Uma Proposta de Incorporação do Diagrama de Classes da UML à Linguagem MASRML Adaptado ao Contexto de Sistema Multiagentes

Abstract

In Software Engineering, the use of modeling languages, techniques, methods, and development processes aids the designer to develop software with higher quality. Among the modeling languages used in software development, one of the best known is the UML – Unified Modeling Language – and its class diagram is one of the most used, as it is capable of demonstrating the logical structure of object-oriented software by means of the representation of their classes, attributes, operations, and relations. Now in the Distributed Artificial Intelligence area, software agents – computer programs characterized by having autonomy, social ability, reactivity, and proactivity – possess a unique structure and requirements, which cannot be represented through the standard UML class diagram. Thus, it is necessary to adapt this diagram to the Multiagent Systems context. Keeping this in mind, we realized the need to verify the existence of UML class diagram extensions for the context of multi-agent systems and evaluate their real usefulness and support for the development of this type of software. Thus, we carried out a systematic literature review, aiming to identify the strengths and weaknesses of these extensions. This way, we analysed the extensions found in the review and, after applying the diagrams presented, we realized the need to propose a new extension and adaptation of the UML class diagram, adding it to the MASRML (Multi-Agent Systems Requirements Modeling Language) – a language derived from UML exclusive to the representation of particular requirements of multiagent systems. The extension was carried out with the objective of allowing the MASRML language to support the structural characteristics of the class diagram, duly adapted to the multiagent systems context, in such a way that this language can represent concepts that we consider important based on the review conducted. After producing an initial version, we evaluated our proposal through an experiment based on comparing models. Key-words: UML. Class Diagram, Multiagent System. Modeling Languages.