Вы находитесь на странице: 1из 6

AHP/ANP a Decision Making Service in PROSIS Model

Yves Dubromelle*. Thamer Louati* Fouzia Ounnar* Patrick Pujo*

* LSIS, Aix Marseille University, Av Esc. Normandie Niémen, 13397 Marseille cedex 20,
France (e-mail: {yves.dubromelle, thamer.louati, fouzia.ounnar, patrick.pujo}@lsis.org).

Abstract: An approach that relies on the use of the holonic paradigm, on an isoarchic architecture and on
a decision-making capacity based on a multicriteria analysis is presented. The various concepts of this
approach are addressed first. Then, the multicriteria decision methods based on AHP (Analytic Hierarchy
Process) and its generalization ANP (Analytic Network Process) are detailed, as well as the
implementation phases. The first obtained results on a case study are presented.
Keywords: Multicriteria decision, AHP/ANP, PROSIS model, Isoarchic architecture.
propose the concept of self scheduling driven by the product
1. INTRODUCTION
or by the interaction product-process. A product carries
information that it is able to communicate to the decision
The quest for better productivity has lead to search solutions
centres associated to resources. The entity managing
aiming to eradicate the production loss and to lower the immaterial aspects (information, communication and
products obtention cost. These solutions include new ways of decision) is called I-product. In parallel, (Mc Farlane et al.,
controlling production systems, some of them based on
2003) defines the intelligent object concept as a dual object
heterarchic or holonic architectures. This paper describes an (physical and virtual) with information processing capability
aspect of the PROSIS isoarchic control system (Pujo et al., (memory, communication, computing, action …). The
2009): the multicriteria decision making algorithm that help
‘active’ product can be given means for capturing
the resource in choosing the product to process in a WIP environment variations, making decisions and thus fully
(Work in progress) or the product in choosing the next interacting with its environment (process resources, other
resource to go on. The multicriteria decision aid presents an products, human operators, etc.) (Morel et al., 2007). Finally,
important advantage over other approaches: its flexibility. whatever name given, I-product, virtual product, extended
Indeed, it is possible to change the criteria and/or indicators
product, avatar or other, future operation of manufacturing
and/or their different weights to guide the achievement of systems will rely on this type of entity. However, objects of
results. First, we shortly present actual research works in different nature will have to inter-operate: the holonic
intelligent manufacturing control. Then, we describe the
paradigm (Koestler, 1967) is not restricted in the Holonic
PROSIS model for isoarchic control systems, based on a set Manufacturing Systems (Mathews, 1995) to an oriented
of holonic entities interacting through a set of entities that product vision only; there exist other types of holonic entities
offer ambient services. We describe various holons models
with a role as much important. Among proposals from the
and we present the use of multicriteria model based on the HMS community, Product, Resource and Order Holons are
Analytic Hierarchy Process (AHP) method (Saaty, 1980) in three types of basic Holons that are most recurrent. We are
the decisional process. Then, we present the first results using
referring to the most known holonic architecture: PROSA
the Analytic Network Process (ANP) method (Saaty, 2005) (Van Brussels et al., 1998). In PROSA, a resource holon is
which is the generalization of AHP, to dependences and associated to each resource, product holon corresponds to a
feedbacks.
