A MULTI AGENT SYSTEM MODEL TO EVALUATE THE

DYNAMICS OF A COLLABORATIVE NETWORK

Ilaria Baffo, Giuseppe Confessore, Graziano Galiano and Silvia Rismondo

Institute of Industrial Technologies and Automation, National Research Council

Montelibretti Research Area Roma1, Strada della Neve, 00010 Montelibretti (Rome), Italy

Keywords: Collaborative network, Multi Agent System.

Abstract: The paper provides a model based on the Multi Agent System (MAS) paradigm able to give a

methodological base to evaluate the dynamics in a collaborative environment. The model dynamics is

strictly driven by the competence concept. In the provided MAS the agents represent the actors operating on

a given area. In particular, it is supposed the agents being composed by two distinct typologies: (i) the

territorial agent, and (ii) the enterprise agent. Each agent has its local information and goals, and interacts

with the others by using an interaction protocol. The decision-making processes and the competences

characterize in a specific way each one of the different agent typologies working on the system.

1 INTRODUCTION

The rapid evolution in customer requirements is

forcing major changes in the overall industrial

system. One possible strategy for facing these

changes is based on the adoption of collaborative

way of working where different abilities and

competences are brought together with the goal of

exploiting benefits and sharing the risks.

This paper aims at providing a model for

understanding the dynamics of a network composed

of heterogeneous actors, and suggests a competence-

based collaborative way of working, where the

competence is defined as the ability to perform

activities by using a combination of knowledge, skill

and attitude. As Camarinha-Matos and Afsarmanesh

(2006) argue in their work, the definition of a model

certainly represents one of the main topics

concerning the collaborative network organization

research field.

The model well represents actors working on a

given geographical area, where the area boundaries

are both physical, and due to the existence of

consolidated business connections among the actors.

In particular, the model represents a breakthrough

with respect to the work by Confessore, Liotta, and

Rismondo (2006), where the authors exploited the

concepts of “competence measure” and “competence

map” to solve the problem of assigning to

collaborative enterprises the activities required for

carrying out an emerging business process. They

provided a Multi Agent System -MAS- (see

Jennings and Wooldridge, 1995), in which the actors

share information about their degree of competence

in doing the activities without reveal private data. In

fact, the competence there represent an aggregate

data based on a local evaluation, and all the actors

measure themselves with respect to the same set of

competences given by the competence map (see also

Hammami, Burlat, and Champagne, 2003). On the

basis of the previous MAS, the new model considers

new agent typologies and new features for the

decision-making processes in order to allow the

evaluation of the impact of new configurations of

the network with respect to key performance

indicators. The new configurations of the network

are generated whenever emerging business processes

and possible public funding (e.g., as calls for

National and European research project) arise, and it

is required to define the roles and responsibilities

through the actors.

2 THE SCENARIO

The paper provides a model for the understanding of

the dynamics of a collaborative network. The

network is represented by a coordinator, and private

actors. The coordinator makes decisions for

increasing the territorial attractive capacity respect

480

Baffo I., Confessore G., Galiano G. and Rismondo S. (2008).

A MULTI AGENT SYSTEM MODEL TO EVALUATE THE DYNAMICS OF A COLLABORATIVE NETWORK.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - AIDSS, pages 480-483

DOI: 10.5220/0001696904800483

 SciTePress

to new investments and new projects. Its main tasks

are:

1. The monitoring of new business opportunities

by doing intensive market analysis. The output

are: the proposal of activities to the agents in

order to meet the business opportunities; the

identification of new possible attractive

industrial sectors that could be exploited with

the actual territorial resources;

2. The monitoring of call for National and

International research projects. The output is

to suggest possible combinations of actors in

order to create the suitable composition of

partners meeting the call requirements;

3. Providing to the actors the competence map in

order to meet both the business and funding

opportunities while using a competence-based

criterion as a way for comparing the actors’

capability.

Due to the competence-based criterion, each

private actor has the main goal of increasing its

competences. Indeed, this condition allows it to

obtain an increasing number of activities of an

emerging business process or it allows becoming an

eligible actor for a research project as a partner. A

greater number of activities and of research project

participations, generate greater revenue for the

private actor that could be invested again for

increasing the competences by generating a positive

feedback. Summarizing, the competence-based

criterion driving both the business process activities

assignment, and the public funding exploitation,

generates a process of continuous development of

the territorial competences, stimulating the

collaboration between private actors. Moreover, the

investments growth pulls new research projects and

business processes by attracting also new enterprises

actors that are motivated to join the collaboration or

to work in the profitable territory.

For measuring the benefits given by the

competence-based collaborative network, the

following Social Wellness indicators are suggested:

 Number of new research projects/business

processes approved and finished. These

indicators represent a measure of the territory

attraction with respect to research project and

business processes, respectively.

