Catálogo de publicaciones - libros

Testing and Debugging multi-agent systems (MAS) – which are inherently concurrent and distributed – is a challenging task. While complex application scenarios demand intelligent entities with autonomous reasoning capabilities, the applied reasoning mechanisms impair current approaches to validate MAS implementations. Reactive planning systems, namely the well-known (BDI) architecture, have been successfully applied to implement these intelligent entities by means of goal directed agents. Despite testing and debugging, used to validate whether implementations behave as intended, are crucial to serious development efforts, only minor attention has been payed to corresponding tool support and testing procedures for BDI–based MAS. In this paper, we examine how the reasoning mechanism inside agent implementations can be checked and how static analysis of agent declarations can be used to visualize and check the overall communication structure in closed MAS. We present corresponding tool support, which relies on the definition of crosscutting concerns in BDI agents and enables both approaches to the Jadex Agent Platform.

This paper describes an architecture allowing to verify properties of a multiagent system during its execution. This architecture is the basis of our study whose goal is to check at runtime, if agents and more generally multiagent systems satisfy requirements. Considering that a correct system is a system verifying the properties specified by the designer, we are interested in the “property” notion. That is why we give here a definition of “property” and we present an architecture to validate them. The architecture, a multiagent system itself, is based on a set of agents whose goals are to check at runtime the whole system’s properties. So after a brief description of the “property” notion, we describe our architecture and the way to check systems.

This work analyses the problematic of debugging a multi-agent system. It starts from the fact that MAS are a particular type of distributed systems in which the active entities are autonomous in the sense that behavior and knowledge of the whole system is distributed among agents. It situates the problem by firstly studying the classical approaches for conventional code debugging and also the techniques used in distributed systems in general. From this initial perspective, it tries to situate agent and multi-agent systems debugging. It finally proposes the use of conventional data mining tasks like clustering to, by summarising, help in debugging huge MAS.

The ability to effectively debug agent-oriented applications is vital if agent technologies are to become adopted as a viable alternative for complex systems development. Recent advances in the area have focussed on the provision of support for debugging agent interaction where tools have been provided that allow developers to analyse and debug the messages that are passed between agents.

One potential approach for constructing agent-oriented applications is through the use of agent programming languages. Such languages employ mental notions such as beliefs, goals, commitments, and intentions to facilitate the construction of agent programs that specify the high-level behaviour of the agent. This paper describes how debugging has been supported for one such language, namely the Agent Factory Agent Programming Language (AFAPL).