Quality of information measures for autonomous decision-making

We present a methodology to detect changes in quality of information (QoI) of data received by an autonomous entity. QoI is defined as the inverse of the expected Kullback-Leibler distance between a reference probability distribution and the conditional distribution associated with the data. When the underlying dynamic process that generates the data is real-valued, the interacting multiple model Kalman filter (IMM-KF) can be used to compute QoI. For the case of discrete-event dynamics, we present an IMM Bayes filter to detect changes in QoI. Numerical examples are provided to illustrate the methodology.