An Inspection Technique Proposal for the Verification of Requirements
Specification Documents for Multi-Agent Systems
Giovane D’Avila Mendonc¸a
a
, Gilleanes Thorwald Araujo Guedes
b
and Iderli Pereira de Souza
Filho
c
Software Engineering Post-Graduation Program, Pampa Federal University, Av. Tiaraju, 810, Alegrete, Brazil
Keywords:
Requirements Engineering, Requirements Verification, Requirements Inspection, Multi-Agent Systems,
Perspective-Based Reading.
Abstract:
Requirements engineering is an important area of software engineering dedicated to eliciting, analysing, speci-
fying, and validating software requirements to ensure the correct understanding of what needs to be developed.
The requirements specification objective is to provide a detailed description of what the software must do, it
involves the production of a document that can be systematically reviewed, evaluated, and approved. Prob-
lems in the requirements are appointed among the main causes of failures in software projects. Therefore,
performing requirements verification and validation is extremely important to ensure the software quality.
Multi-Agent systems are a type of software with particular requirements, beyond the commonly found among
other systems, since they are composed by several autonomous and proactive agents that divide the problem
to be solved among them. Thus, requirements engineering needs to be adapted for this kind of system and the
produced documents need to be verified too. Several techniques were proposed for requirements inspection,
among them there is the Perspective-Based Reading. However, as in the other approaches, this technique does
not allow the inspection of particular requirements for multi-agent systems. Taking this in consideration, our
work has as its objective to adapt this technique so as to allow the verification of requirements specification
documents specific for this kind of system.
1 INTRODUCTION
Multi-Agent systems are characterized by being com-
posed of several agents that interact among them-
selves (Wooldridge, 2009). An agent is an au-
tonomous, proactive, reactive, and with social skills
process (Vicari and Gluz, 2007), moreover, an agent
is able to performing actions without the user inter-
vention (Wooldridge and Jennings, 1995) (Pornphol
and Chittayasothorn, 2006) (Sivakumar et al., 2013).
Multi-Agent systems became an alternative for
the complex systems development (Labba et al.,
2015). However, the development of this kind of
system brought several challenges to the software
engineering area, what entailed the emergence of
AOSE (Agent-Oriented Software Engineering) an
area that mixes characteristics of several disciplines
both from Software Engineering and Artificial Intel-
a
https://orcid.org/0000-0001-9445-6831
b
https://orcid.org/0000-0001-5457-2600
c
https://orcid.org/0000-0001-9694-0977
ligence (Cuesta et al., 2007). AOSE has, among its
objectives, to adapt software engineering practices
specifically for the agent-based systems development
(Guedes and Vicari, 2010), that includes to produce
methodologies, processes, techniques, modeling lan-
guages, and tools for the multi-agent systems devel-
opment (Genza and Mighele, 2013) (Mendonc¸a et al.,
2021). How could it be, AOSE also includes the adap-
tation of Requirements Engineering for multi-agent
systems.
Requirements Engineering is a Software Require-
ments area that is concerned with requirements elic-
iting, analysing, specifying, and validating to ensure
the correct understanding of what needs to be devel-
oped (Fuentes-Fern
´
andez et al., 2009). Requirements
Engineering performs a crucial function for the devel-
opment of any software, since, if the software needs
were not understand correctly, the product will not
satisfy those to whom its is destined (Mendonc¸a et al.,
2021).
The main failures verified in software projects are
related to problems in the requirements specification,
Mendonça, G., Guedes, G. and Filho, I.
An Inspection Technique Proposal for the Verification of Requirements Specification Documents for Multi-Agent Systems.
DOI: 10.5220/0010842700003116
In Proceedings of the 14th International Conference on Agents and Artificial Intelligence (ICAART 2022) - Volume 1, pages 257-264
ISBN: 978-989-758-547-0; ISSN: 2184-433X
Copyright
c
2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
257
due, mainly, to difficulties in the understanding of
the users needs (Koscianski and dos Santos Soares,
2007). Therefore, to perform the requirements veri-
fication and validation is essential to ensure the soft-
ware quality, because errors identified in the require-
ments in posterior phases of the software develop-
ment has a very much bigger cost to be corrected (Bi-
lal et al., 2016). Thus, several inspection techniques
were created to the requirements verification and val-
idation. Inspections had proved to be an effective way
to find failures in different software artifacts (Fogel-
str
¨
om and Gorschek, 2007).
