE 6th Int.
Conf. Multi. Comput. Syst., 6th edition, 1999.
ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems
390