REPMAS: A Requirements Engineering Process for MultiAgent

Systems: An Application Example

Iderli Pereira de Souza Filho

, Gilleanes Thorwald Araujo Guedes

and Giovane D’Avila Mendonc¸a

Software Engineering Post-Graduation Program, Pampa Federal University, Av. Tiaraju, 810, Alegrete, Brazil

Keywords:

Requirement Engineering Process, Model BDI, Multiagent Systems.

Abstract:

Requirements engineering is a crucial phase for the development process of any software, including multi-

agent systems. This particular kind of software is composed of agents, autonomous and proactive entities

which can collaborate among themselves to achieve a given goal. However, multi-agent systems have some

particular requirements that are not normally found in other software, making the requirements engineering

general processes and techniques less efﬁcient. Taking this into account, this work presents a speciﬁc require-

ments engineering process for multi-agent systems with emphasis in a consolidated model in the cognitive

agents development (BDI Model). This process supports the requirements engineering subareas of elicita-

tion, analysis, speciﬁcation, and validation, thus, presenting a wide coverage of the requirements engineering

area. The proposed process was evaluated through its application in requirements engineering of Heraclito

multi-agent system. This assessment allowed us to identify future works and improvement points in our work.

1 INTRODUCTION

Agents technology is a software paradigm that pro-

vides agent abstractions for open, distributed, and het-

erogeneous systems development (Gan et al., 2020).

Software agents are autonomous, social, reactive and

proactive entities (Hajduk et al., 2019). A multia-

gent system (MAS) consists of a set of agents that

can interact with each other. These systems can be

used to solve problems that are difﬁcult or impossible

for a monolithic system or for a single agent to solve

(Rodr

ıguez Viruel, 2011).

Ferber in (Ferber, 2014) states that developing a

MAS is not an easy task. This led to the emergence

of AOSE (Agent-Oriented Software Engineering). Sl-

houb in (Slhoub et al., 2019) states that AOSE aims to

support development processes for agent-based sys-

tems, imposing a solid practice of software engineer-

ing in concepts of artiﬁcial intelligence.

AOSE objectives include adapting requirements

engineering to the multiagent context. Although there

are development processes that address requirements

engineering for MAS, we noticed gaps and points for

https://orcid.org/0000-0001-9694-0977

https://orcid.org/0000-0001-5457-2600

https://orcid.org/0000-0001-9445-6831

improvement in their approaches which were discov-

ered by means of a primary study, where we carried

out a literature systematic review (MendonC¸ a. et al.,

2021). Among these gaps, we considered the most

important the need of a process that encompasses the

four subareas of requirements engineering deﬁned in

SWEBOK (Bourque et al., 2014) and a support to re-

quirements needed to cognitive agents based on the

Belief-Desire-Intention (BDI) model.

Moreover, according to (Javed et al., 2022) re-

quirements engineering is an essential activity in

the software development lifecycle and, according to

(Herzig et al., 2017) the concepts of beliefs and goals

(desires) play a central role in the conception and im-

plementation of agents. Thus, we realized the need to

propose a requirements engineering process for MAS,

allowing to represent particular requirements for this

kind of system and surpassing the main gaps we found

in current processes.

We called this new process REPMAS (Require-

ments Engineering Process for MultiAgent Systems).

We incorporated in REPMAS some existing works for

MAS requirements engineering, such as the Homer

technique and the MASRML language.

Homer technique allows agent-oriented require-

ments elicitation and it has the advantage of being

easily usable in other approaches (Wilmann and Ster-

Filho, I., Guedes, G. and Mendonça, G.

REPMAS: A Requirements Engineering Process for MultiAgent Systems: An Application Example.

DOI: 10.5220/0011798800003393

In Proceedings of the 15th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2023) - Volume 1, pages 387-395

ISBN: 978-989-758-623-1; ISSN: 2184-433X

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

387

ling, 2005)), while MASRML (Guedes et al., 2020)

is a UML-based language that adapts the use-cases

diagram in order to support the main concepts used

in MAS based on the BDI model. It is important to

highlight that our process was validated by means of

its application in the second version of the Heraclito

system (Galafassi et al., 2019).

