Diagnosing plan failure in multi-agent systems using plan abstract knowledge

Open multi-agent systems (MAS) are decentralised and distributed systems that consist of large number of loosely coupled autonomous agents. Data and control in such systems is highly distributed. In the absence of centralised control they tend to be difficult to manage, especially in an open environment, which is a dynamic, complex, distributed and unpredictable. This dynamism and uncertainty of an open environment gives rise to unexpected plan failures. In this paper we focus on the process of diagnosing causes of exceptions in failed plan. We apply our proposed sentinel agent based architecture in order to diagnose a plan's actions failure interactively. Once an exception is detected the sentinel agents involve in a dialogue with associated agents using Agent Communication Language (ACL) messages in order to obtain relevant information. This information consists of abstract knowledge about the failed plan and knowledge about agents' mental attitudes regarding the failed plan. The sentinel agent then uses this abstract knowledge and agent's mental attitudes to diagnose the underlying cause of the plan failure. Agents voluntarily inform the sentinel about their mental state regarding the failed plan.