A Bayesian Approach to Sequential Change Detection and Isolation Problems

The problem of sequential change detection and isolation under the Bayesian setting is investigated, where the change point is a random variable with a known distribution. A recursive algorithm is proposed, which utilizes the prior distribution of the change point. We show that the proposed decision procedure is guaranteed to control the false alarm probability and the false isolation probability separately under certain regularity conditions, and it is asymptotically optimal with respect to a Bayesian criterion.