Information Structures for State-Dependent Decentralized Detection

The problem of decentralized detection over a sensor network where each agent takes a state a priori is considered. It is assumed that the agents’ states have a impact the underlying hypothesis, resulting in correlated observations. The impact of fusion center knowledge is examined: with network state knowledge and without. In the limit of network size, the error exponent for both cases is computed and the relationship between both cases is characterized. A novel error exponent representation facilitates the asymptotic analysis.