Rethinking MAS Infrastructure based on Activity Theory

   page       BibTeX_logo.png   
Andrea Omicini, Alessandro Ricci, Sascha Ossowski
Mark d'Inverno, Carles Sierra, Franco Zambonelli (eds.)
1st European Workshop on Multi-Agent Systems (EUMAS 2003)
December 2003

The engineering of complex systems upon Multi-Agent System (MAS) paradigm cannot be conceived without a suitable infrastructure support, in particular for what concerns organisation and coordination issues. Currently, there are no adequate infrastructure models, but only ad hoc solutions; in particular, if on the one hand the solutions at the infrastructure level provided by mainstream approach such as FIPA seem to be in somewhat sufficient for enabling the agent interaction space (communication, interoperability), on the other hand these solutions are not adequate for governing it (organisation, coordination, security). The same problem can be devised also in other contexts, analogous in terms of the complexity of organisation and coordination activities we aim to achieve in MAS, such as humas society. Here, some models and theories have been developed shaping the features of infrastructure aiming at supporting such complexity. A notably example is Activity Theory, whose conceptual framework has been used and is used to analyse and frame collaborative activities and the infrastructure means useful to support them. Accordingly, in this work we describe how Activity Theory (AT) can be a good conceptual framework also for MAS infrastructure. In particular, the notion of coordination artifact and agent coordination context will be discussed, as governing abstractions provided by a MAS infrastructure in order to support adequately the modelling and engineering of agent society interaction.
The work is articulated as follows: in Section 1 we recall the notion of infrastructure, and discuss the importance of governing abstractions. In Section 2 we introduce the AT as a suitable conceptual framework for modelling/engineering MAS activities and their infrastructure support. The notions of coordination artifact and agent coordination context are then discussed, as infrastructure abstractions useful for the explicit representation and enactment at runtime of coordination and organisation aspects. Finally, in Section 3 we provide the conclusion.

origin event