AN ONTOLOGY FOR COORDINATION

Independent agents that interact within open, distributed, and decentralized environments need to collaboratively regulate their activities in order to facilitate the harmonious and successful achievement of possibly conflicting tasks. Coordination is the process of managing these interactions by identifying and possibly resolving the interdependencies occurring between such activities. A successful coordination mechanism facilitates mutually beneficial interdependencies (e.g., by ensuring activities are not duplicated) while avoiding adverse outcomes (e.g., by preventing two processes simultaneously accessing the same non-shareable resource, potentially causing deadlock). However, for such a mechanism to work effectively within open systems, agents need to communicate and reason about activities, resources, and their properties; i.e., to commit to a shared ontology of coordination that defines the semantics underlying the different coordination regimes. This article describes an ontological approach to coordination in which agents dynamically manage the interdependencies that arise during their interactions. A proof-of-concept implementation in the insurance domain is described and empirically evaluated.