Taking this in consideration, we are proposing
an adaptation of the Perspective-Based Reading tech-
nique (PBR) (Basili et al., 1996) (Shull et al., 2000)
in such a way that it can be applied in the inspection
of software requirements specification documents
(SRSD) specific for multi-agent systems. Whereas
Ebad in (Ebad, 2017), highlights that the Perspective
Based Reading (PBR) technique is one of the most
effective for requirements inspection. This technique
adapted was conceived to be applied on SRSDs pro-
duced by a specific requirements engineering process
for multi-agent systems we developed. The docu-
ments produced by this process use the notation of
Multi-Agent Systems Requirements Modeling Lan-
guage (MASRML) proposed by (Guedes et al., 2020).
We use MASRML because this notation supports
Belief-Desire-Intention (BDI) model, during the re-
quirements specification, different from other nota-
tions that do not support this type of representation.
Furthermore, the MASRML notation uses some con-
cepts, as by example Internal Use Cases (IUC) and
AgentRoleActors.
In this study context, we propose a verification
technique, considering the definition presented in the
study of (Ryan and Wheatcraft, 2017), in which the
use of verification and validation terms and their
definitions according to the literature was analyzed.
In this study, the author defines that the verifica-
tion focus is on the requirements text and structure,
determining whether they are correctly written and
structured according to the organization’s standards,
guidelines, rules, and checklists. On the other hand,
requirements validation makes it possible to deter-
mine whether the requirements are defined in a clear
way and whether they communicate correctly the
stakeholders expectations and needs. It involves es-
tablishing whether the right thing is being built, as
defined by the requirements set.
In order to assess the applicability and effi-
ciency of our proposal to adapt the Perspective-
Based Reading technique, we are applying two quasi-
experiments. The first with students of Verification
and Validation discipline of a Software Engineering
undergraduate course. The second quasi-experiment
is being applied to specialists, master’s students, and
final year students of an undergraduate software en-
gineering course. Most of the subjects have profes-
sional experience in the software engineering area.
Due to the limited space in this work, the results
of the quasi-experiments will be published in a future
work.
2 BACKGROUND
2.1 Requirements Engineering
The objectives of Requirements Engineering are: (I)
identifying software requirements; (II) analysing re-
quirements so to classify them and to derive ad-
ditional requirements, as well as to solve conflicts
among them; (III) documenting the requirements;
(IV) validating the documented requirements (Beren-
bach et al., 2009).
In SWEBOK (Bourque et al., 2014) – a reference
book in the area – it is stated that the requirements en-
gineering process covers four main subareas, (I) Re-
quirements Elicitation; (II) Requirements Analysis;
(III) Requirements Specification; and (IV) Require-
ments Validation.
Requirements elicitation investigates how to ex-
tract requirements and which are its origins. Require-
ments analysis aims to detect and to solve conflicts
among the requirements and to discover the system
limits. Requirements specification, by its turn, pro-
duces requirements documents that can be systemati-
cally reviewed, evaluated, and approved. Finally, re-
quirements validation evaluates the requirements doc-
uments to ensure the requirements are understand-
able, correct, consistent, and complete.
2.2 Requirements Verification
Weak alignment of requirements engineering with
verification can lead to problems in the deliver of soft-
ware products, such as lack in the accomplishments of
deadlines and poor quality (Bjarnason et al., 2014).
According to (Dzida and Freitag, 1998) verification is
related to the requirements correctness.
Requirements verification is the confirmation by
examination that the requirements (individually or in
a set) are well formed (ISO/IEC/IEEE 29148, 2011)
(ISO/IEC/IEEE 29148, 2018). According to (Ryan
and Wheatcraft, 2017), this means that a requirement
or a set of requirements has been revised to ensure that
the characteristics of good requirements are achieved.