 Number of new competences characterizing

the territory. The set of competences is not

static since new competences can be required,

and others can be not more useful to realize a

research project or a business process. This

indicator is useful for understanding how the

competences of the network evolve.

 Number of actors operating on the territory. It

indicates the development of the collaborative

network with respect to the actors’

composition.

3 MULTI AGENT SYSTEM

The MAS is composed of classes of interacting

agents each one having its local information and

goals. In this setting, the decision-making processes

and the competences characterize in a specific way

each one of the different agents working in the

system. In particular, it is supposed the agents being

represented by two distinct typologies: (i) the

Territorial Agent (TA), and (ii) the Enterprise Agent

(EA). TA represents the coordinator, while each EA

(e.g., representing a private company) interacts with

the other agents in order to pursue its goals. The

following notations are used through the paper:

 Let E = {e

,…, e

} be the set of z EAs.

 Let A = {a

,…,a

} be the set of k activities

in which a business process or a research

project can be decomposed.

 Let C = {c

,…,c

} be the competence map,

that is a set of w competences identified in a

certain time on the territory and globally

accepted by all the agents.

Next, the decisional processes are explained.

3.1 Territorial Agent

According to the described scenario, TA acts as a

coordinator and it solve the decisional problem

called Assignment Problem (AP): given a set of

enterprise and a set of activity we want to assign

each activity to exactly one of the EA at minimum

total cost. By solving the AP, TA obtains the

efficient allocation of the activities of a business

process to the EAs. The parameters

represents the

capability of e

∈E with respect to the execution of

the activity a

∈A. These values are provided by each

EA and are computed as described in the next

Section 3.2. Given the binary decision variable x

equal to 1 if the activity a

is assigned to the

enterprise agent e

, and 0 otherwise, it is possible

formulizing the AP as following:

Min

= ∑

i∈A

∑

j∈E

s.t ∑

∈

∩

= 1 for all a

∈A

(1)

3.2 Enterprise Agent

Each EA has to: (i) be able to define its degree of

competence respect to the competences required by

A MULTI AGENT SYSTEM MODEL TO EVALUATE THE DYNAMICS OF A COLLABORATIVE NETWORK

481

the activities of a business process; (ii) solve the

decisional problem of choosing which typology of

investment select in order to increase its

competence, that can be referable to a special case of

the Capital Budgeting Problem (Tobin, 1999); (iii)

decide if a coordinator proposal can be profitable or

not.

Each EA has to evaluate its degree of

competence with respect to each activity, thus it has

to define the value

. This value can be computed

as described in Confessore, Liotta, Rismondo (2006)

by modelling the subsets of competences required by

an activity and declared by an actor as vectors. The

value of

is obtained by computing the Euclidean

distance between the given vectors. Then the model

formulized in (1) aims at assigning each activity to

exactly one enterprise while minimising the distance

among vectors that is obtaining the solution

maximising the competence value.

3.2.1 Capital Budgeting Problem

The problem is to select the investments

maximizing the total return in term of competence

while respecting the budget constraints. The best

solution is the one producing the maximum

competence increment. The following parameters

are introduced:

 Let L be the set of possible investments.

 Let Q(e

)

[h,i]

be a matrix of parameters where

the generic element q

) represents the

return of competence c

∈C for e

∈E, respect

to the investment i∈ L.

 Let

be a positive value in [0, 1]

representing the degree of competence of the

enterprise e

respect to a specific c

∈C.

 A budget B(e

) representing a monetary value,

for all e

∈E.

 A cost of investment b

), for all i∈L, and

∈E.

Given the binary decision variable x

equal to1 if

the investment i is selected, and 0 otherwise, it is

possible formulizing the problem above mentioned

as following:

Max t = ∑

h∈C

∑

i∈L

s.t ∑

i∈L

≤ B(e

) for all e

∈E

∑

i∈L

(

+ q

)) ≤ 1 for all e

∈E

and c

∈C

(2)

The second constraints model the idea that for

each competence exists a threshold value equal to 1.

This value was introduced in Confessore, Liotta,

Rismondo (2006) as the maximum value for doing

as best one activity given the common scale of

benchmark values.

Once the decisional problem is solved, the EAs

update their competences, that is

’=

) if x

equal to1 thus the investment i∈L is selected, for all

∈E and c

∈C. Noteworthy, the system dynamics is

based on the hypothesis that if the actors do not

invest in a competence for long time, the value

decreases.

3.2.2 Project Participation Evaluation

Each EA decides to participate to a business

process or to a research project based on the

following data:

 Capacity availability.

 Profitability with respect to the increment of

the degree of competences.

 Profitability with respect to the collaboration.

For instance the collaboration with other

actors could remain also when the project

ends.

 Project relevance at scientific and research

levels.

4 THE MAS DYNAMIC

This Section summarizes the main features of the

interaction protocol exploited by the agents, then

defining the dynamics of the MAS. It is supposed

that the agents react in response to two possible

events, that is an emerging business or a funding

opportunity arise. The result of each decisional

problem contributes to the definition of a new

system configuration.

