An overview of current trends in European AOSE research

Informatica, Nov, 2005 by Carole Bernon, Massimo Cossentino, Juan Pavon

The agent oriented approach is doing great steps towards its (not yet reached) maturity; from a software engineering point of view, it is today positively used for the analysis and design of complex systems. In this" paper, which is related to the activity of the AgentLink AOSE TFG (Agent Oriented Software Engineering Technical Forum Group), we provide a perspective of current research trends in this area with a specific attention for results' coming from European groups. We start with a discussion of what are agents', specially from the perspective of the software engineer. We present recent trends in modelling agents and multi-agent systems, and then we review the different activities' in the agent development process: J?om analysis and design to implementation, verification and finally testing.

Keywords: Agent Oriented Software Engineering (AOSE), Agent oriented methodologies, Multi-Agent Systems

Received: June 31, 2005

Povzetek: Podan je povzetek evropskega raziskovanja AOSE.

1 Introduction

With the increasing amount of successful applications and techniques based on the agent paradigm, which have validated the feasibility of the approach, there is a big concern on its applicability in an industrial context. This implies the definition of repeatable, reusable, measurable and robust software process and techniques for the development of multi-agent systems (MAS). For this reason, a lot of effort in the agent field has been devoted to the definition of methods and tools for supporting agent oriented software engineering (AOSE). This involves the definition of modelling languages for the specification of MAS, techniques for requirements elicitation and analysis, architectures and methods for designing agents and their organizations, platforms for implementation and deployment of MAS, and validation and verification methods. Taking into account the diversity of influences in the agent paradigm (from distributed objects to knowledge base systems, but also from other fields such as Psychology, Biology and Social sciences) there are many methodological approaches, which should get unified and integrated in a common body of knowledge and practices. This is one of the aims of current actions at EU level, such as the AgentLink (www.agentlink.org) effort, or the collaboration in standardization organizations such as FIPA (www.fipa.org).

In this paper we try to provide a perspective of current research trends in this area, specially in EU groups. This can be useful as a starting reference point to look for specific matters (in this sense there is an extensive bibliography), and is complemented in relevant topics with other papers in this special issue.

The paper starts with a discussion of what are agents, specially from the perspective of the software engineer (section 2). This is followed by a presentation of trends in modelling this kind of systems (section 3). Then, different activities in the development process for MAS are reviewed: analysis and design (section 4), implementation (section 5), verification and testing (section 6). Finally, the conclusions (section 7) provide a view, from the authors of this paper, on what lines of work and trends should follow research in this area.

2 From Objects to Agents and Multi-Agent Systems

When dealing with the agent notion and how to engineer agent-based applications, one question often arises: may agents be considered as an extension of objects and then classical object-oriented software engineering be used as well to build agent-based applications? Several papers have tried to answer this question [76][106], others have compared agents with programs [46] or with components [7]. Many authors agree on the fact that distinguishing agents and objects is difficult because they share some aspects, but they also differ, mainly on notions such as autonomy and interaction. Both agents and objects encapsulate their state, which in objects is determined by the values of a set of variables whilst in agents this can be defined in terms of goals, beliefs, facts, etc., what determines a mental state. Objects may have control over their state by using private attributes or methods but any public method of an object can be invoked by another object forcing the former one to perform the action described by the method. An object, contrary to an agent, has then a limited control over its behaviour because the decision on which method to execute is taken by an external actor (the caller). An agent can determine which behaviour to follow (depending on its goals, its internal state and its knowledge from the environment) and not because someone else forces it to do something.

Therefore, the notion of autonomy is stronger in agents. This autonomy in agents implies that usually they have their own thread of control, whilst, most of the time, objects are passive entities, becoming active just when one of their methods is invoked by another object. This difference may be alleviated by the notion of active objects in which an object has its own thread of control. However, agents have some features which make them something more than active objects. According to Van Parunak and Odell [76], agents exhibit a dynamic autonomy (their behaviour can be reactive as they react to changes in their environment, proactive as they are able to take initiatives to proceed into goal-directed actions, and social as they communicate with other agents in organizations) as well as an unpredictable autonomy (their behaviour depends on their state, their individual goals, and their interactions with others). Active objects would become agents if they are able to take "initiatives". However, this distinction is not always well established. For this reason some works in the agent domain, for instance, on formalization of coordination issues, usually are more related to classical concurrency theory and do not consider intentional aspects of agents.