Sensor Trust-Based Fusion Estimation With Event-Triggered Feedback Under Generalized Abnormal Measurements

Multisensor state estimation is challenging owing to generalized abnormal measurements arising from various sources, including interference, sensor faults, or malicious attacks. Existing methods generally depend on prior knowledge of anomaly characteristics or accurate anomaly detection, which limits their applicability when anomalies occur intermittently and affect an unknown subset of sensors. This article proposes a trust-based fusion estimator with event-triggered feedback for unknown measurement anomalies, mitigating their impact while reducing unnecessary transmissions. Sensor trust, defined as the expected reputation, is updated online through consistency evaluation and is used to determine fusion weights, thus suppressing the influence of unreliable estimates on fused estimate. Event-triggered feedback selectively resets disrupted local filters using the fused estimate. Convergence conditions are derived to quantify the tolerable anomaly magnitude and the maximum number of unreliable sensors, and complexity analysis describes implementability. Simulations demonstrate the accuracy and robustness of the proposed method under multiple anomaly settings.