Decentralized Sequential Change Detection

The problem of decentralized sequential change detection is considered, where an abrupt change occurs in an area that is being monitored by a number of sensors. The goal is to detect this change as soon as possible at a central location (fusion center) which receives information from all sensors subject to quantization and rate constraints. A novel decentralized sequential detection rule is proposed that requires communication from the sensors at random times and transmission of only low-bit messages. The second-order asymptotic optimality of the proposed scheme is established under different statistical models for the sensor observations. Specifically, when each sensor process either has continuous paths and is continuously observed or it is a random walk, it is proved that the inflicted performance loss (with respect to the optimal detection rule that uses the complete sensor observations) is bounded asymptotically as the rate of false alarms goes to 0. The proposed scheme remains asymptotically optimal, but of first-order, even if it induces an asymptotically low communication rate and there is an asymptotically large number of sensors. Finally, simulation experiments illustrate its efficiency in practice and its superiority over alternative decentralized detection rules that rely on communication at deterministic times.