Far-Sighted Diagnosis of Active Systems
Abstract
Active systems are a class of discrete-event systems modeled as networks of non-deterministic automata communicating through either synchronous or asynchronous connection links. The model-based diagnosis of an active system is carried out by first reconstructing its behavior based on the observation, from which faults are later derived. The complexity of behavior reconstruction is exacerbated by the possibility of queuing events within links, thereby making essential the simulation of the order in which events are buffered within links. Unfortunately some sequences of events may lead to blind alleys in the search space. This is especially critical if events exchanged among components are assumed to be uncertain, as the number of alternative sequences of queued events is still larger. Therefore, behavior reconstruction without any prospection in the search space is generally bound to detrimental backtracking. To make diagnosis of active systems more efficient, we present an off-line technique for processing the models inherent to the system at hand so as to automatically generate prospection knowledge relevant to the mode in which events are produced and consumed over links. Such a knowledge is then exploited on-line, when the diagnostic engine is running, to guide the search process, thus reducing both time and space.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 04, 2002
- Accession Number
- ADP012705
Entities
People
- Gianfranco Lamperti
- Marina Zanella
- Roberto Garatti