Adaptive Sampling for Spatiotemporal Anomaly Monitoring in Wireless Sensor Networks
arXiv:2607.15235
The paper introduces a sentinel‑assisted adaptive sampling framework for wireless sensor networks that combines Kalman‑filter‑driven sparse sensing with sentinel nodes performing GLR‑based anomaly detection and local wake‑up alerts to improve detection of short‑lived spatiotemporal anomalies while reducing sampling cost.
Abstract
Long-term environmental monitoring in wireless sensor networks (WSNs) often uses sparse sampling to extend network lifetime, but sparse sensing can miss short-lived, localized, and potentially diffusive anomalies. This paper proposes a sentinel-assisted adaptive sampling framework as a cooperative sensing-control pipeline for WSN anomaly monitoring. During normal periods, nodes perform sparse sensing driven by Kalman filter (KF) predictive uncertainty. During anomalous periods, continuously sampled sentinel nodes perform hybrid GLR-based detection with node-relative thresholds, and local detections trigger one-hop neighborhood wake-up with recovery-aware alert control. Experiments on the Intel Berkeley Research Lab temperature dataset with abrupt random spatiotemporal anomalies show that the proposed method raises the anomaly-window sampling ratio (AWSR) from 0.439 to 0.933 in the main experiment. It also improves AWSR over Adaptive Data Acquisition with Energy Efficiency and Critical-Sensing Guarantee (AAS) and Adapted e-Sampling while reducing total cost by 15.4\% and 2.1\%, respectively. These results show that integrating KF-based sparse sampling, sentinel GLR detection, and local alert propagation improves anomaly-window visibility while maintaining a lower sampling-cost trade-off.
Accepted paper for IEEE ISSC 2026 conference, Limerick, Ireland