ESA satellite telemetry anomaly detection

Short summary

Researchers developed a hierarchical ensemble pipeline that combines shapelet-based and statistical feature extraction to detect anomalies in European Space Agency satellite telemetry. The approach uses per-channel modeling with intra-channel stacking and cross-channel aggregation to prevent information leakage during validation. Results on the ESA Anomaly Detection Benchmark demonstrate strong generalization, showing that hierarchical modeling effectively identifies subtle anomalies in multivariate time-series data.

• •Ensemble pipeline combines shapelet-based and statistical feature extraction for anomaly detection
• •Uses per-channel modeling with cross-channel aggregation to handle multivariate telemetry
• •Validated on ESA Anomaly Detection Benchmark with strong generalization performance

Generated with AI, which can make mistakes.