Showing papers by "Ravi R. Mazumdar published in 1986"

Center for Telecommunications Research

Abstract: The problem of minimal time detection of abrupt parameter changes in linear stochastic systems considered. The problem is posed as an optimal stopping problem for the detection in change of the induced probability measure. Under the assumptionof a prior distribution for the time of change (or disruption) a stopping rule is given which minimizes the average detection delay when there is knowledge of the new measure after the change. When the new induced measure is unknown, a stopping rule is given, based only on the noisy observations and is shown to be better than the a priori knowledge of the disruption time.