product type, and order holon corresponds to the product
instantiation activity (Valckenaers et al., 2009). Starting from
2. INTELLIGENT MANUFACTURING CONTROL there, PROSA shows that the control of manufacturing
systems by products only is not enough. Indeed, the product
Intelligence is referred to as ambient because of the as such does not carry all the operational constraints and all
omnipresence of wireless communicating non apparent related information that would allow making optimal, or at
computing agents. The ambient intelligence concept, also least satisfactory, control decisions. However, the use of
called ubiquity, gives users the possibility to interact from PROSA induces a difficulty to give an active role to each
any place with many interconnected infotronics devices, product. Indeed, the product instantiation is made through the
sensors and actuators, embedded around them and operating order holon. But an order holon can correspond to a batch
through ad hoc networks with distributed architecture. The and thus to multiple products simultaneously. It is then
common point between future production imperatives is the impossible to differentiate each product. It is by taking into
increasing need of adequacy between the actual account the whole set of data and constraints, sometimes
manufacturing system and its associated information system contradictory, including those that are specific to each
(Udoka, 1991), combined with deeper granularity in which product, that the best possible solution at a given time can be
the detail level is at part unit. (Pujo and Brun Picard, 2002) derived for manufacturing system evolution and control. This
is why we support control that integrates viewpoints coming PROSIS aims to explore this approach which is specially
from different interacting entity types. suited to ambient control systems. The absence of a central
decision system forbids any predefined or forecast
3. ISOARCHIC CONTROL: PROSIS organisation of manufacturing system operations. These
should thus be progressively organised by the holons
PROSIS (Product, Resource, Order, Simulation Isoarchic themselves with the support of ambient service entities. This
System) (Pujo et al., 2009) proposes a holonic and isoarchic self-organisation assumes real-time characteristics
approach facilitating the implementation of ambient control considering all information characterising each Holon
solutions for manufacturing systems. contributing to define the operations. We then talk about self-
organised control functions. These functions are integrated
into the intelligence associated to each Holon. For that, we
3.1 PROSIS model presentation
define a Holon as a conceptual entity based on the association
of a Material Structure (the M_holon), an Information System
The PROSIS objective is to study and develop decision
and a Processing System (the I_holon) that provides a
mechanisms with architecture and information system being
decisional intelligence allowing interaction with other
as close as possible of the material system, even to the image
Holons. This structure allows recursive decomposition of
of the organisation of this system, and directly interconnected
manufacturing systems, in compliance with the holonic
to it via infotronics technologies. This approach objective is
paradigm, by clearly showing the duality and parallelism
to gain in terms of structural and decisional flexibility, and
between the real world (material) and the informational world
thus in terms of reactivity and adaptability. This approach
(immaterial, in which data and decision making stands).
marks a break with previous hierarchical models in which
components are of the type ‘master - slave’ following a tree The Management of production related knowledge results
like and not varying topology of decision centres. A Holon from interactions between the three types of holons. PH-RH
has a decisional intelligence giving the possibility to act on interactions provide process knowledge: resource operating
its own behaviour and also to act on the behaviour of the methods, capacity, reachable quantities and possible results.
systems it belongs to. Hierarchical decomposition is replaced PH-OH interactions indicate production knowledge: batches
by Holon recursion and implementation of the Janus effect. description. RH-OH gives execution process knowledge.
Different holonic architectures are proposed in the literature However, a major difference with PROSA is the
for control HMS (Deen, 2003). disappearance of the Staff Holon which is not needed in an
isoarchic context. We define a Simulation Holon having a
These architectures present however the inconvenient of
totally different objective: starting from the manufacturing
giving an important place to the hierarchy concept in decision
system status obtained by analysing interactions between the
making. While the planning is feasible, the holonic execution
other holons, the aim is to simulate the manufacturing system
manufacturing system just executes it. When basic Holons
evolution, to provide evolution indications to the workshop
cannot find in PROSA a scheduling solution, a solution is
manager and to anticipate eventual failures via diagnosis
derived by a Staff Holon which uses a centralised processing
actions. This Holon does not contribute to self organisation,
algorithm. In order to make the staff holon optional,
but it facilitates the role of the workshop manager: it provides
(Valckenaers et al., 2006) (Verstraete et al., 2008) propose to
control with proactive properties. This holon is not addressed
use ant colonies, that find an alternative which resembles the
in this paper.
planning as good as possible. In this solution, every product
is not individually identified.
