Separating Agent-Functioning and Inter-Agent Coordination by Activated Modules: The DECOMAS Architecture

TL;DR

This paper introduces the DECOMAS architecture, a modular programming model for multi-agent systems that separates coordination activities from agent functions, enabling systematic development and reuse of decentralized coordination processes.

Contribution

It presents a novel architecture and formalization for activating modules within agents to encapsulate and manage coordination activities independently.

Findings

Formalization of module activation within MAS

Application to decentralized web service management

Enhanced separation of coordination and agent functions

Abstract

The embedding of self-organizing inter-agent processes in distributed software applications enables the decentralized coordination system elements, solely based on concerted, localized interactions. The separation and encapsulation of the activities that are conceptually related to the coordination, is a crucial concern for systematic development practices in order to prepare the reuse and systematic integration of coordination processes in software systems. Here, we discuss a programming model that is based on the externalization of processes prescriptions and their embedding in Multi-Agent Systems (MAS). One fundamental design concern for a corresponding execution middleware is the minimal-invasive augmentation of the activities that affect coordination. This design challenge is approached by the activation of agent modules. Modules are converted to software elements that reason about and modify their host agent. We discuss and formalize this extension within the context of a generic coordination architecture and exemplify the proposed programming model with the decentralized management of (web) service infrastructures.