This paper is organized as follows, section 2

presents the basic concepts applied in this work. Sec-

tion 3 compares the proposed process with the other

MAS processes that deal somehow with requirements

engineering. Our requirements engineering process is

detailed in Section 4 and the results of its application

in Section 5. Lastly, we present the conclusion.

2 BACKGROUND

2.1 Agents and Multiagent Systems

An agent is a computational process, established in

an environment, designed to achieve a purpose by

means of autonomous and ﬂexible behavior (Vicari

and Gluz, 2007). An agent has autonomy, social

ability, reactivity, and pro-activity. Autonomy is the

agent’s capability to operate without the users’ direct

intervention. Social ability allows agents to interact

among themselves by means of some kind of agent

communication language. Reactivity refers to the ca-

pability of perceiving the environment and answering

timely to the changes that occur in the environment.

Finally, pro-activity is the agent capability to have a

behavior directed to a goal, taking the initiative with-

out receiving orders (Hajduk et al., 2019).

A multiagent system is composed of sev-

eral intelligent agents interacting among themselves

(Wooldridge, 2009). The development of this kind of

system is used in several areas (Liu et al., 2010), due,

among other reasons, to its capability to deal with

complexity (Boes and Migeon, 2017).

2.2 Belief-Desire-Intention (BDI) Model

The Belief-Desire-Intention model is a software

model developed to programming intelligent agents

characterized by including beliefs, desires, and inten-

tions in the agent architecture (Bratman et al., 1987).

Beliefs represent what an agent believe to be true

about the environment, about itself, and about other

agents. Desires represent the goals the agent would

like to achieve. And intentions represent desires the

agent believes he can achieve and take actions to

achieve them (Rao and Georgeff, 1995).

2.3 MASRML

MASRML – Multi-Agent Systems Requirements

Modeling Language – is a UML-based Domain-

Speciﬁc Modeling Language (DSML) conceived for

the requirements modeling in multiagent system

projects (Guedes et al., 2020). This DSML extends

the UML metamodel in order that the use-case dia-

gram can be applied in the speciﬁc domain of multi-

agent systems. MASRML provides mechanisms to

represent the concepts of agent role, goal (desire),

perception, belief, intention, plan, and action, sup-

porting the concepts of the BDI model.

In MASRML, new concepts were created inspired

by concepts of UML. The main contribution was the

creation of the AgentRoleActor and InternalUseCase

metaclasses to represent agent roles and the function-

alities assigned to them, stereotyped in goals, per-

ceptions, actions, and plans. Some associations were

also created in order to associate perceptions, actions,

and plans to the goals that an agent role wishes to

achieve, thus establishing the conditions for a goal to

become an intention and which plan will be executed

in this case, as well as the possible external actions

performed during the execution of a plan.

Furthermore, MASRML internal use-cases docu-

mentation provides a clear view of the structure of

cognitive agents, allowing, for example, to determine

what is needed for a given goal to become an inten-

tion or to establish the steps to be performed when

executing a perception or a plan.

REPMAS uses MASRML notation throughout its

execution, producing partial MASRML use-case dia-

grams for each agent role and a ﬁnal MASRML use-

case diagram representing all agent role actors and all

the internal use cases assigned to each one of them.

2.4 Requirement Engineering

In SWEBOK (Bourque et al., 2014) it is stated that

the requirements engineering process cover four main

subareas: (I) Requirements Elicitation; (II) Require-

ments Analysis; (III) Requirements Speciﬁcation; and

(IV) Requirements Validation.

Requirements elicitation aims to understand the

problem that the software must solve using techniques

such as interviews, scenarios, prototyping, observa-

tions, user histories, among others. Requirements

analysis aims to detect and solve conﬂicts among the

requirements, to classify the requirements, to derive

new software requirements, and to establish how the

functionalities must interact with its organizational

and operational environment.

Requirements speciﬁcation, by its turn, produces

ICAART 2023 - 15th International Conference on Agents and Artiﬁcial Intelligence

388

requirements documents that can be systematically

reviewed, evaluated, and approved. Finally, require-

ments validation evaluates requirements documents to

ensure that the requirements are understandable, con-

sistent, and complete.

