Project Description

Discrete-event systems (DES) can be used to model failures in a wide class of systems; for an overview on the use of DES models for fault detection, see [1]. It is typically assumed in DES diagnosis problems that sensors available to an observer are always active; however, in problems where events may be sensed by sending probes over parts of the network, it would be more practical to only sense these events when it is necessary to do so. In our approach, we demonstrate how to actively update the observation actions to find a policy that minimizes the cost of observing events while retaining the ability to diagnose failures. The figure below shows the structure of a diagnoser actively updating its observation policy; details of the method are discussed in [2].