ICAART 2022 - 14th International Conference on Agents and Artificial Intelligence
258
According to (Khan et al., 2015), experts be-
lieve that review and inspection are the best strate-
gies to remove or mitigate requirements verification
and validation challenges. In this sense, (Ryan and
Wheatcraft, 2017) states that requirements verifica-
tion confirms, through inspection, that the require-
ments contain the necessary elements and have the
characteristics of a well-formed requirement.
2.3 Agents and Multi-Agent Systems
An agent is a process situated in an environment
designed to achieve a purpose by means of an au-
tonomous and flexible behavior. The environment is
the application domain where the agent will act (Vi-
cari and Gluz, 2007).
A multi-agent system (MAS) consists of a set of
agents that can interact among themselves. Agents
are able to act in an environment and have different
“influence spheres”, where they can control or influ-
ence different parts of the environment (Wooldridge,
2009).
Agents use to play roles. An agent role contains a
part of the social behavior of an agent and it is char-
acterized by having a goal and/or providing a service.
The goal of each role is to contribute to the accom-
plishment of a part of the organization (or environ-
ment) requirements in which the agent is contained
(Cossentino and Seidita, 2014). The agent roles rep-
resent the functions the agents can perform inside the
system. The agents can take more than one role, but
generally not at the same time.
2.4 Internal Use Cases and
AgentRoleActors
The specification proposed in our work uses the no-
tation established in MASRML (Guedes et al., 2020)
language, a modeling language extended from UML
specifically to the multi-agent systems requirements
representation.
The AgentRoleActors represent roles taken by
agents. These agent roles are modeled inside the sys-
tem boundary. Internal Use Cases (IUC) are use cases
that are not accessed by external actors. IUCs are
accessed by Agent Roles (AgentRoleActors). These
IUCs can have stereotypes of the Goal, Plan, Percep-
tion, or Action kind.
2.5 Belief-Desire-Intention Model
The Belief-Desire-Intention (BDI) model is a soft-
ware model developed for programming intelligent
agents. It includes beliefs, desires, and intentions in
the agent architecture (Bratman et al., 1987).
Beliefs represent the information state the agent
has about the environment, about himself, and even
about the other agents. The desires represent the goals
or situations that the agent wants to achieve. Lastly,
intentions represent the desires the agent believes he
can achieve and acts to achieve it (Rao and Georgeff,
1995).
According to (Bordini et al., 2021) asserts that the
unique strengths of BDI agent languages provide an
ideal framework for integrating the wide range of AI
capabilities needed to progress towards the next gen-
eration of intelligent systems.
BDI agents are among the most widely studied
models of rational agents (Torre and Parlato, 2020).
The Belief-Desire-Intention model is one of the most
popular models for developing rational agents based
on how humans act and on information derived from
an environment (Ujjwal and Chodorowski, 2019).
Furthermore, according to (Xu et al., 2018), the
BDI architecture, where agents are modeled based on
their beliefs, desires, and intentions, provides a practi-
cal approach to developing intelligent agent systems.
2.6 Multi-Agent Systems Requirements
Modeling Language
MASRML Multi-Agent Systems Requirements
Modeling Language is a UML-based Domain-
Specific Modeling Language (DSML) conceived for
requirements modeling in multi-agent system projects
(Guedes et al., 2020). This DSML extends the UML
metamodel in order the use-case diagram can be ap-
plied in the specific domain of multiagent systems.
MASRML provides mechanisms to represent the con-
cepts of agent role, goal (desire), perception, belief,
intention, plan, and action, supporting the concepts of
the BDI model.
In MASRML, new concepts were created inspired
by the concepts of the standard UML. The main
contribution was the creation of the AgentRoleActor
and InternalUseCase metaclasses to represent agent
roles and the internal functionalities assigned to them,
stereotyped in goals, perceptions, actions, and plans.
Some associations were also created in order to as-
sociate perceptions, actions, and plans to the goals
that an agent role wishes to achieve, thus establish-
ing 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
executing a plan.
In addition, MASRML internal use-cases docu-
mentation provides a clear view of the structure of
An Inspection Technique Proposal for the Verification of Requirements Specification Documents for Multi-Agent Systems
259
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.