3.2 Product Holon
In order to simplify the implementation of an ambient control
system, we suggest that all interacting entities be at the same A Product Holon (PH) is made with a M_product (the
decision level. This means an isoarchic architecture. The material object) and an I_product containing the
isoarchy concept (word made from Greek iso (equal) and manufacturing process and also its state model and all
archy (power)) refers to the same decision power and thus to traceability information. Therefore, there exist in PROSIS as
a complete absence of hierarchy. In a decision system made many PH instances as manufactured products or/and WIP
of several decision centres, a decisional architecture can be products. This is a major difference regarding PROSA in
qualified as isoarchic when each decision centre has the same which the Product Holon acts as an information server to the
decision capability. This property can easily be obtained other Holons of the HMS, delivering technical information
when decision mechanisms are duplicated in each decision for a given type of product but not containing products state
centre and appropriately parameterised. Isoarchy appears as a information. By definition in manufacturing, a product is a
particular specification of the concept of heterarchy and as nomadic entity. It is thus necessary to identify it and, for that,
the opposite of the concept of hierarchy (Mesarovi et al., to tag it. Unit identification goes through the deployment of
1980). However, within this category it expresses an even ad hoc technologies linking each M_product to its I_product.
concept that can be applied only to truly and totally A good example of these technologies is RFID with an ID tag
equalitarian architectures. This particular situation between attached to the M_product carrying at least an identification
Holons has been foreseen in holonic systems through the number, eventually completed with key information of the
concept of ‘flat holonic form’ (Bongaerts et al., 2000). I_product. Other information stored in the central database
However, this architecture was not really deeply studied.
and accessible through a network can be associated to this Multicriteria decision making (MCDM) is a field that
nomadic information. chooses the best of a discrete set of alternatives. Unlike the
usual methods of optimization that assume the availability of
3.3 Ressource Holon measurements, measurements in MCDM are assumed to be
derived or interpreted subjectively as indicators of preference
The Resource Holon (RH) is conceptually similar to the and of the strength of preference (Saaty, 2005). One person’s
definition proposed in PROSA: it includes a material part, preference is different than another person’s and thus the
outcome depends on who is making the decisions and what
like automated equipment (NC machine tools, industrial
robot), making an M_resource, and an information processing his/her preferences and goals are. The consideration of
part, the I_resource, which drives the equipment and multiple conflicting objectives in a decision model has made
the area of MCDM very challenging (Muralidharan et al.
contributes to allocate tasks to the resources. Resource
allocation methods for the I_resource are not the same as 2002). A survey of methods using multicriteria (Ounnar
1999, Ounnar et al., 2007) has led the authors to select a
those in PROSA since interactions with the other types of
Holons are defined in an isoarchical context. Furthermore, method called Analytic Hierarchy Process (AHP) (Saaty
RF identification is implemented only in the case of nomadic 1980, Muralidharan et al. 2002, Rong et al. 2003). AHP has
advantages over other decision-making approaches (Vargas
resources, like mobile robots, shuttles, etc.
1990) these include its ability to: (i) handle tangible and
intangible attributes; (ii) structure the problems hierarchically
3.4 Order Holon to gain insights into the decision-making process; and (iii)
monitor the consistency of the judgments of a decision-
An Order Holon (OH) represents a task in a manufacturing maker. The AHP has demonstrated robustness across a range
system: a manufacturing order concerning in general a set of of application domains.
PH. It is responsible for the performance of assigned work Recently, we are interested in an extension of AHP, called
within specified times. It is thus closely linked to the concept Analytic Network Process (ANP), which allows to introduce
of batch, WIP and delays / lead time. This is a nomadic entity more precision into the multicriteria decision mechanism.
with a strong link with one (or several) Product(s) Holon(s).
