A Rule-Based Agent Framework for Weakly-Structured Scientific Workflows

Instead of focusing on the orchestration of low-level computational intensive tasks in efficiency critical structured scientific processes, as most of available scientific workflow systems do, this work explicitly addresses the coordination and decision-centric higher levels of weakly-structured scientific workflows supporting the choreography of human agents and IT services/agents in their scientific problem solving tasks. From a technical perspective, this work proposes a rule-based agent framework, which exploits the benefits of both the declarative knowledge representation with rules and ontologies and the abstraction with multi-agent technology to support the dynamic and adaptive execution of weakly-structured scientific workflows. Based on the evaluation in terms of the workflow patterns, the rule-based workflow specification in this work demonstrates a high expressiveness level required to represent different types of tasks. The concrete use cases from different scientific domains implemented by the framework also demonstrate that the work fulfills the requirements of weakly-structured scientific workflows.