8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia

“Evaluation and optimization of innovative production systems of goods and services”

VERIFYING MOBILE AGENT BASED WORKFLOWS

IN REWRITING LOGIC

Faiza Bouchoul* and Mohammed Mostefai**

Laboratory of Automatics of Sétif (L.A.S)

Ferhat Abbas University
Sétif Algeria
*faizabouchoul@univ-setif.dz, **mostefai@univ-setif.dz

ABSTRACT: Traditionally, Workflow management systems (WfMS) support static processes which do not change
frequently. Modern workflows systems require a lot of flexibility i.e. the capability to react to changes in the workflow
during its execution and a great degree of adaptivity. The agent technology seems to be a good candidate to these
requirements. One solution is that the workflow enactment is done by a mobile agent. These systems are so complex
such as the usage of formal tools for verification, simulation and prototyping to facilitate their design and their
validation is essential and of great interest. In general, improved methods are needed to insure reliability, security and
robustness of different components of automated business processes systems such as workflows. The aim of this paper
is twofold: firstly we discuss the importance of applying agent technology to workflows and then a new formal method
based on rewriting logic is proposed to verify mobile agent based workflows and to enable their rapid prototyping.

KEYWORDS: flexible workflows, mobile agents, formal verification, rapid prototyping, rewriting logic, Maude.

