Anomaly Detection in Force/Torque Time Series From Delta Robot
Collection of Jupyter notebooks and Python code implementations for anomaly detection in time series data, particularly in the context of robotic assembly.
Asset Description
This Python package contains a collection of Jupyter notebooks and Python code implementations for anomaly detection in multivariate time series, specifically focusing on force and torque data collected during robotic assembly processes using a delta robot. The methods include traditional statistical approaches as well as advanced deep learning models like LSTM networks.
Asset Details
Asset Content
The package offers multiple methods for anomaly detection, including Dynamic Time Warping (DTW), feature-based classification, LSTM neural networks, and AR models. Each method is provided in its own directory with corresponding Python scripts. The repository also includes pre-trained models, evaluation scripts, and example Jupyter notebooks that demonstrate the application of these methods on the publicly available dataset.
Usage
The repository is ideal for researchers and developers working on anomaly detection in time series data, particularly in the context of robotic assembly.
To use this repository, install the package:
| pip install ctuFaultDetector |
Detailed instructions and example code are provided in the Jupyter notebooks (presentation_LSTM, presentation_feature, presentation_nsigma). These notebooks demonstrate how to load the dataset, preprocess the data, train the models, and evaluate their performance.
Maturity
First version finished, but new versions are being developed, a more comprehensive documentation will be added in future
Licence
Open source - MIT License
Resources
- Python package available here
- Provided by Czech Technical University in Prague - CTU
Acknowledgement
This work was co-funded partly by AI REDGIO 5.0 and partly by the European Union under the project ROBOPROX(reg. no. CZ.02.01.01/00/22_008/0004590). In the AI REDGIO 5.0 project it will be used in the Didactic Factory Pilot DFXI: AI-driven Monitoring of Robotic Assembly Process