2.5 The Homer Technique

Homer (Human Oriented Method for Eliciting Re-

quirements), proposed by (Wilmann and Sterling,

2005), is a technique for eliciting requirements proper

to MAS. When applying this technique, the inter-

viewer asks the Stakeholders questions through or-

ganizational metaphors, as making the stakeholders

think in the software as a company that is hiring new

employees. This style of elicitation aims to more eas-

ily discover the agent roles and its goals within the

system, besides the system boundaries, business rules,

and relationships between the agents.

It is important to highlight that Homer assumes

that the use of multiagent systems is valid and that the

customer desires an agent-oriented solution. How-

ever, Homer does not perform a feasibility test to de-

termine whether a multiagent system is suitable for

solving the problem or not.

2.6 The Heraclito System

Heraclito is a multiagent system for the interactive

and dialectical learning with focus on the natural

deduction teaching of propositional logic (Galafassi

et al., 2019). This Intelligent Tutor System is based

in a Multiagent System and identiﬁes the individual

knowledge of each student in the context of Natural

Deduction in Propositional Logic.

Heraclito assists students in solving various types

of Logic exercises and provides the LOGOS Elec-

tronic Notebook to create and edit formulas, truth

tables and proofs of Propositional Logic (Galafassi

et al., 2019). We applied REPMAS in the require-

ments engineering of Heraclito second version.

3 RELATED WORKS

With the objective to understand the state-of-art of

the requirement engineering processes for multiagent

systems, mainly in what concerns to the coverage

of the requirements engineering subareas deﬁned in

SWEBOK (Bourque et al., 2014) and its support to

the BDI model (Rao et al., 1995), we performed a lit-

erature systematic review (MendonC¸ a. et al., 2021).

The systematic review analysed a total of 43 stud-

ies and did not ﬁnd any methodology that supports the

use of BDI model and cover the four requirements en-

gineering subareas deﬁned in SWEBOK. Moreover,

we perceived a strong lack, in a general way, in the

areas of requirements elicitation and validation.

This review was the basis for our proposal of

a MAS requirements engineering process, since we

search for acting in the gaps found in the current pro-

cesses and contributing to the area as a whole.

The support to the BDI model was mentioned in 8

processes, however, performing a deeper analysis, we

realized that only the PRACTIONIST process (Mor-

reale et al., 2006) supports the use of this model dur-

ing requirements engineering. Nevertheless, this pro-

cess only contemplates the requirements analysis and

speciﬁcation subareas, it does not support the elicita-

tion and validation phases.

4 REPMAS - A REQUIREMENTS

ENGINEERING PROCESS FOR

MULTIAGENT SYSTEMS

We identiﬁed several processes that cover require-

ments engineering for multiagent systems through our

literature systematic review (MendonC¸ a. et al., 2021).

The focus of our study was to observe the coverage of

requirements engineering of these processes and their

support for the BDI model in order to identify gaps in

both requirements engineering and BDI support.

From this systematic review, we identiﬁed the

need to create a requirements engineering process for

multiagent systems. Thus, we developed the REP-

MAS process with the aim of covering all the re-

quirements engineering subareas deﬁned in SWE-

BOK (Bourque et al., 2014) adapted speciﬁcally for

the context of multiagent systems.

It is important to highlight that our process also

aims to adapt requirements speciﬁc for MAS, such as

agent roles, perceptions, goals, and actions. In this

way, we intend to provide a greater support to require-

ments engineering for cognitive agents.

Another relevant point to be highlighted is that, al-

though this process is limited to Requirements Engi-

neering, our research group is working on techniques

and notations that can help to expand this process to

other areas of Software Engineering, such as the Soft-

ware Design subarea.

4.1 General Process Description

We believe it is important to highlight the choice of

using the Homer technique and the MASRML lan-

guage. We decided to use them in REPMAS because

REPMAS: A Requirements Engineering Process for MultiAgent Systems: An Application Example

389

both were created to ﬁll gaps in requirements engi-

neering for MAS. According to Willmann (Wilmann

and Sterling, 2005), Homer technique was conceived

due to the little attention AOSE devotes to require-

ments models and notations focused on elicitation.