1 INTRODUCTION The paper is composed of three main sections, section 1

The enterprise activities are generally governed by a
business process. A business process can be seen as a
flow of related activities that together create a service or
a product. A workflow is an automation of the business
process. Traditionally, Worflow management systems
(WfMS) are used to support static processes, i.e.,
processes which do not change frequently. When a
workflow is assumed to act in mobile environments, it is
said a mobile workflow. Mobile workflows extend the
main business process of the enterprise, since the latter
incorporates the functionalities of the first within its own
processes out of the traditional boundaries of the
organization. The main restrictions of mobile devices for
the execution of workflow tasks are essentially limited
bandwidth, limited resources such as CPU power and
memory and limited energy supply; so that and in most
cases the performances of mobile applications is greatly
restricted due to architectural and design considerations
and thus very subject to frequent failures. In such
environments, mobile workflows not only inherit the
wireless technologies weakness but also are often
Internet-based so that mobile users periodically become
unavailable due to the lack of network service
guarantees. The result is limited or very difficult
business information access and activity coordination.
The obvious and general consequence is that traditional
centralized WfMS have not been designed for dynamic
environments requiring adaptive situation induced by
new business requirements and sometimes by mobility.
will show how multi-agents technology can offer good
solutions for modern workflow systems. Since these
systems are so complex formal verifications seem in our
sense essential; in section 2 and 3 rewriting logic is
used to specify and verify in a very simple manner
mobile agent based workflow enactment.
2 AGENTS BASED WORKFLOWS.
The agent technology seems to be a good candidate to
handle efficiently modern workflow systems
requirements. The multi-agent system has some
common characteristics, such as distribution, autonomy,
interaction and openness, which are helpful to transform
traditional architectures into a distributed and
cooperative architecture in an intelligent manufacturing
system ( Guo and Zhang 2009).
In practice an agent-based workflow is a workflow in
which agents perform, coordinate, and support the whole
workflow or parts of the workflow. An agent-based
WfMS is a set of software agents that manage and
coordinate the flow of work defined by a business
process (Odgers et al. 1999). Using agents to perform
WfMS functions is an emerging field where researchers
are exploring the ability of agents to improve process
integration, interoperability, reusability and adaptability,
some examples of works proposing agents solutions for
workflows issues are (Stormer 2000); (Huang et al.
2006); (Trappey et al. 2004); (Kuo 2007); (Xu and
Wang 2002); (Zhuge 2003); (Madhusudan 2005) and
(Filho et al. 2003).
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”
In Bouchoul and Mostefai (2007) and Bouchoul and
Mostefai (2009a) MOBIFLEX a generic architecture for
mobile workflows is proposed. This architecture
integrates mobile agents, static agents with intelligent
capabilities and Web-Services, since these new
technologies when put together can generate a flexible,
reconfigurable, adaptive and integrable framework to
fulfil easily requirements of mobile workflows. The
architecture is empowered by a fault-tolerance
mechanism and validated analytically through a
performance analysis which proves that mobile agent
based workflows outperform centralized ones in time
consuming and network load (Bouchoul and Mostefai
2009b). This architecture is implemented on
JadeLeap/Jadex platform. As shown in Figure. 1, the
architecture proposed is composed of three types of
agents.
Figure 1. MOBIFLEX architecture.
A workflow manager (WF-Manager): the WF-Manager
is the core component of the mobile workflow engine on
the mobile device. The WF-Manager in this scenario
takes full responsibilities of an WfMS, which means
that it has possibilities to enact and oversees its
workflow.
An agent planner (PL-Agent): It is an intelligent agent
which resides on the static server. IA mechanisms and
performances metrics can be used by the agent planner
to choose the most appropriate and most coherent
sequences of Web services composing the workflow to
be enacted. When the WF-Manager is initiated by the
user by a specific request, it sends the request to the PL-
Agent that returns details of the workflow susceptible to
achieve the goal.
A mobile worker agent (M-Worker): the M-Worker is a
mobile agent that enacts effectively the workflow. To
enact the workflow the WF-Manager sends an
M-Worker to execute it. The agent carries with him an
itinerary which offloads the execution sequence for tasks
The WF-Manager can launch multiple mobile agents for
a given workflow which are not dependent upon each
other and can, therefore, be executed in parallel. As an
applicative area for MOBIFLEX, the healthcare domain
was chosen (Bouchoul and Mostefai 2009b), healthcare
information systems (HCIS) are complex,
heterogeneous, and naturally mobile.
3 THEORETICAL BACKGROUND : THE
REWRITING LOGIC.
The rewriting logic (Meseguer 1992) is chosen as the
basis of the formal framework proposed here since this
logic is a very a powerful unifying paradigm of most of
formal models of concurrency. Rewriting logic is a
computational logic that can be efficiently implemented
and that has good properties as a general and flexible
logical and semantic framework, in which a wide range
of logics and models of computation can be represented.
In particular, for programming language semantics
(Şerbănuţă et al. 2009).
3.1 Rewriting logic semantics.
Although the rules of rewriting logic (Meseguer 1992)
(Meseguer and Rosu 2007) resemble those of equational
logic, their meaning is very different; rewriting logic is a
logic to reason about changes in a concurrent system, not
about equalities. Each rewrite rule is a general pattern for
a basic action that can occur concurrently with other
actions in a concurrent system. The models of rewriting
logic are precisely concurrent systems whose state is
distributed and can change by actions taking place
simultaneously.
The rewriting logic is nothing but a generalization of
equational logic in order to adapt it to changes
(Meseguer 1992). The rules are similar to those of
equational logic but have a completely different
significance. A rule T  T’ do not mean any more T
equal T’ but T becomes T’. The rule is a basic action
allowing the transition of the system from one state to
another. The rewriting logic describes the changes of the
system so that the state is represented by an algebraic
term, the transition becomes a rewriting rule and the
distributed structure, an algebraic structure modulo a set
of axioms E. Syntax in logic of rewriting is given by a
signature ( ,E) where  is a set of functions and E a
set of axioms. A rewriting theory T=(Σ, E, L, R) in
rewriting logic is composed of a signature (,E) and by
a set of labelled rules R with labels in L. These rules
describe the behaviour of the system and the rewritings
are performed on the classes of equivalences of the
terms modulo the axioms E.
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”

