This paper studies how to keep remote estimators informed about the state of a Markov source while minimizing two critical metrics: the Age of Missed Alarms (AoMA) and the Age of False Alarms (AoFA). These ages quantify how long an estimator either fails to detect a true change (missed alarm) or incorrectly believes a change has occurred (false alarm). Such metrics are essential in monitoring and control systems, where outdated or incorrect information can lead to significant performance degradation. The setup involves a source generating states governed by a Markov chain, a sensor that observes this source, and a remote estimator that receives selective status updates over a communication channel. Because sending every update is costly or constrained, the sensor must decide when to transmit. These transmission decisions play a central role in shaping both AoMA and AoFA.
The authors analyze different threshold-based transmission policies, where the sensor sends an update only when the difference between its local state estimate and the remote estimator’s belief exceeds a predefined threshold. Under this structure, the paper derives closed-form expressions for the expected ages of missed and false alarms, revealing how threshold selection impacts system performance. A key insight is that reducing missed-alarm age and reducing false-alarm age often conflict with each other. Tight thresholds lead to more frequent updates, reducing missed alarms but causing more false alarms. Loose thresholds reduce communication but make the estimator slower to detect real changes. The paper characterizes this trade-off and provides optimization guidelines for choosing thresholds that balance both objectives depending on system requirements. Through analytical results and validation experiments, the work offers a systematic framework for understanding how communication constraints and decision policies affect alarm freshness in remote monitoring of Markov processes. The findings are useful for applications such as industrial automation, sensor networks, and anomaly detection systems, where timely and accurate updates are essential.
Minimizing the Age of Missed and False Alarms in RemoteEstimation of Markov Sources