A dynamic decentralized sequential multi-hypothesis testing problem under uniformly distributed nonstationary observations

Abstract

A dynamic team of multiple communicating decision makers perform Bayesian sequential multi-hypothesis testing with nonstationary observations. A member-by-member optimal decision rule is derived under the conditions that the observations are independent and uniformly distributed conditioned on each hypothesis, that the observation costs are adapted to the quality of observations, and that the cost for decision errors is sufficiently large and does not impose additional penalty for "vague" terminal decisions. The "vague" terminal decisions can be made "clear" by performing a finite number of member-by-member optimal sequential decisions successively.