3 RELATED WORKS
Several methodologies were proposed supporting re-
quirements engineering. However, the requirements
validation is a subarea that has been neglected in these
methodologies. Based on the study of (Mendonc¸a
et al., 2021), in which, by means of a systematic re-
view sought to establish the state-of-art of methodolo-
gies or processes that covers requirement engineering
for multi-agent systems, we observed that only three
studies support the requirements validation subarea.
Although these three studies ((Cysneiros and Yu,
2002), (Haumer et al., 1999), (Bonjean et al., 2014))
contain in its phases the requirements validation
subarea, none of them proposed a specific valida-
tion technique for multi-agent systems requirements.
Thus, is evident the need of proposing a new approach
of requirements validation for this kind of system.
Several techniques have been proposed for in-
specting traditional systems, among which we can
quote Fagan in (Fagan, 1976), who proposed the use
of checklists for defect detection; Porter and Votta in
(Porter and Votta, 1994), who presented a scenario-
based technique that provides inspectors with more
specific instructions than checklists, allowing them to
classify defects and develop questions for each type
of defect; Thelin et al. in (Thelin et al., 2004) who in-
troduced the Usage-Based Reading (UBR), in which
the inspectors use a predefined list of prioritized use
cases; UBR-ir (Thelin et al., 2003), a variation of
UBR that allows inspectors to classify use cases indi-
vidually at the beginning of the inspection process ac-
cording to the inspectors’ experience and understand-
ing. Dunsmore et al. in (Dunsmore et al., 2003),
who presented Use-Case Reading (UCR) to inspect
object-oriented systems considering dynamic interac-
tion with collaborative objects.
However, these techniques were proposed to be
applied in traditional systems, therefore they can-
not be fully used in the requirements verification for
multi-agent systems due to the specific features of this
type of system.
4 PROPOSED ADAPTATION OF
PERSPECTIVE-BASED
READING TO INSPECT
REQUIREMENTS
SPECIFICATION DOCUMENTS
FOR MULTI-AGENT SYSTEMS
The study presented by Ebad in (Ebad, 2017), eval-
uated the evidence about the efficacy of the reading
techniques in the inspection of artifacts produced dur-
ing the requirements, project, and coding phases. In
the requirements phase (requirements engineering),
Ebad states that the ad hoc technique is the simplest
of the techniques, but in this technique the inspec-
tors do not follow guidelines or orientations. In this
same study, the author highlights that the Perspective-
Based Reading (PBR) technique is one of the most
effective for requirements inspection.
PBR provides a set of procedures that can help
developers to solve software requirements inspection
problems. PBR reviewers represent specific stake-
holders interested in the document (as designers or
testers) to verify the requirements quality (Shull et al.,
2000). The combination of different perspectives re-
sult in a better analysis of the requirements specifica-
tion document, making it possible that it can be com-
pletely checked (Pagliuso et al., 2002).
The PBR technique provides questions developed
specifically for each step of procedure and this way
it creates representations in order the reviewers can
answer a series of questions about the work product
(Pagliuso et al., 2002). Each inspector must define
his interest in the document so that it is inspected ac-
cording to the perspective of a specific involved part.
Thus, considering the evidences pointed by Ebad
in (Ebad, 2017) that PBR is one of the more efficient
techniques for requirements inspection and consider-
ing the statement of Shull in (Shull et al., 2000) that,
depending on the environment where PBR is applied,
it can be find a different set of more applicable per-
spectives, we decided to adapt this technique to the
inspection of a requirements specification document
produced based on the ISO/IEC/IEEE 29148:2018
standard (ISO/IEC/IEEE 29148, 2018) extended for
the multi-agent system dominion.
That said, in this work, we are proposing the
Agent Simulator Perspective to SRSD inspecting con-
sidering multi-agent systems requirements specific
characteristics. This perspective aims to determine
whether the requirements describe properly the func-
tionalities to be performed by the agents. This per-
spective inspects Internal Use Cases (IUC), evaluat-
ing if the main, alternative, and exception scenarios
ICAART 2022 - 14th International Conference on Agents and Artificial Intelligence
260
are described correctly, as well as if the pre-conditions
are defined.
Moreover, basing on the Agent Simulator perspec-
tive, it is possible to verify whether the agent roles
are identified and detailed, whether the initial beliefs
are described, whether the perceptions are associated
with the goals, and to verify the correctness of the
stereotypes or associations of the internal use cases.