And, according to Guedes (Guedes et al., 2020).

MASRML was created to solve the lack of mecha-

nisms to model functional requirements for multia-

gent systems.

Thus, as can be seen in Figure 1, REPMAS starts

in the requirements elicitation subprocess (detailed in

Figure 2), that generates the preliminary requirements

document. After eliciting the requirements, the ana-

lyst must identify the roles that agents can play. Agent

roles can be discovered by verifying which function-

alities can be described in the preliminary scenarios

and grouping these functionalities by speciality.

For each agent role, the subprocess of agents in-

ternal scenarios identiﬁcation (detailed in Figure 3) is

executed. In this subprocess are generated a partial

MASRML use-case diagrams (UCDs) for each agent

goal, resulting in the phase where the ﬁnal UCD is

produced. After producing the ﬁnal UCD, the docu-

ments created must be validated.

Figure 1: General Process.

4.2 Requirements Elicitation

Subprocess

Analysing the processes of requirements engineering

for multiagent systems, we veriﬁed that Homer tech-

nique (Wilmann and Sterling, 2005), stood out among

the analysed methodologies. This was the only tech-

nique that presented a way to elicit requirements fo-

cused directly with the stakeholder. Moreover, Homer

was made speciﬁcally to be integrated into other de-

velopment process, that is the reason why we decided

to integrate this technique in our process.

Another important point is that none methodol-

ogy presented an appropriate approach for the BDI

model, we veriﬁed that Tropos methodology (My-

lopoulos et al., 2013) supports BDI, however although

Tropos states to have an elicitation phase, it does not

present ways for eliciting requirements. Tropos elici-

tation phase is not described, this appears to be a kind

of free phase where it is allowed to its users to apply

the elicitation technique they want.

Thus, we understand that Homer technique is the

most suitable for collecting speciﬁc requirements for

multiagent systems and we integrate this technique

into our process. The requirements elicitation sub-

process can be seen in Figure 2.

The application of these questions intends to iden-

tify generic characteristics of the environment to ex-

tract the preliminar scenarios that will serve as a basis

to the agents internal scenarios identiﬁcation and to

the diagrams production.

Figure 2: Subprocess of Requirements Elicitation.

4.3 Subprocess of Agent Internal

Scenarios Identiﬁcation

The subprocess of agent internal scenarios identiﬁca-

tion (Figure 3), intends to extract agents speciﬁc char-

acteristics as well as documenting internal use cases

and producing partial use-cases diagrams.

For each agent role deﬁned previously, the analyst

must identify the initial beliefs associated with this

role and identify a goal for this agent role. With the

identiﬁcation of this goal, the analist must determine

the beliefs needed for the goal to become an intention,

identify perceptions needed for the identiﬁed goal, de-

termine which beliefs can be affected by the percep-

tions identiﬁed and the possible external actions asso-

ciated with the goal.

The analyst must verify whether for this goal is

necessary a plan. It is possible that a goal does not

have plans associated (in situations in which the agent

ICAART 2023 - 15th International Conference on Agents and Artiﬁcial Intelligence

390

Figure 3: Subprocess of Agent Internal Scenarios.

reactions are simple and demand little reasoning),

however it is also possible for a goal to have more

than a plan, it will depend on the goal complexity. If

it is realized that a plan is necessary, then it is needed

to identify the steps to accomplish that plan and the

external actions associated with the plan.

The subprocess is ﬁnished with the production of

a partial MASRML use-case diagram and the docu-

mentation of the scenarios relative to the use cases

that compose it. This subprocess presents how is

made the requirements specﬁcation with the applica-

tion of MASRML. This DSML allows to model BDI

agents, demonstrating its cogntive features.

It is important to highlight that the documentation

of MASRML scenarios presents a description of the

system functioning and agents’ characteristics, in ad-

dition to what is presented in the diagrams, such as

the agents’ beliefs.

5 PROCESS EXECUTION

This section has as its objective to present the exe-

cution of the proposed process in the developing of

the requirements engineering for the second version

of the Heraclito System.

