Flexible and Emergent Workflows using Adaptive Agents

Abstract

Most of existing workflow systems are rigid since they require to completely specify processes before their enactment and they also lack flexibility during their execution. This work proposes to view a workflow as a set of cooperative and adaptive agents interleaving its design and its execution leading to an emergent workflow. We use the theory of Adaptive Multi-Agent Systems (AMAS) to provide agents with adaptive capabilities and the whole multi-agent system with emergent "feature". We provide a meta-model linking workflow and AMAS concepts, and the specification of agent behavior and the resulting collaborations. A simulator has been implemented with the Make Agent Yourself platform.