The Agent Simulator Perspective follows a se-
quence of steps. Through these steps, the inspector
simulates the agents behaviors searching for errors or
failures in the requirements specification document.
This process can be observed in Figure 1.
As can be seen in Figure 1 the inspection process
has a sequence of activities, however, due to lack of
space, in this paper we will discuss only the activities
performed by the inspector.
After receiving the SRSD, the inspector will per-
form a sequence of six steps. Each of these steps con-
tains a verification list that must be followed during
the SRSD inspection. These steps correspond to I)
general inspection of the SRSD; II) general inspec-
tion of each IUC; III) inspection of the internal use
cases of Goal type; IV) inspection of the internal use
cases of Perception type; V) inspection of each partial
use-case diagram for each AgentRoleActor; and VI)
inspection of the general use-case diagram.
The activities performed by the requirements en-
gineering team and by the inspection manager are
similar to the traditional inspection process, however,
the differences between the traditional process and
ours are the activities performed by the inspector.
In our process, the inspector takes the agent per-
spective as discussed previously, so the SRSD will
be inspected considering the specific characteristics
of multi-agent systems, following the checklists that
we are developing for each of the inspection steps.
Next we will describe every step of our inspection
technique proposal.
4.1 First Step
In this step, the requirements specification document
is inspected in a general way. The inspector will
search for completeness, correctness, and consistency
failures types. Some of them are related to the defini-
tion of agent roles, the functionalities associated with
the agents, communication patterns used, and about
the beliefs identified.
4.2 Second Step
In this step, the inspector must analyse each IUC
searching for types of failures like incorrectness, am-
biguity, untraceability, incompleteness, and inconsis-
tency. Some of these failures are related to the general
structure of each IUC, lack of main scenarios, inter-
pretation ambiguities, or incorrectness in the use-case
scenario texts.
4.3 Third Step
In this step, the inspector must analyse each one of
the internal use cases of the Goal type, searching for
types of failures like correctness, consistency, and
completeness. Some of these failures are related to
stereotype correctness, clarity in the perceptions de-
scription that turns a goal into an intention, or the ex-
istence of perceptions associated with a goal.
4.4 Fourth Step
In this phase, the inspector will analyse each one of
the internal use cases of the Perception type, search-
ing for types of failures like correctness and complete-
ness. Some of these failures are related to stereotype
incorrectness, lack of identification or clarity of the
initial beliefs, and lack of pre-conditions definition.
4.5 Fifth Step
In this step, the inspector must analyse each partial
use case diagram developed for each AgentRoleAc-
tor. The inspector will search for failure types like
correctness and consistency. Some of them are re-
lated to stereotypes incorrectness, incorrectness of the
associations among the diagram components, and in-
correctness of the associations direction.
4.6 Sixth Step
Lastly, in this step, the inspector must analyse the gen-
eral use-case diagram, searching to locate failures of
correctness and completeness type. Some of them are
related to agent roles stereotypes incorrectness, agent
roles and external users identified in incorrect places,
and the lack of attribution of at least one goal to each
agent role.
5 CONCLUSION AND FUTURE
WORKS
The requirements verification and validation is an im-
portant phase of the requirements engineering and
plays a crucial role in the successful of any project
(Gupta et al., 2019). During the verification and val-
idation phase the requirements specification must be
An Inspection Technique Proposal for the Verification of Requirements Specification Documents for Multi-Agent Systems
261
Figure 1: Inspection Process of Multi-Agent Systems Requirements.
examined to ensure that all the requirements had been
declared in a non ambiguous way and free of incon-
sistencies, omissions, and errors (Pressman, 2011).
Several techniques were presented for require-
ments inspecting, as for example, the checklist-based
reading, scenarios-based inspection, perspective-
based reading (PBR), use-cases reading (UCR), and
Use-Based Reading (UBR) (Fagan, 1976) (Porter and
Votta, 1994) (Basili et al., 1996) (Dunsmore et al.,
2003) (Thelin et al., 2004).
However, all these techniques were applied to
inspect traditional systems, mainly object-oriented.
