On Modelling Multi-agent Systems Declaratively

Abstract

Abstract. We propose a declarative framework for modelling multi-agent systems and specify a number of properties of these systems and agents within them. The framework is parametric with respect to an input/output semantics for agents, whereby inputs are the agents' observations, and outputs are their actions. The observations include actions performed by other agents and events happening in the world. We define the semantics of a multi-agent system via a stability condition over the individual agents' semantics. We instantiate the framework with respect to simple abductive logic agents. We illustrate the framework and the proposed properties by means of a simple example of agent negotiation.

Logic programming with social features

“…Logic-based Multi-Agent Systems -A related approach, where the semantics of a collection of abductive logic agents is given in terms of the stability of their interaction can be found in (Bracciali et al 2004) where the authors define the semantics of a multi-agent system via a definition of stability on the set of all actions per-formed by all agents in the system, possibly arising from their communication and interaction via observation. According to the authors, a set of actions committed by different agents is stable if, assuming that an "oracle" could feed each of the agents with all the actions in the set performed by the other agents, each agent would do exactly what is in the set.…”

Solving abduction by computing joint explanations

“…Close to this approach is the framework of [7], where the semantics of Multi-Agent Systems is given in terms of a stable set of actions for different agents, where a set is stable if assuming an oracle that could feed each of the agents with all the other actions in the outcome, then agents' decision would be what the set envisages. …”