Agent-supported simulation for coherence-driven workflow discovery and evaluation

This article proposes a generic agent-supported symbiotic simulation architecture that generates and evaluates competing coherence-driven workflows using genetic programming. Workflows are examined by an agent-supported multi-simulation environment that allows inducing variation and assessing the outcome of the candidate workflows under a variety of environmental scenarios. The evaluation strategy builds on a coherence-driven selection mechanism that views assessment as a constraint satisfaction problem. The proposed system is also based on an introspective architecture that facilitates monitoring the activities of competing agent simulations as well as the status of the environment to determine the success of the candidate workflow with respect to given or emergent goals.