A Distributed Architecture for Norm Management in Multi-Agent Systems

Norms, that is, obligations, prohibitions and permissions, are useful abstractions to facilitate coordination in open, heterogeneous multi-agent systems. We observe a lack of distributed architectures and non-centralised computational models for norms. We propose a model, viz., normative structures, to regulate the behaviour of autonomous agents taking part in simultaneous and possibly related activities within a multiagent system. This artifact allows the propagation of normative positions (that is, the obligations, prohibitions and permissions associated to individual agents) as a consequence of agents’ actions. Within a normative structure, conflicts may arise – one same action can be simultaneousely forbidden and obliged/permitted. This is due to the concurrent and dynamic nature of agents’ interactions in a multi-agent system. However, ensuring conflict freedom of normative structures at design time is computationally intractable, and thus real-time conflict resolution is required: our architecture support the distributed management of normative positions, including conflict detection and resolution.