The I_order checks dates satisfaction during work
4.2 Introduction to AHP method
performance and about the consideration of economic factors
(batch size, WIP volume, minimisation of production
In AHP, the hierarchy of the decision-making process is
changes, batch partition, etc.). The M_order will be,
according to the case, the manufacturing order with an ID defined by a quadruplet <N1, N2, N3, N4>; where: N1=
tag, or the container also tagged, allowing the manipulation Global Objective level; N2= Criteria level; N3= Indicators
level; N4= Alternatives level. This process organizes a
of one or several M_products.
hierarchical decision-making problem in a mathematically
rigorous manner to ensure proper results of the decision-
4. MULTICRITERIA DECISION TOOL
making process. The central 'ingredient' of the AHP method
is comparisons. A pair-wise comparison assesses the relative
The self-evaluation process of each RH, in order to rank the
importance of two elements of a same level with respect to
set of products in the WIP, is based on a multicriteria method.
the decision maker contributing to reaching the objective of
The multicriteria decision aims at providing tools to a
the adjacent higher level. For that, in the case of a qualitative
decision maker enabling him to progress in the resolution of a
comparison, it is necessary to choose a scale of values to
problem where several points of view, often contradictory,
specify the degree of importance (weight, priority) of an
must be taken into account. Three classes of multicriteria
element with regard to another. We adopted the scale of value
methods can be distinguished: decision aid methods,
(1-9) used in the AHP method (Saaty, 1980). We propose to
elementary methods and mathematical optimization methods.
implement AHP through two main phases: configuration and
The choice of one class of methods can depend either on the
exploitation (Ounnar and Pujo, 2009).
data available to treat the multicriteria problem, or on the way
the decision maker models its preferences. The selection The configuration phase first consists in deciding the relative
process of the product (PH) that will be executed by the importance between the criteria and deciding, in each
resource (RH) supposes knowledge of the various possible criterion, the relative importance between their associated
alternatives in order to carry out sorting with respect to a set indicators. This is done by filling pair wise comparison
of criteria. So, the use of optimization methods is not matrices for the root element and each criterion. These
possible. In addition, the objective being to carry out a matrices are then analyzed to look for any incoherencies in
classification, the elementary methods are not considered. the judgments made. When this is done the matrixes are
Multicriteria decision aid methods make it possible to bring converted into priority vectors that give the relative
help to the decision maker during the refinement of the importance of each compared element.
decision-making process related to the choice of an
alternative among a set of potential alternatives. The exploitation phase allows to obtain the vector giving the
relative importance of the alternatives compared to the global
4.1 Choice of an MCDM method objective (level 4 compared to level 1). This aggregation
principle consists in carrying out matrices products. It is
based on the different relative vectors of a level compared to select the first product that will be treated. Table 1 shows a
its adjacent level. In this phase, the real data (value of the small part of the information related to these products that are
indicators regarding the alternatives) are taken into account. stored in the PH.
The component ‘y’ that has the highest value in the priority
vector corresponds to the product (y) chosen. Table 1: Product’ information
Reference Typ_Prod Typ_Prod2 Typ_Prod3
P1 Principal Flow Shop Spare
4.3 Introduction to ANP method P2 Complement Jop Shop Spare
P3 Principal Flow Shop Assembly
ANP is a generalization of AHP (Saaty, 2005). While AHP P4 Principal Jop Shop Spare
represents a structure with a unidirectional hierarchical
5.2 Configuration phase of ANP algorithm
relationship, ANP reflects more complex correlations
between the decision levels and attributes (the elements of
each level). A feedback system can be represented by a Step 1.1: Network structure for the decision problem
network where nodes correspond to levels (Saaty, 1980). In
this Feedback Network, the elements in a component and The control criteria will help to properly structure the
those in another component may influence other elements in problem and establish the comparisons between the
the same component (inner dependence) and those in other components. This control criterion is used to answer the
components (outer dependence) with respect to each of question of dominance, and for the decomposition of a
several properties. To summarize, AHP provides a decision complex problem with a variety of influences.