4.1 Business Process

Whenever a business opportunity occurs, two levels

of interaction between the agents can be defined.

First level: The information flow is from TA to

the EAs and vice versa.

Information Domain

: TA manages a list of

codified competences (i.e. the competence map,

globally accepted by the agents), a list of agents

operating in the system. TA decomposes the

business process in a set of activities, and defines the

subset of competences required for each activity.

Goal

: TA has to allocate each activities of the

business process according to competence-based

criterion, thus maximizing the total degree of

competences for realizing the project.

Communication Protocol

: TA communicates to

the EAs the set of activities and the related set of

competences required for doing them. Each EA then

ICEIS 2008 - International Conference on Enterprise Information Systems

482

communicates to TA an aggregate data describing its

level of competences.

Second level: It corresponds to the EAs actions in

response to the business opportunities. Since at the

first level TA decides by using the competence-

based criterion, each EA has to improve its degree

on competence by selecting possible investment.

Information Domain

: Each EA knows the

competence map.

Goal

: Each EA solves the problem of selecting

from a set of profitable investment the sub-set of

them maximizing the return of competence while

satisfying its budget.

Communication Protocol

: Each EA

communicates its availability in executing the

activities, or its degree of competences in order to

stimulate new collaborations

4.2 Funding Opportunities

Also in this scenario, whenever a funding

opportunity occurs, two levels of interaction

between the agents can be defined. In this paper, the

funding opportunities arise when the TA observes a

call for research project.

First level: The information flow is from TA to

the EAs. It is important to notice that for the EAs the

participation to a research project can be view as an

alternative profitable investment.

Information Domain

: TA manages a list of

codified competences, a list of agents acting on the

system. Furthermore, TA knows the competences

that best suite a call for project, and codifies the

composition of agent typologies that have the greater

probability of obtaining the financial fund approval.

Goal

: TA has to decide the best composition of

agents.

Communication Protocol

: TA contacts the agents

for proposing the project participation.

Second level: it corresponds to the EA and PA

actions in response to the funding opportunities.

Information Domain

: each agent knows its

degree of competences.

Goal

: EAs aim at carrying out the activities of

the research project by collaborating.

Communication Protocol

: The EAs response to

the TA request by communicating their availability

to execute the project activities. If they decide to

participate then communicate to TA their

competences. The probability of obtaining the public

funding will be a function of the agents’

composition and on the total degree of competences.

The agents collaborate during the project duration

and they have a return of competences due to the

research project collaboration.

5 CONCLUSIONS

The paper analyses a competence-based

collaborative network, identifying roles, decision

making processes and the interaction protocol

between the actors. The model is based on the Multi

Agent System paradigm and it is driven by the

competence concept. Even if the model does not

capture all the aspects of the collaboration, it

represents a further step toward the representation of

the collaborative networks, and the understanding of

what and how a network of actors has benefits from

the collaboration. Actually, both the dynamics and

the decisional problems are faced by the

implementation of ad hoc algorithms. In the future,

the plan is to add new features to the MAS in order

to suggest the model as a valid way for studying the

complex connections between collaborative actors,

and then to exploit it in a real case-study as

preliminary done in Baffo et al. (2006).

REFERENCES

Baffo I, Dedonno L, Confessore G, Rismondo S., 2006,

Descrizione ex-post di una rete relazionale territoriale

e realizzazione di uno strumento per la simulazione

dinamica ex-ante, in Proceedings of the XXVII Italian

Conference of Regional Science - Impresa, Mercato,

Lealtà Territoriale.

Camarinha-Matos, L. M., Afsarmanesh, H., 2006, in IFIP

International Federation for Information Processing,

Volume 224, Nerwork-Centric Collaboration and

Supporting Frameworks, eds. Camarinha-Matos, L.,

Afsarmanesh, H., Ollus, M., (Boston: Springer), 2006,

pp. 3-14.

Confessore G, Liotta G, Rismondo S., 2006, in IFIP

International Federation for Information Processing,

Volume 224, Nerwork-Centric Collaboration and

Supporting Frameworks, eds. Camarinha-Matos, L.,

Afsarmanesh, H., Ollus, M., (Boston: Springer), pp.

121-128.

Hammami, A., P. Burlat, and J.P. Champagne. 2003,

Evaluating orders allocation within networks of firms.

In International Journal of Production Economics; 86 :

233-249.

Jennings N.R., and M.Wooldridge, 1995, Applying agent

technology, Applied Artificial Intelligence; 9: 357-

369.

Tobin, R.L., 1999, A fast interactive solution method for

large capital expenditure selection problems.

European Journal of Operational Research 116, 1–15.

A MULTI AGENT SYSTEM MODEL TO EVALUATE THE DYNAMICS OF A COLLABORATIVE NETWORK

483