ADISTES

Active diagnosis aims at significantly improving system reliability by using diagnostic information at run-time for fault isolation and online error recovery. Active diagnosis for open embedded real-time systems (e.g., health management and medical systems) is an open research problem due to stringent real-time and reliability requirements in combination with constituent components that are unknown at design time.

The proposed project will extend semantic techniques, usually used in large-scale IT systems, for active diagnosis in open embedded real-time systems. We will develop modeling techniques for expressing diagnostic features, symptoms, faults and recovery actions. Methods for distributed knowledge management will establish relaxed consistency while ensuring real-time constraints. Real-time inference will be investigated based on the time-triggered scheduling of diagnostic queries. The goal of query transformations, semantic transformations and goal-oriented learning will be improved schedulability and reliability. The methods and algorithms will be prototypically implemented, as well as experimentally and analytically evaluated concerning reliability and timeliness.

Major contributions beyond the state-of-the-art include

• modeling techniques for a diagnostic knowledge base,
• time-triggered scheduling and optimizations of diagnostic queries for real-time inference
• distributed knowledge base management with relaxed consistency, and
• goal-oriented self-learning for active diagnosis in open embedded systems.