Exploiting System Structure in Model-Based Diagnosis of Discrete-Event Systems

We describe an implemented system for model-based diagnosis of discrete-event systems, which are continuous systems governed by discrete controllers. Our approach contributes to model-based diagnosis by taking advantage of the system structure (a directed graph depicting component interconnectivity) to eeciently diagnose dynamic systems, and by providing computational guarantees based on such structure. Specifically , we use a propositional temporal logic to describe discrete-event systems, but constrain the temporal sentences deening a system using the topology of its structure. We describe a computational machinery for computing and focusing consistency-based diagnoses based on a structured set of temporal sentences, the complexity of which is dependent on the topology of the given structure. We explain how to engineer the system structure to ensure that our algorithms are eecient. We also explain why our approach has proven eeective for diagnosing discrete-event systems by identifying characteristics of this domain that make it amenable to structure-based methods. Motor AM+ AM-Actuators P1 P2 P3 Figure 1: A discrete-event model for a conveyor system, with the following components: three sensors xSensor; ySensor and zSensor to detect presence of objects at the corresponding positions; a motor capable of forward and reverse motion; two actuators AM + and AM ? for driving the motor in forward and reverse directions; a controller (not shown here) that maps sensor readings into actuator settings.