Nevertheless, according to Padmanaban et al. in (Pad-
manaban et al., 2016) conventional object-oriented
software tests cannot be applied to agent-oriented sys-
tems due the agents’ features, such as autonomy, pro-
activity, social capabilities, reactivity, and mobility.
Moreover, the new concepts and properties that arise
with the use of the agent paradigm, implies in the
construction of new verification and validation ap-
proaches, different from the conventional software
engineering (Fuentes-Fern
´
andez et al., 2004).
Thus, we proposed in this work a verification tech-
nique for a kind of SRSD produced based on an exten-
sion of ISO/IEC/IEEE 29148:2018 standard and on
the MASRML notation. To develop this verification
technique we adapted the Perspective-Based Reading
technique, creating the Agent-Simulator perspective.
PBR must be adaptable to the particular docu-
ment it will be inspecting and to the notation used
to produce artifacts and diagrams contained in the
document. Moreover, any technique, including PBR
should be adapted to the particularities of a given en-
vironment in order to be more successful (Ciolkowski
et al., 1997).
To evaluate the applicability and efficiency of our
proposed inspection technique, we are applying our
technique in two quasi-experiments. The inspection
is being carried out from a document of requirements
specification produced based on the ISO/IEC/IEEE
29148:2018 standard, that was extended to support
requirements for multi-agent systems. This extended
standard uses MASRML notation. The results of the
quasi-experiments will be published in a future work.
As a partial result of the first quasi-experiment, we
noticed that, of the 23 failures out of the 25 injected
into the Software Requirements Specification Docu-
ment (SRSD), about 92%, were discovered by at least
one of the twelve subjects who participated in the in-
spection.
Therefore, we concluded that the partial results
obtained in the first quasi-experiment can indicate
the efficiency and applicability of our adaptation of
the PBR technique for the inspection of specifica-
tion requirements documents for multi-agent systems.
Moreover, the quasi-experiment contributed to in-
crease the experience in software inspections of the
students participants.
We are finalizing the tabulation of the data
collected during the execution of the two quasi-
experiments. We intend to publish the full results in
future works.
ICAART 2022 - 14th International Conference on Agents and Artificial Intelligence
262
Furthermore, to complement our PBR adaptation
proposal, we are developing checklists for each of the
six steps presented in section 4. These checklists will
have a sequence of pre-defined questions intended to
guide the inspectors through the evaluation process of
the requirements specification document.
This work is part of a bigger project that aims to
propose a complete requirements engineering process
for the multi-agent systems development. Thus, this
study contributes to the proposition of a new approach
for multi-agent systems requirements verification.
Our inspection proposal was developed for in-
specting SRSDs produced by a specific requirements
engineering process. These documents are structured
according to the standard ISO/IEC/IEEE 29148:2018
extended to the multi-agent systems context and its
use-case diagrams and documentation are produced
by means of the notation provided by MASRML
(Guedes et al., 2020). However, we believe our in-
spection technique proposal can be adapted to in-
specting other SRSDs produced by other processes of
requirements engineering for multi-agent systems.
As a future work we are developing a guide with
application orientations about our inspection tech-
nique. Thus, we also intend to add this guide to the
requirements engineering process currently in devel-
opment.
REFERENCES
Basili, V. R., Green, S., Laitenberger, O., Lanubile, F.,
Shull, F., Sørumg
˚
ard, S., and Zelkowitz, M. V. (1996).
The empirical investigation of perspective-based read-
ing. Empirical Software Engineering, 1:133–164.
Berenbach, B., Paulish, D., Kazmeier, J., and Rudorfer, A.
(2009). Software & systems requirements engineer-
ing: in practice. McGraw-Hill, Inc.
Bilal, H. A., Ilyas, M., Tariq, Q., and Hummayun, M.
(2016). Requirements validation techniques: An em-
pirical study. International Journal of Computer Ap-
plications, 148(14):5–10.
Bjarnason, E., Per Runeson, M. B., Unterkalmsteiner, M.,
Engstr
¨
om, E., Regnell, B., Sabaliauskaite, G., Locon-
sole, A., Gorschek, T., and Feldt, R. (2014). Chal-
lenges and practices in aligning requirements with
verification and validation: a case study of six com-
panies. Empirical Software Engineering.
