Rubén Fuentes

Model driven development of multi-agent systems

From a software engineering point of view, the agent paradigm has shown its potential for modelling in different domains. However, moving from agent models to implementation is not fully addressed by most agent-oriented methodologies in a systematic way. In most cases, they focus on the agent concept at the analysis level or look for visual or formal representations of elements present in an already implemented agent framework. Here we show that model driven development with agent-based models can facilitate the implementation of methods and tools for the development of multi-agent systems. This is illustrated with the INGENIAS methodology and its tools, the INGENIAS Development Kit (IDK), which are founded on the definition of the INGENIAS meta-models for multi-agent systems, and implement most of the facilities that a model driven development approach requires.

Activity Theory for the Analysis and Design of Multi-agent Systems

Modeling a Multi-Agent System (MAS) involves a large number of entities and relationships. This implies the need for defining an organization, in order to structure and manage the complexity of the whole system. In this work we propose the use of human organization metaphors and their application to the verification of MAS specification. In concrete, we apply Activity Theory, which has its roots in Sociology, to study agent systems and obtain relationship patterns that can be applied to the analysis of MAS. These patterns guide analysis and design refinements, and help to detect inconsistencies. This technique has been implemented and integrated in the INGENIAS IDE platform, and proved with some case studies, in particular, for agent-based web applications.

Integrating agent-oriented methodologies with UML-AT

There are many methodological approaches for Agent-Oriented Software Engineering, each one focusing on some features of multi-agent systems, but leaving others underdefined. For this reason, it would be interesting to have the possibility of applying different methods, according to their suitability to each particular problem domain and system view. Here, a key issue is how to integrate the information resulting from different methods in a common specification. We propose the use of an intermediate language called UML-AT, which would enable bidirectional transformations between models in different languages. These transformations allow representing views of those models in the language of choice at every moment. UML-AT is a UML profile based on the Activity Theory (AT) framework, which includes the concepts to describe societies of actors that are both autonomous and intentional. The translation with UML-AT mappings makes the integration process independent of any given methodology. At the same time, these mappings can be a basis to study missing features in design languages, according to what is needed in the study of the intentional and social aspects in multi-agent systems and human organizations following AT. This integration architecture by means of UML-AT mappings is illustrated with a case study about a legacy tourism application. It uses a support tool that automates the translation.

Social Analysis of Multi-agent Systems with Activity Theory ∗

The development of software systems is a complex task that requires support techniques to guide the process and solve inconsistencies in its products. In the agent paradigm, the use of social and intentional abstractions facilitates the application of these techniques in theories for the analysis of human societies, such as the Activity Theory. The Activity Theory proposes patterns that appear in human groups to explain their behaviour, and the study of their internal contradictions to understand its evolution. These patterns can be translated to the agent paradigm to reason about the behaviour of multi-agent systems. To ease the use of such techniques, a partially automatized process is proposed. The aim is to create an assistant that helps developers in the specification of their multi-agent systems. The use of this tool is evaluated through case studies specified with the INGENIAS notation.

Requirements elicitation for agent-based applications

Requirements Elicitation for a software system is a key stage in a successful development. At the same time, it is one of the most challenging, because requirements have to consider the mutual influences between the envisioned system and the human context where it will work. These influences cover aspects such as organization, motivation, evolution, and cognition, taking place in a specific setting. The agent paradigm facilitates the analysis of these features because of its intentional and social nature. Nevertheless, determining the information that should be obtained and the way it should be modelled is not a trivial task. Developers are experts in software systems but they are not always familiarized with the concrete domain of those systems. The Requirements Elicitation Guide, a technique based on the Activity Theory from Social Sciences, can be applied to support developers in these issues. This guide empowers the development team with the experience of Social Sciences in these issues. This paper introduces the guide and shows its application in a case study about a web application.

A Sociological Framework for Multi-agent Systems Validation and Verification

Social and intentional behaviours appear as two main components of the agent paradigm. Methods of conventional software engineering do not seem to be appropriate to gain a full knowledge of these behavioural aspects, as they are not traditional software components. Their study involves new concepts and techniques, belonging to social sciences, to be integrated into software development. In this paper we show how to use the Activity Theory as a support for current Multi-Agent System methodologies. The application of the same social concepts at every stage of the development cycle allows requirements traceability, and provides methods to describe social properties and decide whether the specification satisfies them or not. The use of the approach is shown with a case study.

Checking Social Properties of Multi-agent Systems with Activity Theory

Many approaches of the agent paradigm emphasize the social and intentional features of their systems, what are called social properties. The study of these aspects demands their own new techniques. Traditional Software Engineering approaches cannot manage with all the information about these components, which are as related with software development as with social disciplines. Following previous work, this paper presents a framework based in the Activity Theory to specify and verify social properties in a development process for multi-agent systems. Using this framework developers acquire tools for requirements elicitation and traceability, to detect inconsistencies in their specifications, and to get new insights into their systems. The way of working with these tools is shown with a case study.

Agent behavior representation in INGENIAS

Nowadays, we have different agent oriented methodologies that enable developers to produce agent oriented designs. One of the recurrent problems of these methodologies is how to describe the behaviour of agents within a system. A developer needs primitives to express autonomy, proactivity, and social concerns of his agents, but there are problems in understanding what does these elements mean, beyond any natural language explanation. There is a clear need of semantic models understandable by average engineers. These models could help in foreseeing the impact of autonomy with respect system goals, or determining if, in an agent specification, a task will ever be scheduled for execution. This paper presents a proposal of semantic model for the visual modelling language used in INGENIAS, a project started in 2002 and considered the inheritor of MESSAGE/UML.

Agent oriented software engineering with web applications

This paper proposes an agent oriented approach to design web applications. Contrary to other research work, this paper considers agents only at the conceptual level because they are convenient abstractions. In the implementation, agents are substituted by more adequate computational entities like Enterprise Java Beans or servlets. The translation from the conceptual level to the implementation is based on the relevance of interactions in both web-applications and Multi-Agent Systems. Throughout the paper, we show the current situation in Multi-Agent systems and what elements are required in a Multi-Agent System implementation.