Distributed Chernoff Test: Optimal Decision Systems Over Networks

We study “active” decision making over sensor networks where the sensors' sequential probing actions are actively chosen by continuously learning from past observations. consider two network settings: with and without central coordination. In first case, nodes interact each other through a entity, which plays role of fusion center. second in fully distributed fashion. both these scenarios, we propose adaptive hypothesis tests extending classic Chernoff test. compare performance proposed to optimal presence center, our test achieves same asymptotic optimality test, minimizing risk, expressed expected cost required reach plus wrong decision, when observation per unit time tends zero. The is also asymptotically higher moments decision. Additionally, parsimonious terms communications, number channel uses node small constant. setup, Chernoff's up multiplicative constant risk time. communications comparison state-of-the-art schemes literature. analysis extended account for message quantization communication channels random erasures.