Bonjean, N., Mefteh, W., Gleizes, M. P., Maurel, C.,
and Migeon, F. (2014). ADELFE 2.0, pages 19–63.
Springer Berlin Heidelberg, Berlin, Heidelberg.
Bordini, R. H., El Fallah Seghrouchni, A., Hindriks, K., Lo-
gan, B., and Ricci, A. (2021). Agent programming in
the cognitive era. In Proceedings of the 20th Interna-
tional Conference on Autonomous Agents and Multi-
Agent Systems, AAMAS ’21, page 1718–1720, Rich-
land, SC. International Foundation for Autonomous
Agents and Multiagent Systems.
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.
Bratman, M. et al. (1987). Intention, plans, and practical
reason, volume 10. Harvard University Press Cam-
bridge, MA. 1-57586-192-5.
Ciolkowski, M., Differding, C., Laitenberger, O., and
M
¨
unch, J. (1997). Empirical investigation of
perspective-based reading: A replicated experiment.
International Software Engineering Research Net-
work (ISERN).
Cossentino, M. and Seidita, V. (2014). PASSI: Process
for agent societies specification and implementation,
pages 287–329. Handbook on Agent-Oriented Design
Processes.
Cuesta, P., G
´
omez, A., and Gonz
´
alez, J. C. (2007). Agent
oriented software engineering. Issues in Multi-Agent
Systems.
Cysneiros, L. and Yu, E. (2002). Requirements engineer-
ing for large-scale multi-agent systems. volume 2603,
pages 39–56.
Dunsmore, A., Roper, M., and Wood, M. (2003). The devel-
opment and evaluation of three diverse techniques for
object-oriented code inspection. IEEE Trans. Softw.
Eng., 29(8):677–686.
Dzida, W. and Freitag, R. (1998). Making use of scenarios
for validating analysis and design. IEEE Transactions
on Software Engineering, 24(12):1182–1196.
Ebad, S. A. (2017). Inspection reading techniques applied
to software artifacts - a systematic review. Comput.
Syst. Sci. Eng., 32.
Fagan, M. E. (1976). Design and code inspections to reduce
errors in program development. IBM Systems Journal,
15(3):182–211.
Fogelstr
¨
om, N. and Gorschek, T. (2007). Test-case driven
versus checklist-based inspections of software re-
quirements - an experimental evaluation. In Proceed-
ings of the 10th Workshop on Requirements Engineer-
ing, pages 116–126.
Fuentes-Fern
´
andez, R., G
´
omez-Sanz, J. J., and Pav
´
on, J.
(2004). Verification and validation techniques for
multi-agent systems. Upgrade, 5:15–19.
Fuentes-Fern
´
andez, R., G
´
omez-Sanz, J. J., and Pav
´
on, J.
(2009). Requirements elicitation and analysis of mul-
tiagent systems using activity theory. IEEE.
Genza, N. and Mighele, E. (2013). Review on multi-agent
oriented software engineering implementation. Inter-
national Journal of Computer and Information Tech-
nology, 2.
Guedes, G., Souza Filho, I., Gaedicke, L., Mendonc¸a, G.,
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).
Guedes, G. T. A. and Vicari, R. M. (2010). A uml profile
oriented to the requirements modeling in intelligent
tutoring systems projects. In Bramer, M., editor, Ar-
tificial Intelligence in Theory and Practice III, pages
An Inspection Technique Proposal for the Verification of Requirements Specification Documents for Multi-Agent Systems
263
133–142, Berlin, Heidelberg. Springer Berlin Heidel-
berg.
Gupta, A., Deraman, A., and Siddiqui, S. (2019). Bdi logics
for bdi architectures: old problems, new perspectives.
ARPN Journal of Engineering and Applied Sciences,
14(17):3046–3061.
Haumer, P., Heymans, P., Jarke, M., and Pohl, K. (1999).
Bridging the gap between past and future in re: a
scenario-based approach. In Proceedings IEEE In-
ternational Symposium on Requirements Engineering
(Cat. No.PR00188), pages 66–73.
ISO/IEC/IEEE 29148 (2011). Iso/iec/ieee interna-
tional standard - systems and software engineering
life cycle processes –requirements engineering.
