A rule-based agent-oriented approach for supporting weakly-structured scientific workflows

Weakly-structured Scientific Workflows (WsSWFs) often contain goal-oriented tasks that are logical and complicated, but they are vital for workflow results. They may involve interactions between multiple participants or have complicated logic to express scientific policies and cater to dynamic execution environments. In general, such WsSWFs not only need a rich process and (domain-specific) decision logic specification, but also require a flexible execution and human interaction. In this paper, we propose a Rule-based Agent-oriented Framework (RbAF) to support the WsSWF execution by combining rule-based knowledge representation with agent technology. We describe workflows by messaging reaction rules, which go beyond global Event-Condition-Action (ECA) rules and support performing complex actions locally within certain contexts. We describe (domain-specific) decision logic in workflows by exploiting the benefits of both Logic Programming (LP) and Description Logic (DL). Our evaluation results show that, RbAF well supports the WsSWFs and has higher expressive power than other three considered scientific workflow systems.