Dataset for "Deep learning based robot cognitive architecture for collaborative assembly tasks"
Dataset used for training action recognition method and task status prediction method for use in human-robot collaborative assembly scenario. Inertial measurement unit and skeleton tracking data included while users perform assembly tasks and gestures. Dataset has two parts; 1) human action recognition where sensor data is recorded along with ground truth of action being performed at that time 2) task status prediction where sensor data is recorded along with how long until each action in the task is completed.
Cite this dataset as:
Male, J.,
Martinez Hernandez, U.,
2023.
Dataset for "Deep learning based robot cognitive architecture for collaborative assembly tasks".
Bath: University of Bath Research Data Archive.
Available from: https://doi.org/10.15125/BATH-01161.
Export
Data
har_training_data.zip
application/zip (140MB)
Creative Commons: Attribution 4.0
Data files for use in training human action recognition methods in manufacturing style tasks. IMU and skeleton tracking data recorded with ground truth labels. Actions are screwdriver, Allen key, hand screw, hammer and various gestures.
status_prediction … data.zip
application/zip (59MB)
Creative Commons: Attribution 4.0
Data files for use in developing methods to predict the status of each user action in an assembly task. IMU and skeleton tracking data recorded along with ground truth data on how long until each action in a task is complete and whether the action has started and finished.
Contributors
University of Bath
Rights Holder
Documentation
Data collection method:
The methodology can be found in the associated publication.
Additional information:
[Corrigendum 2023-10-09] An error during the recording process has led to data from the arm sensor being triplicated. This can be seen in the IMU data where columns 3–8, 9–14 and 15–20 in the files in har_training_data.zip are identical, as are columns 2–7, 8–13 and 14–19 in the IMU files in status_prediction_training_data.zip. Different recognition results may therefore be achieved when using the methods detailed in the associated publication on new data, though the underlying methods should not need modification. The authors apologise for this oversight.
Methodology link:
Male, J., and Martinez-Hernandez, U., 2023. Deep learning based robot cognitive architecture for collaborative assembly tasks. Robotics and Computer-Integrated Manufacturing, 83, 102572. Available from: https://doi.org/10.1016/j.rcim.2023.102572.
Documentation Files
readme.txt
text/plain (1kB)
Overview of the subsets of this dataset
har_readme.txt
text/plain (2kB)
Documentation of human action recognition training data subset
task_status_prediction_readme.txt
text/plain (2kB)
Documentation of status prediction training data subset
Funders
Engineering and Physical Sciences Research Council
https://doi.org/10.13039/501100000266
PhD Studentship
Publication details
Publication date: 4 April 2023
by: University of Bath
Version: 1
DOI: https://doi.org/10.15125/BATH-01161
URL for this record: https://researchdata.bath.ac.uk/id/eprint/1161
Related papers and books
Male, J., and Martinez-Hernandez, U., 2023. Deep learning based robot cognitive architecture for collaborative assembly tasks. Robotics and Computer-Integrated Manufacturing, 83, 102572. Available from: https://doi.org/10.1016/j.rcim.2023.102572.
Contact information
Please contact the Research Data Service in the first instance for all matters concerning this item.
Contact person: James Male
Faculty of Engineering & Design
Electronic & Electrical Engineering
Research Centres & Institutes
Centre for Autonomous Robotics (CENTAUR)