ISO/IEC/IEEE 29148, pages 1–94.
ISO/IEC/IEEE 29148 (2018). Iso/iec/ieee international
standard - systems and software engineering
life cycle processes requirements engineering.
ISO/IEC/IEEE 29148:2018(E), pages 1–104.
Khan, H., Asghar, I., Ghayyur, S., and Raza, M. (2015).
An empirical study of software requirements verifica-
tion and validation techniques along their mitigation
strategies. pages 2321–5658.
Koscianski, A. and dos Santos Soares, M. (2007). Quali-
dade de Software-2
a
Edic¸
˜
ao: Aprenda as metodolo-
gias e t
´
ecnicas mais modernas para o desenvolvi-
mento de software. Novatec Editora.
Labba, C., Bellamine ben Saoud, N., and Dugdale, J.
(2015). Towards a conceptual framework to support
adaptative agent-based systems partitioning. In 2015
IEEE/ACIS 16th SNPD, pages 1–5.
Mendonc¸a, G., Souza, Filho, I., and Guedes, G. (2021).
A systematic review about requirements engineering
processes for multi-agent systems. In 13th Inter-
national Conference on Agents and Artificial Intelli-
gence (ICAART 2021), volume 1, pages 69–79.
Padmanaban, R., Thirumaran, M., Suganya, K., and Priya,
R. (2016). Aose methodologies and comparison of
object oriented and agent oriented software testing.
pages 1–16.
Pagliuso, P., Tambascia, C., and Villas-Boas, A. (2002).
Melhoria da inspec¸
˜
ao de requisitos segundo a t
´
ecnica
de leitura baseada em perspectiva. XI Semin
´
ario de
Computac¸
˜
ao - XI SEMINCO, 32.
Pornphol, P. and Chittayasothorn, S. (2006). An agent-
based quality assurance assessment system. 5th
WSEAS International Conference on E-ACTIVITIES.
Porter, A. and Votta, L. (1994). An experiment to assess
different defect detection methods for software re-
quirements inspections. In Proceedings of 16th Inter-
national Conference on Software Engineering, pages
103–112.
Pressman, R. (2011). Software Engineering: A Practi-
tioner’s Approach. Bookman.
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.
Ryan, M. and Wheatcraft, L. (2017). On the use of the
terms verification and validation. INCOSE Interna-
tional Symposium, 27:1277–1290.
Shull, F., Rus, I., and Basili, V. (2000). How perspective-
based reading can improve requirements inspections.
Computer, 33(7):73–79.
Sivakumar, N., Vivekan, K., and Kanimozhi, M. (2013).
Testing in prometheus methodology plan oriented
approach.
Thelin, T., Runeson, P., and Wohlin, C. (2003).
An experimental comparison of usage-based and
checklist-based reading. IEEE Trans. Softw. Eng.,
29(8):687–704.
Thelin, T., Runeson, P., Wohlin, C., Olsson, T., and
Andersson, C. (2004). Evaluation of usage-
based reading—conclusions after three experiments.
9(1–2):77–110.
Torre, S. L. and Parlato, G. (2020). A fixed-point model-
checker for bdi logics over finite-state worlds. In
OVERLAY.
Ujjwal, K. and Chodorowski, J. (2019). A case study of
adding proactivity in indoor social robots using be-
lief–desire–intention (bdi) model. Biomimetics, 4(74).
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.
Wooldridge, M. (2009). An introduction to multiagent sys-
tems. John Wiley & Sons.
Wooldridge, M. and Jennings, N. R. (1995). Intelligent
agents: Theory and practice. The knowledge engineer-
ing review, 10(2):115–152.
Xu, M., Bauters, K., McAreavey, K., and Liu, W. (2018).
A formal approach to embedding first-principles plan-
ning in BDI agent systems. In Ciucci, D., Pasi, G.,
and Vantaggi, B., editors, Scalable Uncertainty Man-
agement - 12th International Conference, SUM 2018,
Milan, Italy, October 3-5, 2018, Proceedings, volume
11142 of Lecture Notes in Computer Science, pages
333–347. Springer.
ICAART 2022 - 14th International Conference on Agents and Artificial Intelligence
264