In practice a theory of rewriting T=(Σ, E, L, R) can be
used as an executable specification allowing a rapid
prototyping of the modelled system and its checking.
The states of the system are specified as algebraic data
types, the basic changes which may occur in the system
and in parallel are specified by rewriting rules, the set of
axioms E capture the structural properties of the system,
a possible state of the system is represented by an
equivalence class  t  of a term t
modulo the structural
axioms E .
Given a rewrite theory R, we say that R entails a
sequent  t    t  and write
R   t    t  if and only if  t    t  can be obtained
by finite application of the following rules of deduction :
Reflexivity.
  t   T , E ( X )
This rule has no meaning in rewriting logic because
changes in time are irreversible.
3.2 The MAUDE Language.
MAUDE (Duran and Meseguer 2007) is a fully reflective
programming language and development environment
based on rewriting logic and its equational logic
sublanguage to specify formal executable environments.
The reflective nature of rewriting logic allows MAUDE
programs to define their own syntax, operations, and
data types to model the behaviour of concurrent systems.
Full MAUDE is an object-oriented formal language
derived from MAUDE. In Full MAUDE a class defines
the structure of an object, and objects are specific
instances of a class.
A class consists of a class identifier (Cid), which is a
sort, and a list of attributes that are sorts, including class
or object identifiers. Classes also support multiple
inheritance with subclasses that inherit all the attributes
of parent classes.

t   t  class C | attribute1: Sort1, … , attributen: Sortn .

< Oid_Name : C | attribute1: variable1, … , attributen:
Congruence. variablen >
msg syntax : Oid Sort1 Sortn -> Msg .
For each f (function symbol)  n n IN
Each object has an object identifier (Oid), a class
identifier, and a list of attributes. Each message type has
 t1    t '1 ... t n    t 'n  a name and a list of arguments.
The configuration is the distributed state of the
 f (t 1,...,t n )   f (t '1,...,t 'n ) concurrent object-oriented system and is represented as a
multiset of objects and messages according to the
following syntax:
Replacement.
subsorts Object Message < Configuration .
For each rewrite rule op -- : Configuration Configuration => Configuration
(assoc comm id: )
r :  t ( x )    t ' ( x ) we have
Where the operator -- is associative and commutative
 w1    w1  ...  wn    wn  with identity . The system evolves by concurrent
rewriting (modulo ACI) of the configuration by means of
 t ( w / x)   t ' ( w  / x) rewrite rules specific to each particular system, whose
left-hand and right-hand sides may in general involve
patterns for several objects and messages.
Rewrite rules in Full MAUDE transition the system from
Transitivity. a configuration of objects and messages to a new
configuration of objects and messages, so that they
 t1    t 2   t 2   t 3  define all the possible transitions for the concurrent
system.
 t1    t3  In MAUDE, the general form required of rewrite rules
used to specify the behaviour of an object-oriented
Notice that equational logic is obtained by adding the system is as follows:
next symmetry rule
  m1 ,..., m n < O1: C1 / atts1 > ... < O m:

 t    t  Cm / attsm > =>

 t    t  < Oi1: Ci1 / attsi1 > ... < Oik: Cik / attsik >
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”
"
< Q1: D1 / atts1 > … < Op: Dp /
atts"
p > m1' ,..., m 'q if C
where the m1,…,mn are message expressions,
O1,…,On ; Oi1, … ,Oik and Q1,…, Qn are objects such as
{ Oi1, … ,Oik }  { O1,…,On  }
C is the rule’s condition. A rule of this kind expresses a
communication event in which n messages and m
distinct objects participate. The outcome of such an
event is as follows:
 The messages m1,. . . , mn disappear;
 The state and possibly even the class of the
objects Oi1, … ,Oik may change;
 All other objects vanish;
 New objects Q1,…, Qn are created ;
m1' ,..., m'q
 New messages are sent.
4 VERIFYING MOBILE AGENT BASED
WORKFLOW ENACTMENT IN REWRITING
LOGIC.
This section tries to propose a formal framework for
specifying and verifying mobile agents based workflows,
where the enactment is performed by mobile agents.
MOBIFLEX as shown above is a typical example of such
systems, the workflow enactment can be seen as a
complex combination of mobile agents moving over a
set of nodes where reside Web-Services, these Web-
Services can be simple ones or composite ones i.e. BPEL
processes. Figure 2 shows the two layers of MOBIFLEX
architecture, a static service oriented layer (called here
the SO_Layer) is the Web-Services platform of the
enterprise, and a dynamic layer obtained by a
reconfigurable and very flexible workflow enactment
patterns to be performed by mobile agents over the
static layer (called the MA_Layer).
Figure 2. The two abstract layers of MOBIFLEX
architecture.
The workflow enactment in MOBIFLEX is performed
by an M-Worker which is a mobile worker. In this work
a particular interest is given to the behaviour of the
mobile agent when enacting the workflow; i.e. when
pursuing its itinerary.
4.1 related works.
In literature, only few works dealing with formal
specification of mobile agents migrations can be found.
In Ling and Loke (2002) the authors discuss how
mobile agent enabled interorganizational workflows can
be usefully modelled using advanced Petri Net
techniques such as Interorganizational Workflow Nets.
This model provides a means to verify the correctness
(and, so the viability) of the itineraries of agents used in
enacting interorganizational workflows. the authors used
the a special class of Petri nets, called Workflow nets
(WF-nets) (Aalst 1997), workflow concepts are
modelled by Petri net elements (places and transitions).
In Ferreira et al. (2004) the paper presents results of a
performance comparative study of the three mobile
agents design patterns presented in Ferreira et al. (2003).
The Itinerary, Star-Shaped and Branching migration
patterns were investigated as solutions to a distributed
information retrieval system. The itinerary assumes the
migration of the agent over a sequence of hosts to do a
given job and then return back to the source agency. In
the Star-Shaped pattern, the agent migrates to the first
destination agency in the list, executes the relevant job
and comes back to the source agency, the agent repeats
this cycle until the last agency on its list is visited. In the
Branching pattern, the agent clones itself according to
the numbers of agencies in the defined itinerary, each
clone migrates to one agency, executes its job and
notifies the source agency when the job is completed.
The three solutions were modelled with timed Coloured
Petri Nets.
In Satoh (2004), the paper an agent itinerary
specification language is defined, the specification
language is basically inherited from those of existing
process algebras, e.g., CCS and π-calculus, because,
according to the authors, they provide well-studied
foundations.
In Lu and Xu (2005) an other itinerary language, MAIL,
is introduced to model the mobile behaviour of
proactive agents. The language is structured and
compositional so that an itinerary can be constructed
recursively from primitive itineraries. Special constructs
to support the specifications of different kinds of
itineraries are provided: sequential, conditional,
concurrent, and loop itineraries. The operational
semantics of the language is defined in terms of a set of
inference rules. MAIL is amenable to formal methods to
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”

reason about mobility and verify correctness and safety

properties. Example the next graph (figure 3), shows in an abstract
manner a complex itinerary built by composing atomic
In Yang et al. (2006) The authors present a new itineraries by the constructs listed above.
approach to ensure the secure execution of itinerary-
driven mobile agents, in which the specification of the This itinerary can be described by the next algebraic
navigational behaviour of an agent is separated from the expression
specification of its computational behaviour. Each host   Aa 3  Asa55  
Asnil0  Asa11  Asa22   sa34 
is empowered with an access control policy so that the   A   Aa 6  Aa 7   Aa 8  
host will deny the access from an agent whose itinerary  s4 s6 s7 s8 

does not conform to the host’s access control policy. A

host uses model checking algorithms to check if the
itinerary of the agent conforms to its access control
policy written in μ-calculus.

In Peschanski et al. (2007) the goal was to provide both

a specification language to model and support reasoning
about the coordination of mobile agent systems, and an
executable language to control their deployment and
operation at runtime. Agents can migrate and thus “hop”
from host to host in order to perform distributed
computations. This is modelled through the π-calculus
notion of channel mobility.

4.2 An Algebra of itineraries for the MA_Layer.

This proposition is based on the algebra of itineraries

proposed in Loke et al. (1999) modified and enhanced
with an operational semantics (next paragraph) in term
of rewriting logic rules.

