Article Preview
TopIntroduction
A Multi-Agent System (MAS) is composed of entities called agents that interact together in order to achieve a common goal. An agent, in general terms, is anything that can be viewed as perceiving its environment through sensors, and acting upon that environment through effectors (Russell & Norvig, 1995). Computational MASs consist entirely of artificial agents executing as software programs and running on a computer system. Some researchers have treated the concept of formal descriptions of MAS, while the aim is to assess their properties and formal specifications. In fact, there are formal descriptions of both agent paradigms and interaction paradigms; however, some formalisms provide representations of both of them. In the literature, as in Wooldridge and Jennings (2005), formal descriptions are also called theories. The MAS theories are rules that constitute a basis for specification (Glaser, 2002; Occello, 2002\), design (D’Inverno, Kinny, Luck & Woolridge (1998), Ricordel & Demazeau (2002)) implementation and verification of MAS (Arlabrosse, Gleizes & Occello (2004), D’inverno, Fisher, Lomuscio, Luck, Rijke, Ryan & Woolridge (1997), Picard & Gleizes (2004)). There are other methodologies and theories dealing with this concept, such as The Multi-Agent Scenario-Based Method (Moulin & Brassard (1996), Moulin & Cloutier (1994)) that is intended to be applied in the field of cooperative work.
Moreover, software agents represent a fundamental way of considering complex distributed systems and societies of cooperating autonomous components. When building such systems, special techniques are required for their design and implementation. In our work, we aim at assisting the development of such systems by providing a model and formalism that can be used to specify the desirable behavior of MAS, where one requirement is to be able to move from specifications of such systems to implementation. The characteristics identified by using a formalism serve to measure and evaluate implementations of agent systems. On the other hand, CSCW is the field of study that examines how technology affects group interaction, and how technology can be best designed and built to facilitate group work. Activities in that domain are known as groupware.
Ellis, Gibbs & Rein (1991) defines groupware as “computer-based systems that support groups of people engaged in a common task (or goal) and that provide an interface to a shared environment”. While groupware refers to real computer-based systems, this means that the notion CSCW is the study of tools and techniques of groupware as well as their psychological, social and organizational effects. Some questions arise such as how to conceive and develop adequate collaborative work architectures taking into account communication, coordination and production features, and how to evaluate the developed collaborative systems. In this paper, we attempt to answer these questions by putting forward a new conceptual model of an agent called Collaborator Agent (CA) used to develop and evaluate software architectures of collaborative applications. The proposed CA is based on an approach combining astutely two research area: MAS and CSCW (Ellis, Gibbs & Rein (1991), Laurillau & Laurence (2002)). In the first area, a formal model of the MAS (Ferber approach in Ferber & Muller (1996)) is used to bring out the features of the agent. The second area provides a useful specification of collaboration based on the communication, coordination and production theories. Then, we apply our conceptual model on the ARITI-C system for collaborative online robot teleoperation. This system offers web interfaces in order to assist users in collaboration to remotely manipulate real objects via a robot.