Multi-Agent Safety Monitor

On-line safety monitoring, i.e. the real-time detection, diagnosis and correction of hazardous failures, is an essential task in the operational phase of most safety-critical systems. Although current safety monitoring schemes deliver this task to some extent, the problem of effective and timely safety monitoring is still largely unresolved. In this paper, we address the safety monitoring problem through an intelligent distributed safety monitor. This monitor is based on a Multi-agent system (MAS) and uses knowledge about normal and failure operational behaviour that is derived from a systematic safety analysis process. By exploiting this knowledge and real-time observations of the target system, agents deliver a wide range of monitoring tasks which include detection, diagnosis and correction of failures, both locally and globally via MAS collaboration.