Let {A{i}*} where {i}* denotes strings of integers

identifying an agent or a clone of an agent be a finite set
of mobile agents. For generating names for agents and
their clones the following assumption is done: When an
agent A{i} is cloned k times, its clones are named A{i}1, …
, A{i}k.
Figure 3. An example of an M-Worker itinerary.
Let A ={a1,a2 , … ,an} be a finite set of activities and S
={s1,s2 , … ,sn} a finite set of sites to be visited.
Itineraries (denoted by I) are formed as follows
representing the null activity, atomic activity, parallel, 4.3 Formal patterns for Mobile agent based
sequential, nondeterministic, conditional workflows.
nondeterministic behaviour, and have the following
syntax:  An approach that can simplify and improve the
 The null activity :  development of workflows based on mobile agents are
a the use of design patterns similar to classical workflow
 The atomic activity, As which means the
patterns (Aalst et al 2003 ). The use of design patterns is
agent A moves to site s and executes the action a an approach to improve the development process of
 The parallel behaviour : I  I’  which means applications and the quality of the final products. Some
the itinerary I in parallel with the itinerary I’  mobile agent design patterns have already been proposed
 Sequential behaviour : I  I’ which means although for other purposes than workflows : (Aridor
itineray I then the itinerary I’  and Lange 1998); (Kendal et al. 1998); (Tahara and
 Nondeterministic behaviour : I  I’  which Ohsuga 2001); (Ojha et al. 2007); (Al-Shrouf 2008) and
means I or I’  (Ferreira et al. 2003).
 Conditional nondeterministic behaviour :
I I ' In Loke et al. (1999) it is shown that itineraries are
which means if condition c is true then associative with an element identity , so it is easy to
c
I else I’ , c is a Boolean expression. conclude that itineraries are a special case of strings
rewriting theory modulo a set of axioms E=AI. This
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”
section proposes a rewriting logic based operational
semantics in term of MAUDE rules. The configuration is
in this case a set of agents and messages, messages will
be used to mark important steps of the workflow such as
the achievement of activities.
Additionally two sets (Site) for places to be visited by
the M-worker and (Activity) for the activities to be
performed by the M-Worker are assumed to be
predefined. A class M-Worker with two attributes :
'site' and 'activity' is also defined. A mobile agent will be
specified by the term < A:MWorker s:s0 , a:a0>
which at the right-side of a rule means : The M-Worker
A moves to the site s0 to perform the activity a0, and at
the left-side of a rule it means that the activity a0 is
already performed by the agent A in the side s0. The
activity a0 is essentially built upon the invocation of a
Web-process. Since the aim of the study is the
behavioural aspect of the workflow enactment and for
the sake of simplicity, the state of the agent is not taken
into account. It can be modelled in MAUDE easily by a
set of additional attributes.
A workflow instance enactment is performed by one
agent and starts at a source site and ends at a target
site, at intermediary steps, the agent may clone itself to
perform some parallel tasks so that the configuration
may sometimes involve more than one agent. After
achieving parallel tasks, clones have to declone
themselves so that the workflow terminate with only one
agent which is exactly the original one who launches the
workflow at the source site. In what follows the different
itinerary constructs are modelled as MAUDE rules.
4.3.1 Agent basic Movement.
Different cases have to be envisaged for agent
movements:
The agent is in site s1 with no activities to perform here
and has to move to site s2 to perform activity a1
<A: MWorker |s:s1, a: nil>=> <A: MWorker|s: s2,
a:a1>
After achieving activity a1 in site s1, the agent moves to
site s2 to perform activity a2
<A: MWorker| s1:site, a1: activity>=> <A: MWorker|
s2:site, a2: activity>
The agent have to perform sequentially two activities a1
then a2 in the same site s1
<A: MWorker| s1:site, a1: activity>=> <A: MWorker|
s1:site, a2: activity>
4.3.2 Sequential behaviour.
The basic rule presented above can be used to specify a
sequence of more than one activity, so that for n steps (n
sequential activities), n rules are necessary, for example
an agent A is in a site s0 and have to perform
sequentially 3 activities (a1,a2, a3) in three sites
respectively s1,s2 and s3 ; this can be expressed as
follows: As 0  As1  As 2  As 3
nil a1 a2 a3
This sequence can be modelled in MAUDE like this
<A: MWorker| s:s0, a: nil>=>
<A: MWorker|s: s1, a: a1>
<A: MWorker| s:s1, a: a1>=>
<A: MWorker|s: s2, a: a2>
<A: MWorker| s:s2, a: a2>
 <A: MWorker|s: s3, a: a3>
4.3.3 Parallel composition.
Two itineraries composed by "" are executed in
parallel. As1  As 2 means that the agent A has to
a1 a2
perform activities a1 in site s1 and a2 in site s2
concurrently. Parallelism imply cloning of agents. Since
agent A has to perform actions at both s1 and s2 in
parallel. The agent A is assumed in site s0 and have to
perform the itinerary As 0  ( As1  As 2 )
nil a1 a2
The MAUDE rules for this itinerary are as follows :
<A: MWorker| s:s0, a: nil> =>
<A1: MWorker|s:s1, a:a1><A2: MWorker|s:s2,a: a2>
<A1: MWorker| s:s1, a:a1><A2: MWorker| s:s2,a: a2>
=> <A: MWorker| s:s0, a:nil>
The second rule is the decloning rule , it may be used in
different manners and to represent different situations,
the clones can before decloning themselves perform
complex itineraries, when these latter achieved they
declone themselves in a site which can one of the two
sites from which the clones reside or a third site where
the rebuilt agent can continue its itinerary.
4.3.4 Nondeterministic behaviour.
 
