Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding

By Kyle Hundman, Valentino Constantinou, Christopher Laporte, Ian Colwell, Tom Söderström. Knowledge Discovery and Data Mining, 2018

1,881 citations recorded by the source corpus.

As spacecraft send back increasing amounts of telemetry data, improved anomaly detection systems are needed to lessen the monitoring burden placed on operations engineers and reduce operational risk. Current spacecraft monitoring systems only target a subset of anomaly types and often require costly expert knowledge to develop and maintain due to challenges involving scale and complexity. We demonstrate the effectiveness of Long Short-Term Memory (LSTMs) networks, a type of Recurrent Neural Network (RNN), in overcoming these issues using expert-labeled telemetry anomaly data from the Soil Moisture Active Passive (SMAP) satellite and the Mars Science Laboratory (MSL) rover, Curiosity. We also propose a complementary unsupervised and nonparametric anomaly thresholding approach developed during a pilot implementation of an anomaly detection system for SMAP, and offer false positive mitigation strategies along with other key improvements and lessons learned during development.

Subfields: Artificial Intelligence, Signal Processing

View Paper: https://www.semanticscholar.org/paper/67628543f7ae51979acadfbb8860568b25f263da

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Signal Processing

Citations in the corpus

Multivariate Time-series Anomaly Detection via Graph Attention Network (2020, methodology): “al [2] leverages LSTM to detect anomalies in multivariate time-series metrics of spacecraft based on prediction errors.”
An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series (2021, methodology): “learning based techniques have received significant attention for MVTS anomaly detection owing to (a) their ability to scale to high dimensions and model complex patterns in various domains, compared to straightforward statistical approaches such as out of limits approaches [5], [29], [30], (b) fast”
Learning Graph Structures With Transformer for Multivariate Time-Series Anomaly Detection in IoT (2021, background): “cn) In this work, we focus on anomaly detection for multivariate time series [8] as a copious amount of IoT sensors in many real-life scenarios consecutively generate substantial volumes of time series data.”
Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding (2021, methodology): “LSTM-NDT [14] comes from the authors’ implementation on https: //github.com/khundman/telemanom.”
TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis (2022, methodology): “For the short-term forecasting, following the N-BEATS (Oreshkin et al., 2019), we adopt the symmetric mean absolute percentage error (SMAPE), mean absolute scaled error (MASE) and overall weighted average (OWA) as the metrics, where OWA is a special metric used in M4 competition.”
ModernTCN: A Modern Pure Convolution Structure for General Time Series Analysis (2024, methodology): “For anomaly detection, we compare models on ﬁve widely-used benchmarks: SMD (Su et al., 2019), SWaT (Mathur & Tippenhauer, 2016), PSM (Abdulaal et al., 2021), MSL and SMAP (Hundman et al., 2018).”
Practical Approach to Asynchronous Multivariate Time Series Anomaly Detection and Localization (2021, methodology): “In [13], a method based on Long Short Term Memory with unsupervised Nonparametric Dynamic Thresholding (LSTM-NDT) for anomaly inference on predicted results was proposed.”
MOMENT: A Family of Open Time-series Foundation Models (2024, methodology): “In the zero-shot setting, MOMENT consistently outperformed all statistical interpolation methods with the exception of linear interpolation.”
Graph Neural Network-Based Anomaly Detection in Multivariate Time Series (2021, methodology): “…sensors, or even a single sensor) : To dampen abrupt changes in values are often not perfectly predicted and result in sharp spikes in error values even when this behavior is normal, similar with (Hundman et al. 2018b), we use a simple moving average(SMA) to generate the smoothed scores A s ( t ) .”
DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection (2023, methodology): “(2) SMAP (Soil Moisture Active Passive dataset) is also collected by NASA and presents the soil samples and telemetry information used by the Mars rover [32].”
Deep Learning for Time Series Anomaly Detection: A Survey (2022, neutral): “In all three models, LSTMs were stacked; however, LSTM-NDT [94] combines techniques from various domains, LSTM and RNN, to achieve high prediction performance by extracting historical information from MTS.”
Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network (2020, background): “These attacks were launched with different intents and diverse lasting durations (from a few minutes to an hour); (v) Mars Science Laboratory rover ( MSL ); and (vi) Soil Moisture Active Passive satellite ( SMAP ) data: Both MSL and SMAP are public data sets from NASA [8].”
Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy (2021, methodology): “(3) Both MSL (Mars Science Laboratory rover) and SMAP (Soil Moisture Active Passive satellite) are public datasets from NASA (Hundman et al., 2018) with 55 and 25 dimensions respectively, which contain the telemetry anomaly data derived from the Incident Surprise Anomaly (ISA) reports of spacecraft ”
TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data (2022, methodology): “The LSTM-NDT [20] method relies on an LSTM based deep neural network model that uses the input sequence as training data and, for each input timestamp, forecasts data for the next timestamp.”