Adaptive Kalman Filter Based Data Aggregation in Fault-Resilient Underwater Sensor Networks

Sensor Networks in harsh underwater environments are prone to faults and anomalies that may lead to deteriorated data quality or even failures. This paper proposes a fault-resilient underwater sensor network based on sensor data aggregation by updating the measurement error matrix of an adaptive Kalman filter, where the error matrix is updated using adjusted measured value difference from the predicted value as well as the age of the latest measurement (i.e. latency). A case study on a real-world harbor water flow monitoring use-case shows the advantages of the proposed method. The experiments indicate that the adaptive and difference based Kalman filter aggregation provides for a significantly smoother aggregation in case of high fault rates in sensors’ readings when compared to traditional Kalman filter and median value based aggregation techniques.