5.1 Applying the Requirements

Elicitation

To verify the functioning and applicability of Homer,

we applied this technique in interviews with the stake-

holders of the Heraclito project. We hope that the re-

quirements engineering performed in this work will

contribute to the development of the second version

of Heraclito.

We performed three interviews with Heraclito sys-

tem stakeholders. It is important to highlight that the

conﬂicts found during the interviews were solved by

means of new interviews with the stakeholders.

The performed interviews allowed to extract the

positions in the organizational metaphor proposed and

its tasks in the enterprise. We also got to extract the

dependencies relations between the positions and its

limitations. Heraclito system does not have a position

hierarchy, however we believe that it would be pos-

sible to delimit a hierarchy with these questions for

other systems.

By means of applying Homer technique, we got

to extract the agent roles (deﬁned by positions in the

organizational metaphor) and the goals of these roles.

We also veriﬁed whether these goals depend on other

goals to be executed. For each goal, we also extracted

tasks for the goal to be achieved. Besides these char-

acteristics, the Homer technique would allow us to ex-

tract the hierarchy levels of each agent role, however,

we did not identify hierarchy levels in the agent roles

in the Heraclito system.

With the content elicited we were able to create a

preliminary description of Heraclito scenarios. These

scenarios allowed us to identify agent roles and to de-

velop the use-case models representing the system in-

ternal scenarios.

However, we would like to highlight that we were

not able to elicit the functionalities accessed by the

external agents (users), because Homer technique

does not cover the users actions. In this case, we

asked questions not included in Homer, with the ob-

jective to understand which functionalities could be

accessed by the system users.

5.2 Identiﬁcation Agent Roles

Applying Homer methodology questions allowed us

to acquire a description of the functions needed in the

organization.

When verifying the answers, we noticed that all

the interviewees presented three base-functions for

Heraclito. However, the interviewee 2 showed an-

swers very disconnected from the others. Analysing

his answers we realized that several parts of the sys-

REPMAS: A Requirements Engineering Process for MultiAgent Systems: An Application Example

391

tem were missing. On the other hand, the intervie-

wees 1 and 3, provided very similar answers, mostly

changing only the nomenclature.

Regarding the answers of the interviewee 2, we

made the decision to use those that were compatible

with those of the others interviewee and to validate in

other interviews those answers that were disconnected

with the answers of the others interviewee.

From the description of these functions, we delim-

ited the system agents. Next, analysing the answers,

not only to the questions, but also of the tasks de-

scribed, we got to identify similarities in the agents

described, deﬁning the following agent roles:

• Technical Specialist: It is responsible for all that

needs technical knowledge. This agent role pre-

pares the student content, corrects tests and ques-

tions, and follows the student in each action in

solving a question;

• Mediator: It is the single responsible for inter-

mediating the system with the student. It follows

the student development, presenting feedback and

providing clues. This agent role also knows the

student proﬁle;

• Bayesian Analyst: This agent role has knowl-

edge about the historical of the student questions

and tests. Based on this historical, the Bayesian

Analyst prepares a Bayesian Network containing

several statistics of the student results.

We also highlighted that the agents contained in

the Heraclito System are cognitive agents, i. e., these

agents present a mechanism of decision taking, ad-

vanced interactions, and have goals strongly estab-

lished.

5.3 Production of the Agents Internal

Scenarios

Aiming to demonstrate the operation of MASRML,

together with the process execution, in this section we

will present a scenario analysed from the Heraclito

System.

The partial use-case diagram (Figure 4), presents

the necessary steps for the Mediator Agent Role to

perform the ”Inform question to the Student” inter-

nal functionaly. The agent playing this role perceives

the solicitation of a new question and the goal ”In-

form question to the Student” became an intention.

This way, the plan ”Prepare to apply the question” is

triggered, ending with the presenting of the question

to the student. The description of the scenario steps

needed to perform this Goal can be viewed in the Ta-

bles 1, 2, 3, and 4.

Figure 4: Partial Use-Case Diagram - Inform a Question to

the Student.

Table 1: Goal - Inform Question to the Student.

Internal Use Case Name Inform Question to the Student.