In this proposition the nondeterminism presented in
Loke et al. (1999) is extended to more than two
choices, since MAUDE has an interleaving semantics,
we can specify nondeterministic behaviour in a very
natural manner, as example the next itinerary is
0  ( As1 As 2 As 3 ) .
assumed: Asnil a1 a2 a3
The MAUDE specification can be like this
<A:MWorker|s:s0,a:nil>=><A1: MWorker| s:s1, a:a1>
<A: MWorker|s:s0,a:nil>=><A1:MWorker| s:s2, a:a2>
<A: MWorker|s:s0,a:nil>=><A1:MWorker| s:s3, a:a3>
Executing any rule of these three rules in a
nondeterministic manner will disable automatically the
two others.
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”

4.3.5 Conditional nondeterministic behaviour.

  In figure 6 we tried with a wrong specification (see
A conditional nondeterministic choice between two figure 5 for the itinerary graph ) where a decloning of
activities a1 and a2 can be for example expressed like two clones is omitted, the workflow is launched by an
this : M-Worker 'Sim' but at the last stage not only some
activities are not performed but we end the workflow
 A a1 Asa22  with two clones 'Sim1' and 'Sim2' rather than 'Sim'
Asnil0   s1  alone which is the initiator of the workflow.
 c 

Since MAUDE rules are conditional this expression can

be specified easily as follows:

<A: MWorker| s:s0, a: nil> =>

<A1: MWorker| s:s1, a:a1> if c
<A: MWorker| s:s0, a: nil> =>
<A1: MWorker| s:s2, a:a2> if not
c

4.4 Verifying in MAUDE.

Based on the operational semantics proposed above, the

Figure 5. A wrong itinerary for an M-Worker.
workflow of the example presented in figure 3 can be
verified in MAUDE, we used Maude for windows under
Eclipse, next listing (figure 4) is the source of the
specification in MAUDE as an object oriented module of
this workflow. MAUDE offers a number of commands
for verifying properties of the modelled system, the most
powerful are rewrite command and search command,
rewrite command causes a specified term to be rewritten
using the rules and equations, in the given module. In
this example we used the rewrite command to verify that
an enactment of the workflow of the previous example
by a mobile agent "Mos" is correct i.e. the correct
activities are performed in the right order and at the final
stage we have the original M-Worker that launched the
enactment at the home site (s0).

