OPEN INTERACTION FRAMEWORKS

This project is carried out in cooperation between HES-SO and University of Lugano in the framework of the COST Agreement Technologies. Its goal is to prove that open interaction systems, conceived as systems of interacting artificial institutions, can be used to support human collaboration. To reach this goal we extend the existing meta-model OCeAN of artificial institutions, and use this extension to specify and design a demonstrative system as a negotiation platform interconnecting hospitals.

The possibility to specify open interaction systems, which can be dynamically entered and left by heterogeneous, autonomous, and self-interested agents, is widely recognized to be a crucial requirement for the development of intrinsically distributed applications on the Internet. In its past research, Prof. Colombetti's team in University of Lugano has developed OCeAN, a normative meta-model for the specification of artificial institutions, and a complete Agent Communication Language (ACL) that can be used to model open interaction systems where heterogeneous software and human agents may interact. This approach on normative multi-agent systems (MAS) is focusing on the social properties of the system in order to predict the evolution of open MASs in time.

The current problem with artificial institutions approaches is that while they offer an interaction space and a normative framework for the agents, there is neither a clear model nor an appropriate infrastructure to specify how the mediation of agents’ interaction is being guaranteed.

For that, we propose the usage of a governing environment to structure and shape the space of interaction within open MAS. This governing perspective mainly allows managing agent interactions from an external point-of-view, and has the following advantages: i) it provides a software layer for the mediation of agent interactions; ii) it provides an infrastructure that offers all the necessary services; and iii) it offers an interface for the agents to use the low level resources of the system. Furthermore, it provides a middleware to run the agents, providing them services for their existence.

In summary, we want to further investigate and specify this governing environment applied to open interaction systems by expanding the OCeAN meta model with a focus on mixing human and artificial agents in hybrid multi agent systems. In particular, we plan to develop a demonstrative system able to support the activities of an e-health marketplace.

Publications concerning this project are available on http://publications.hevs.ch/index.php/topics/single/95