Stereotype Goal

AgentRoleActor Mediator

Abstract This internal use case describes the steps followed

by the agent that plays the role of Mediator

to inform new study questions to the student.

Initial Beliefs Negative solicitation of a new question

Perceptions Probing of new solicitations of questions

Main Scenario

Actions to AgentRoleActor

1. Execute the internal use case ”Perceive new question request”

Alternative Scenario - Solicitation of a New Question Perceived

AgentRoleActor Actions

1. Change belief of Solicitation of a New Question to positive

2. Make the goal an intention.

3. Execute internal use case ”Prepare question application”

Table 2: Perception - Perceive a new question request.

Internal Use Case Name Perceive a new question request

Stereotype Perception

AgentRoleActor Mediator

Abstract This internal use case describes the probing

of the environment to verify

if a student asked for a new study question.

Pre-conditions The Goal Internal Use Case ”Inform

question to the student” must be in execution.

Initial Beliefs There was no request for a new question.

Main Scenario

AgentRoleActor Actions

1. Probe the environment to verify the occurrence of a new request of question.

Alternative Scenario - True Perception

AgentRoleActor Actions

1. Starts to believe that a student requested a new question.

5.4 Requirements Validation

In the requirements validation we veriﬁed the accep-

tance of the proposed scenarios. This review valida-

tion was applied after the ﬁrst version of the scenarios

speciﬁcation and served as a method to identify errors

in the scenario description, together with a validation

of the agent’s goals.

This validation was made by presenting the sce-

ICAART 2023 - 15th International Conference on Agents and Artiﬁcial Intelligence

392

Table 3: Perception - Perceive the submission at a new ques-

tion by the technical Specialist.

Internal Use Case Name Perceive the submission at a new question by the

Technical Specialist

Stereotype Perception

AgentRoleActor Mediator

Abstract This internal use case describes the probing of the

environment to verify if the Technical Specialist

sent a new question.

Pre-conditions The Plan Internal Use Case ”Prepare question

application” must be in execution.

Initial Beliefs Technical Specialist did not send a new question.

Main Scenario

AgentRoleActor Actions

1. Probe the environment to verify if a new question from the student was

sent by the Technical Specialist.

Alternative Scenario - New question perceived.

AgentRoleActor Actions

1. Starts to believe that the Technical Specialis sent a new question.

2. Receives a new question from the Technical Specialist.

Table 4: Plan - Prepare to apply the question.

Internal Use Case Name Prepare to apply the question

Stereotype Plan

AgentRoleActor Mediator

Abstract This internal use case describes the steps

followed by the agent that plays the Mediator

role to request a question to the Technical

Specialist to allow its application.

Main Scenario

AgentRoleActor Actions

1. Request a new question to the Technical Specialist

2. Execute internal use case ”Perceive the submission of a new question

by the Technical Specialist”

Alternative Scenario - Sending of new question from the student perceived

AgentRoleActor Actions

1. Apply student question.

narios to the stakeholders, asking if they were correct

and asking for suggestions. If the stakeholder found

an error or if there was something missing in the sce-

nario description, a comment suggesting alterations in

the scenario should be made.

With this validation we identiﬁed several errors

presented in the speciﬁcation ﬁrst version. Among

the errors found there was the lack of depth in some

goals and points not covered in the interviews.

Even containing errors, the validation gave us a

conﬁrmation that the elicitation and analysis were

satisfactory. The requirements elicited and analysed

were mostly correct and the stakeholders were satis-

ﬁed with the documentation produced.

5.5 Improvements Foressen to the

Process

Applying REPMAS in a real environment, allowed us

to extract some information that can contribute with

future works.

The technique Homer has several weaknesses,

most of them regarding its coverage or difﬁculty to

extract some particular requirements for MAS. How-

ever, we considered Homer organizational metaphor

very useful for extracting requirements for MAS.

Currently, in our research group, we are develop-

ing a technique for requirements elicitation based on

the organizational metaphor present in Homer tech-

nique. We intend to include this technique in the ini-

tial phase of this requirements engineering process, in

order to make it easier to obtain scenarios that will be

analyzed and speciﬁed with help of MASRML.