Figure 6. Verification in MAUDE of a wrong workflow

enactment.

REFERENCES

Aalst Van der, W.M.P. Ter Hofstede, A.H.M.

Kiepuszewski, B. Barros, A.P. 2003.
Workflow Patterns. Distributed and Parallel
Databases, 14(3), 5-51.
Aalst Van der, W.M.P. 1997. Verification of workflow
nets. In Proceedings of 18th International
Conference on Application and Theory of Petri
Nets. LNCS 1248, Springer-Verlag. 407–26
Figure 4. Verification in MAUDE of a mobile agent
Al-Shrouf, F. M. Facilitator Agent Design Pattern of
based workflow. Procurement Business Systems. 2008. 32nd
Annual IEEE International Computer Software
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”
and Applications Conference, compsac, 505-
510.
Aridor Y. and Lange. D.B. Agent design patterns:
Elements of agent application design. 1998 In
Proceedings of the Second International
Conference on Autonomous Agents. ACM
Press.
Berry, G. and Boudol G. 1992: The Chemical Abstract
Machine . Theoretical Computer Science .
96(1), 217-248.
Bouchoul, F. and Mostefai, M. 2007. MOBIFLEX:
A Generic Architecture for Integrated M-Health
Applications. in Proc. of IEEE: First
international conference on E-medical systems:
E-medisys’07 October 2007, FEZ Morocco
Bouchoul, F. and Mostefai, M. 2009a. Agent-services
and mobile agents for an integrated HCIS.
International Journal of Computer Integrated
Manufacturing (IJCIM) Francis and Taylor
editor. 22(5) 458-471
Bouchoul, F. and Mostefai, M. 2009b. Validating
MOBIFLEX: A Performance Analysis Model.
to be presented in IEEE: International
conference on applied informatics ICAI'09
Bordj Bou Arréridj Algeria. November 15-17
2009.
Chandy K. M. and Misra J. 1988. Parallel Program
Design: a Foundation. Addison-Wesley, 1988.
Department of Computer Science, Washington
University, March 1988
Duran F. and Meseguer J. 2007. MAUDE’s module
algebra. Science of Computer Programming.
66(2), 125-153.
Ferreira de Araújo Lima, E. Duarte de Lima Machado, P.
Abrantes de Figueiredo J. C. and Ronison
Sampaio, F. 2003. Implementing Mobile
Agent Design Patterns in the JADE framework.
In Special Issue on JADE of the TILAB Journal
EXP.   Odgers, B. R. Shepherdson J. W. and S. G. Thompson.
Ferreira de Araujo Lima, E. de Figueiredo, J. C. A.
Guerrero, D. D. S. 2004. Using Coloured Petri
Nets to Compare Mobile Agent Design
Patterns. Electronic Notes in Theoretical
Computer Science 95 287–305.
Filho, R. S. Wainer, J. and Madeira, E. R. M. 2003. A
Fully Distributed Architecture for Large Scale
Workflow Enactment. International Journal of
Cooperative Information Systems. 12(4),
411-440.
Guo, Q. and Zhang, M. 2009. A novel approach for
Multi-Agent-Based Intelligent Manufacturing
System. Information Sciences. 179(18), 3079-
3090. published online may 2009
Huang, C. J. Trappey A. J. C. and Yao, Y. H. 2006.
Developing an agent-based workflow
management system for collaborative product
design. Industrial Management and Data
Systems. 106(5) 680–699.
Jennings, N. R. Norman, T. J. Faratin, P. O’Brien, P.
Odger, B. 2000. Autonomous Agents for
Business Process Management. International
Journal of Applied Artificial Intelligence.
14(2), 145-189.
Kendall, E.A. Krishna, P.V.M. Pathak, C.V. and Suresh.
C.B. 1998. Patterns of intelligent and mobile
agents. Proceedings of the 2nd International
Conference on Autonomous Agents (Agents’98),
New York. ACM Press.
Kuo, M. H. 2007. An intelligent agent-based
collaborative information security framework.
Expert Systems with Applications. 32(2),
585–598.
Ling, S.Seng and Loke, W. 2002. Advanced Petri Nets
for modelling mobile agent enabled
interorganizational workflows. Proceedings of
Ninth Annual IEEE International Conference
and Workshop on the Engineering of
Computer-Based Systems, 8-11 April 2002 245
– 252
Loke, S. W. Schmidt H. and Zaslavsky  A. 1999.
Programming the Mobility Behaviour of Agents
by Composing Itineraries. Proc. of the 5th
Asian Computing Science Conference on
Advances in Computing Science. 214 – 226. 
Lu, S. and Xu, C. 2005. A formal framework for agent
itinerary specification, security reasoning and
logic analysis. 25th IEEE International
Conference on Distributed Computing Systems
Workshops. 6-10 June 2005 580 – 586
Madhusudan T. 2005. An agent-based approach for
coordinating product design workflows.
Computers in Industry. 56(3) 235–259.
Meseguer, J. and Rosu, G. 2007. The Rewriting Logic
Semantics Project. Theoretical Computer
Science. 373(3), 213-237.
Meseguer, J. 1992. Conditional rewriting logic as a
unified model of concurrency. Theoretical
Computer Science. 96(1) 73-155.
1999. Distributed workflow co-ordination by
proactive software agents. Proc. of intelligent
workflow and process management, IJCAI-99
Workshop.
Ojha, A. C. Pradhan  S. K. and Patra M. R. 2007.
Pattern-Based Design for Intelligent Mobile
Agents.  4th International Conference on
Innovations in Information Technology. 18-20
Nov. 501-505
Peschanski, F. Darrasse, A. Guts, N. Bobbio, J. 2007.
Coordinating mobile agents in interaction
spaces. Science of Computer Programming. 66
246–265
Satoh, I. 2004. Selection of mobile agents. Proceedings.
24th International Conference on Distributed
Computing System. 484 – 493
Şerbănuţă, T.F. Roşu G. and Meseguer J. 2009. A
rewriting logic approach to operational
semantics. Information and Computation,
207(2), 305-340.
 8th International Conference of Modeling and Simulation - MOSIM’10 - May 10-12, 2010 - Hammamet - Tunisia
“Evaluation and optimization of innovative production systems of goods and services”

