UNIVERSIDADE FEDERAL DO RIO GRANDE DO SUL
INSTITUTO DE INFORMÁTICA
PROGRAMA DE PÓS-GRADUAÇÃO EM COMPUTAÇÃO
DEFESA DE TESE DE DOUTORADO
Aluno: Fernando dos Santos
Orientadora: Profª. Drª. Ana Lucia Cetertich Bazzan – videoconferência
Coorientadora: Profª. Drª. Ingrid Oliveira De Nunes
Título: Model-Driven Agent-based Simulation Development
Linha de Pesquisa: Planejamento, Sistemas Multiagentes e Robótica
Local: Prédio 43412 – Auditório 1 (AUD-1) do Instituto de Informática da UFRGS.
– Prof. Dr. Jaime Simão Sichman (USP – por videoconferência)
– Profª. Drª. Diana Francisca Adamatti (FURG– por videoconferência)
– Profª. Drª. Leila Ribeiro (UFRGS)
Presidente da Banca: Profª. Drª. Ana Lucia Cetertich Bazzan
Abstract: The Agent-based Modeling and Simulation (ABMS) paradigm has been used to analyze,reproduce, and predict phenomena in many application areas, such as traffic and epi-demiology Building agent-based simulations is a challenging task, to which researchershave already stressed the importance of providing building blocks and tools at a higherabstraction level to promote expressiveness and productivity. Model-driven Develop-ment (MDD) is an approach to software development, in which high-level modelingartifacts drive the production of effort-consuming low-level artifacts, such as source code.Previous studies on the use ofMDDin mainstream software development have alreadyshown that it significantly increases productivity. However, in theABMSparadigmMDDhas been exploited in a limited way. Most of the existing proposals consider modelingand code generation of just a few simulation aspects, leaving much left to be implementedby developers. Nevertheless, there is a lack of empirical studies that demonstrate whethertheseMDDapproaches are indeed effective. In this thesis, we exploitMDDin thecontext ofABMS. We propose MDD4ABMS, a model-driven approach for developingagent-based simulations. MDD4ABMS is composed of the following elements, whichare the main contributions of this thesis: (i) a metamodel for agent-based simulationsthat was built following a bottom-up approach so as to abstract aspects recurrently usedin simulations and make them available for modeling; (ii) a domain-specific modelinglanguage with building blocks to instantiate agent-based simulation models; and (iii)model-to-code transformations to generate source code for NetLogo, a widely usedagent-based simulation platform. While abstractions provided by the metamodel allowdevelopers to focus onwhichfeatures to consider in simulations instead ofhowtoimplement and integrate them (a task that may introduce inconsistencies in simulations),building blocks provided by the modeling language promote expressive modeling. Empir-ical studies showed that (i) MDD4ABMS requires less effort to develop simulations withsimilar (sometimes better) design quality than NetLogo, and (ii) MDD4ABMS meetsqualitative aspects related to the user experience, such as ease of comprehension andusability. These results give evidence of the benefits that MDD4ABMS provide to ABMS.
Keywords: Agent-based Modeling and Simulation. Model-driven Development.Domain-specific Modeling Language. User Study.