tree structure with pair-wise comparisons at each level of the ANP is applied to improve the control mode of the
tree while ANP proposes a decision network structure with a production system. In PROSIS, where there is no scheduling
generalized pair-wise comparisons. task prior to the execution, we use ANP in order to decide in
The above section describes a process for scheduling real time which product, in the WIP queue of a given
products among a resource. The scheduling is a real-time resource. This is the global objective in the network (Fig. 1).
scheduling. Only the determination of the next product to be To achieve this objective we chose to take into account five
executed by the resource is interesting. The multicriteria criteria that represent three points of view regarding the
classification determines the product ranked first, which will holons (Ounnar and Pujo, 2009). First, the PH need to
also be chosen. It is based on the generalization of the AHP progress through the production system (C2: Product
to dependence and feedback, the ‘Analytic Network Process progression), to be sure not to be buried forever in the WIP
(ANP)’ method. queue (C4: queue time). We also define the criterion C1
(Product type), related to PH, which allows to prioritize some
5. ANP EXPLOITATION IN PROSIS types of products according to their belonging to different
classes. The RH aims maximal productivity, we don’t want it
The ANP algorithm has been implemented in the simulation to be too long unused so we load it with secondary products,
prototype based on the holonic, isoarchic and multicriteria but they mustn’t disturb the principal products. The resource
control (PROSIS) for a mechanical production workshop. A also has to be careful not being monopolized by a single
multicriteria decision making algorithm applying AHP / ANP product (C5: Resource workload). The OH’s principal worry
is established in each RH and defines a PH classification, is to have the final products delivered in time. So it takes care
taking into account RH, PH and OH constraints. The product along the production process that the product has enough
ranked in first position is the next one to be processed by the margin (C3: Remaining slack). To define the dependence
resource. matrices, an economic control criterion was used.
The data concerning the first criterion are static data
5.1 Presentation of the case study
presented in Table 1. For the other criteria, the majority of
values needed to calculate indicators are calculated
The production system test, the Unit 4-5, is a mechanical dynamically, “in the loop” during the simulation.
production workshop that manufacture in a recurring way a
score of different parts types, mainly ‘shafts’ and ‘sleeves’ of
Step 1.2: Relative importance matrices and dependence
big sizes, having close tolerances and complex routings.
matrices
These parts are in the heart of the kinematics diagram of the
aircraft mechanical engineering manufactured by the In ANP, like AHP, the components themselves are compared
company. This workshop comprises about fifteen CNC pair-wise with respect to their contribution to the goal.
machining centre. This tool machines are versatile, of high Decision makers are asked to respond to a series of pair-wise
precision, with great capacities and equipped with all the ad comparisons where two elements or two components at a
hoc peripheral equipments and tools. The batches size seldom time will be compared in terms of how they contribute to
exceeds about fifteen parts, each part representing a value of their particular upper level criterion.
many hundreds of thousands Euros. If there are dependencies among components, the decision
A resource having in its queue four products to be executed elements at each component are compared pair-wise with
have been considered. ANP method has been applied to respect to their importance towards a control criterion. In
addition, if there are interdependencies among elements of a the influence of other elements on it. The relative importance
component, pair-wise comparisons also need to be created, values are determined with the scale (1-9).
and an eigenvector can be obtained for each element to show

Fig. 1. ANP Network


Figure 2 shows an example of a relative importance matrix [CrOg] represents the relative importance vector of each
(the criteria compared to the global objective). According to criterion regarding the global objective; [CC]dep represents
the ANP network (Fig. 1), there are internal dependencies (in the dependence matrix between the criteria; I represents the
criterion C1) and other external (between C2 C3 C4 C5) Identity matrix.