Another consideration we can quote is that the

agents’ beliefs are not expressed in the models. MAS-

RML use-case models, in spite of supporting partic-

ular MAS requirements, do not provide a way to ex-

press the agents’ beliefs, only in the use cases doc-

umentation. So it is necessary to perform a study

to verify if there are ways of belief modeling or if

it is needed to propose a method and/or notation for

it. This study is being performed in parallel with this

work and we are verifying the possibility to model

beliefs through conceptual models produced by ex-

tended UML class diagrams to represent agents’ be-

liefs, besides other structural information as goals,

perceptions, and plans representations.

Finally, from a wider perspective, we considered

that our work contributed to the requirements engi-

neering of the Heraclito project in a satisfactory way,

eliciting and analysing a wide range of speciﬁc re-

quirements for MAS, specially when dealing with

BDI agents.

6 CONCLUSIONS AND FUTURE

WORKS

In this study we proposed REPMAS - a require-

ments engineering process for multiagent systems.

The proposed process encompassed the four require-

ments engineering subareas inspired in the SWEBOK

(Bourque et al., 2014): elicitation, analysis, speciﬁca-

tion, and validation. With this process we tried to give

support to the software engineer when performing a

requirements engineering for this kind of system.

The elicitation stage of this process aims to dis-

cover speciﬁc requirements for MAS by means of

a direct contact with the Stakeholder, eliciting soft-

ware agents and its functions, by means of interview,

performing an organization metaphor with the hiring

of new employees for an enterprise. This elicitation

was performed by means of the Homer methodology

(Wilmann and Sterling, 2005). We have also made

REPMAS: A Requirements Engineering Process for MultiAgent Systems: An Application Example

393

an evaluation of this methodology execution, where

there were found several weak points that can serve

as a base for future works.

The analysis was proposed to be performed by

means of scenario identiﬁcation, produced by means

of MASRML (Guedes, 2012), a language that extends

UML use-case diagram for the context of multiagent

systems. This analysis aims to identify speciﬁc re-

quirements for MAS focusing on the support of the

BDI model. Finally, the validation phase of this pro-

cess aims to validate the artifacts speciﬁed in the elic-

itation and analysis phases.

To evaluate the proposed process, we applied it in

the second version of the Heraclito systems (Galafassi

et al., 2019) - a multiagent system that aims to aid in

logical teaching. The requirements engineering pre-

sented in this process can help in the development of

the second version of this system.

In the process execution, we performed interviews

with Heraclito Stakeholders, with the objective of

eliciting the requirements and to identify preliminar

scenarios for the system. These preliminar scenarios

served as the base to the elaboration of the preliminar

use case diagrams, in which we identiﬁed the goals,

perceptions, plans, and actions of the agent roles of

Heraclito System. We validated these scenarios by

means of a complete reading with the stakeholders.

Based on this study, we identiﬁed opportunities

for future work. One of them is related to a possi-

ble extension of the Homer methodology, to provide

a bigger support to particular requirements of MAS,

such as actions, perceptions, and plans. Another op-

portunity for future work is the need of a notation

that expresses the agents’ beliefs and how they can

change. Moreover, we are currently extending our

process in order to adapt other phases of software en-

gineering for the multi-agent systems development.

REFERENCES

Boes, J. and Migeon, F. (2017). Self-organizing multi-agent

systems for the control of complex systems. Journal

of Systems and Software, 134:12–28.

Bourque, P., Fairley, R. E., et al. (2014). Guide to the soft-

ware engineering body of knowledge (SWEBOK (R)):

Version 3.0. IEEE Computer Society Press, Washing-

ton, DC, USA.

Bratman, M. et al. (1987). Intention, plans, and practical

reason, volume 10. Harvard University Press Cam-

bridge, MA, Chicago, USA.

Ferber (2014). Handbook on agent-oriented design pro-

cesses, pref

acio. Springer, USA.

Galafassi, F. F. P., Galafassi, C., Vicari, R. M., and Gluz,

J. C. (2019). Her

aclito: Intelligent tutoring system for

logic. In International Conference on Practical Ap-

plications of Agents and Multi-Agent Systems, pages

251–254. Springer, Cham.

