ML for telemetry anomaly detection

Not started

Self-directed study

Statistical and ML methods for detecting anomalies in spacecraft telemetry time series.

Practice in the Flight Lab: Spot the anomaly · Anomaly Machine

Subareas

Baselines (Z-score, MAD, limits)Isolation Forest & LOFLSTM forecasting + dynamic thresholdsLabeled fault datasets (SMAP/MSL)Precision/recall & alert fatigue

What to learn

Reproduce telemanom on SMAP/MSL
Compare against MAD baseline
Deploy detector on own telemetry

Resources

Hundman et al. 2018 ↗arxiv.org
telemanom ↗github.com
scikit-learn outlier detection ↗scikit-learn.org