Global Criteria Indicators Alternatives
The dependence matrices between criteria and indicators are Objective
then established from these influences. Global Objective (Og) I
Criteria (Cr) [CrOg] [CrCr]dep
C1 C2 C3 C4 C5
Indicators (In) [InCr] [In In]dep
C1 1 1/3 1/3 3 1 Real Data
Alternatives I
C2 3 1 1 9 3
C3 3 1 1 9 3 Fig. 4.Super-matrice structure
C4 1/3 1/9 1/9 1 1/3 Etape 2.2: Selection of the better alternative
C5 1 1/3 1/3 3 1
The supermatrix must be transformed first to make it
Fig. 2. Criteria importance matrix stochastic, that is, each column of the matrix sums to unity.
Figure 3 shows an example of a dependence matrix for the The blocks in each column of the supermatrix are weighted,
criterion C2 which is dependent on the criterion C3. and the result is known as the weighted supermatrix, which is
stochastic. Then, the stochastic matrix is multiply by itself
C1 C2 C3 C4 C5 several times to reach a stationary matrix (the limit
C1 0 0 0 0 0 supermatrix). Indeed, raising a matrix to powers gives the
C2 0 1 5 0 0 long term relative influences of the elements on each other.
C3 0 1/5 1 0 0 To achieve a convergence on the importance weights, the
C4 0 0 0 0 0 weighted supermatrix is raised to the power of 2k+1; where k
C5 0 0 0 0 0 is an arbitrarily large number. This matrix will indicate the
Fig. 3. Dependence matrix relative to C2 criterion rank of alternatives compared to the global objective.
The ranking of alternatives is therefore in the block 4:1
53 Exploitation phase of ANP algorithm (Alternatives: Global Objective (Og)) of the Figure 4).

Etape 2.1 : Etablishment of the Supermatrix The super-matrix is based on the relative importance of
A supermatrix is a partitioned matrix, where each matrix vectors calculated in the configuration phase and also on real
segment represents a relationship between two nodes in a data (indicators values for each alternative). In the PROSIS
system. The vector of priorities (or relative importance) model, the real data for the indicators are derived from
calculated in the previous phase are elements of the information available in the different holons.
supermatrix. This latter represents all the dependencies From the data in Table 1, the values of indicators of criterion
between components (criteria) or elements (indicators) of the C1 (Product Type) are highlighted. Similarly, the other
network. Figure 4 shows the supermatrix [SM]. Where:
indicators values are obtained by analysing the information in challenges and issues. Control Engineering Practice, 15
the RH and OH. (11), 1321-1331.
The priority vector of the products is given by (1): Muralidharan, C; Anantharaman, N., and Deshmukh, S.G.,
P1 → % 0,294 " % P3 "
(2002). A multi-Criteria Group Decision-making Model
# # for supplier Rating. Supply Chain Management, 38, (4),
P 2 → # 0,186 # P1
& # (1) 22–33.
P3 → ## 0,309 # P4 Ounnar, F. (1999) Prise en compte des aspects décision dans
P 4 → #$ 0,211 ! # P2
$ ! la modélisation par réseaux de Petri des systèmes
The same experiments were conducted without the flexibles de production. PhD Institut National
dependencies; the result is given by (2): Polytechnique de Grenoble.
Ounnar F., Pujo P., Mekaouche L., Giambiasi N. (2007).
P1 → % 0,329 " % P1 " Customer-supplier relation-ship management in an
# #
P 2 → # 0,160 # P3 intelligent supply chain network. Production Planning &
# & # (2)
P3 → # 0,304 Control, 18(5), 377-387.
# P4
P 4 → #$ 0,207 ! # P2
$ !
Ounnar, F., Pujo, P. (2009). Pull control for Job Shop:
Holonic Manufacturing System approach using
The first product to be executed by the resource is P1. multicriteria decision-making, Journal of Intelligent
Applying ANP (thus with taking into account the Manufacturing, DOI 10.1007/s10845-009-0288-4.
dependencies), we note that the classification of products in Pujo, P. and Brun-Picard, D. (2002). Pilotage sans plan
the queue of the resource has changed. It shows the existing prévisionnel ni ordonnancement préalable. In: P. Pujo
dependencies and especially the dependence between C2 and and J.P. Kieffer, Eds. Collection IC2 - Productique :
C3 which are the most important criteria. Méthodes du pilotage des systèmes de production. Paris:
Hermès Science Europe Ltd.