Stormer H. 2000. Task Scheduling in Agent-Based
Workflow in Proceeding of International ICSC
Congress on Intelligent Systems and
Applications (MAMA'2000: Multi-Agents and
Mobile Agents in Virtual Organizations and E-
Commerce). December 11-13, Wollongong,
Australia
Tahara Y. and Ohsuga  A. Behaviour Patterns for
Mobile Agent Systems from the Development
Process Viewpoint . 2001. Proc. of the 5th
International Symposium on Autonomous
Decentralized Systems, 239 – 242.
Trappey, A. J. C. Trappey, C. V. Hou, J. L. and Chen,
B. J. G. 2004. Mobile agent technology and
application for online global logistic services.
Industrial Management and Data System.
104(2) 169–183.
WFMC 99. 1999. Workflow Management Coalition
Terminology and Glossary [online]. Document
Number WFMC-TC-1011. Available from
http://www.wfmc.org/standards/docs/TC-
1011_term_glossary_v3.pdf [Accessed date
mars 09].
Xu, D. and Wang, H. 2002. Multi-agent collaboration
for B2B workflow monitoring. Knowledge-
Based Systems. 15(1) 485–491.
Yang; Z. Lu, S. Yang, P. 2006. Runtime Security
Verification for Itinerary-Driven Mobile
Agents. 2nd IEEE International Symposium on
Dependable, Autonomic and Secure Computin.,
177 – 186.
Zhuge, H. Workflow and agent-based cognitive flow
management for distributed team cooperation. 2003.
Information and Management. 40(1) 419–429