Gan, K., Anthony, P., On, C., and Hamdan, A. (2020).

Enforcing Social Semantic in FIPA-ACL Using SPIN,

pages 3–13. Springer, Singapore.

Guedes, G., Souza F., I., Gaedicke, L., Mendonc¸a, G. D.,

Vicari, R., and Brusius, C. (2020). Masrml - a domain-

speciﬁc modeling language for multi-agent systems

requirements. International Journal of Software En-

gineering & Applications (IJSEA), 11(5):21.

Guedes, G. T. A. (2012). Um metamodelo UML para a

modelagem de requisitos em projetos de sistemas mul-

tiagentes. PhD thesis, Universidade Federal do Rio

Grande do Sul (UFRGS).

Hajduk, M., Sukop, M., and Haun, M. (2019). Cognitive

Multi-agent Systems. Springer, USA.

Herzig, A., Lorini, E., Perrussel, L., and Xiao, Z. (2017).

Bdi logics for bdi architectures: old problems, new

perspectives. KI-K

unstliche Intelligenz, 31(1):73–83.

Javed, S., Alam, K. A., Ajmal, S., and Iqbal, U. (2022). Re-

quirements engineering education: A systematic lit-

erature review. In Proceedings of International Con-

ference on Information Technology and Applications,

pages 469–480. Springer.

Liu, G., Zheng, L., and Zhu, T. (2010). Cooperative plan-

ning process modeling on daily dispatching plan in

railway bureau based on multi-agent. In 2010 Inter-

national Conference on Intelligent System Design and

Engineering Application, volume 2, pages 616–621.

MendonC¸ a., G., Filho., I., and Guedes., G. (2021). A sys-

tematic review about requirements engineering pro-

cesses for multi-agent systems. In Proceedings of the

13th International Conference on Agents and Artiﬁ-

cial Intelligence - Volume 1: ICAART,, pages 69–79,

USA. INSTICC, SciTePress.

Morreale, V., Bonura, S., Francaviglia, G., Centineo, F.,

Cossentino, M., and Gaglio, S. (2006). Goal-oriented

development of bdi agents: the practionist approach.

2006 IEEE/WIC/ACM International Conference on

Intelligent Agent Technology, 1(10221975):66–72.

Mylopoulos, J., Castro, J., and Kolp, M. (2013). The Evo-

lution of Tropos, pages 281–287. Springer Berlin Hei-

delberg, Berlin, Heidelberg.

Rao, A. S. and Georgeff, M. P. (1995). Bdi agents: From

theory to practice. In in proceedings of the ﬁrst Inter-

national Conference on Multi-Agent Systems (ICMAS-

95, pages 312–319, San Francisco, CA, USA. IC-

MAS.

Rao, A. S., Georgeff, M. P., et al. (1995). Bdi agents: from

theory to practice. In ICMAS, volume 95, pages 312–

319.

Rodr

ıguez Viruel, M. L. (2011). Requirements Modeling for

Multi-Agent Systems, chapter 1, pages 3–22. InTech,

Asunci

on, Paraguay.

Slhoub, K., Carvalho, M., and Nembhard, F. (2019). Eval-

uation and comparison of agent-oriented methodolo-

gies: A software engineering viewpoint. In 2019 IEEE

International Systems Conference (SysCon), pages 1–

8, Orlando, FL, USA. IEEE.

ICAART 2023 - 15th International Conference on Agents and Artiﬁcial Intelligence

394

Vicari, R. M. and Gluz, J. C. (2007). An intelligent tutor-

ing system (its) view on aose. International Journal

of Agent-Oriented Software Engineering, 1(3-4):295–

333.

Wilmann, D. and Sterling, L. (2005). Guiding agent-

oriented requirements elicitation: Homer. In

Fifth International Conference on Quality Software

(QSIC’05), pages 419–424, Melbourne, VIC, Aus-

tralia. IEEE.

Wooldridge, M. (2009). An introduction to multiagent sys-

tems. John Wiley & Sons, USA.

REPMAS: A Requirements Engineering Process for MultiAgent Systems: An Application Example

395