CONCLUSION Pujo, P., Broissin, N., and Ounnar, F. (2009). PROSIS: An
isoarchic structure for HMS control, Engineering
After presenting the control system based on a holonic and Applications of Artificial Intelligence 22: 7. 1034-1045
isoarchic approach (PROSIS model), we have demonstrated Rong, C., Takahashi, K., and Wang, J. (2003). Enterprise
the interest of using multicriteria methods AHP / ANP as waste evaluation using the analytic hierarchy process and
decision support tool in this context. ANP is based on the fuzzy set theory. Production Planning & Control, 14,
generalization of the AHP to dependence and feedback. For (1), 90–103.
this, the AHP method was presented and then the various Saaty, T.L. (1980). The Analytic Hierarchy Process, Mc Hill.
phases and steps of the ANP method were described. The use Saaty, T. L. (2005). Theory and Applications of the Analytic
of this method as a decision support in the PROSIS control Network Process. RWS Publications, Pittsburgh, PA,
model has been presented while applying on case study. It Udoka, S.J. (1991). Automated data capture techniques: A
was highlighted that decision support in PROSIS take into prerequisite for effective integrated manufacturing
account dependencies between criteria and indicators for systems. Computers & Industrial Engineering, 21 (1-4),
choosing the better solution. The presented work represents 217-221.
the premises of a new control orientation research. Valckenaers, P., Hadeli, K., Saint Germain, B., Verstraete, P.
and Van Brussel, H. (2006). Emergent short-term
REFERENCES forecasting through ant colony engineering in
Bongaerts, L., Monostori, L., McFarlane, D. and Kadar, B. coordination and control systems. Advanced Engineering
(2000). Hierarchy in distributed shop floor control. Informatics, 20, 261–278.
Computers In Industry, 43, 123-137. Valckenaers, P., Saint Germain, B., Verstraete, P., Van
Deen, S.M. (2003). Agent-Based Manufacturing – Advances Brussel, Van Belle, J., Hadeli, K. and Van Brussel, H.
in the Holonic Approach. Springer-Verlag Ed, ISBN 3- (2009). Intelligent products: Agere versus Essere.
540-44069-0. Computers in Industry, 60 (3), 217-228.
Koestler, A. (1967). The ghost in the machine. London: Van Brussel, H., Wyns, J., Valckenaers, P., Bongaerts, L. and
Editions Hutchinson. Peeters, P. (1998). Reference architecture for holonic
Mathews, J. (1995). Organizational foundations of intelligent manufacturing systems: PROSA. Computers in Industry,
manufacturing systems; the holonic viewpoint. Computer 37, 255-274.
Integrated Manufacturing Systems, 8 (4), 237-243. Vargas L.G. (1990). An Overview of the Analytic Hierarchy
Mc Farlane, D., Sarma, S., Chirn, J.L., Wong, C.Y. and Process and its Applications. European Journal of
Ashton, K. (2003). Auto ID systems and intelligent Operational Research, 48, 2-8.
manufacturing control. Engineering Application of Verstraete, P., Valckenaers, P., Van Brussel, H., Saint
Artificial Intelligence, 16 (4), 365-376. Germain, B., Hadeli, K. and Van Belle, J. (2008).
Mesarovic, M.D., Macko, D. and Takahara, J. (1970). Theory Towards robust and efficient planning execution.
of Hierarchical Multilevel Systems. New York: Engineering Applications of Artificial Intelligence, 21,
Academic Press. 304–314.
Morel, G., Va1ckenaers, P., Faure, J.M., Pereira C., and
Dietrich, C. (2007). Manufacturing plan control

Вам также может понравиться