This is a dataset maintained by the Medical Robotics and Devices Lab at the University of Minnesota, for the purpose of allowing researchers to evaluate different models of torque and jaw angle estimation.
This dataset is made available under Open Database License whose full text can be found here. Any rights in individual contents of the database are licensed under the Database Contents License whose text can be found here.
GNU General Public License v3.0
This folder contains a series of data runs characterizing the input and output of a da Vinci tool for creation and validation of grip force/torque and jaw angle estimation methods.
The general file naming scheme is either XY.TXT or ZXY.TXT, where:
- X Corresponds to different frequencies
- Y Corresponds to different resistive torque levels
- Z Corresponds to different jaw angle stopping positions (also referred to as yaw changes) The Z value is only used in the roll/pitch/yaw varying dataset, to denote different yaw values. The different roll and pitch values are separated by folders, and the roll and pitch tested are ONLY denoted in the readme.txt for that folder, not in the header.
The SerialLog.txt file may or may not be included in a directory, and contains the debug output of the microcontroller during each run. It is not anticipated that this file will be necessary for analysis.
Each .TXT file has header data which will provide some insight into what the data contains, however for torque, position and frequency the data itself should be able to be analyzed to detect these values, and that should take precedence over any values found in the header, as some header text was changed manually during data collection and could be inaccurate.
The data files should have 10 columns:
- Back-end Time [ms] (interrupt running at 1 kHz)
- Back-end CUI Encoder [counts] (8192 counts/rev)
- Back-end Futek Torque [counts] (0.000039811 Nm/count)
- Back-end Measured Current [Amps] (from Maxon controller)
- Front-end Maxon Encoder [counts] (4000 counts/rev)
- Front-end Futek Torque [counts] (0.000039811 Nm/count)
- Front-end Command [amps] (analog out, sent to Maxon controller)
- Back-end Command [-1:1] (analog out, sent to Maxon controller)
- Back-end Maxon [counts] (4196*35 counts/rev; 35:1 gearbox)
- Back-end Trajectory Target [counts] (fraction of counts from trajectory)
For convenience and clarity, MATLAB code is provided in the visualization directory, which has a file called basic_data.m will load and parse the data, and provide some basic plots. The plotting code should work on GNU Octave as well.
In the interest of full disclosure, the full code used for the RA-L paper "Evaluation of Torque Measurement Surrogates as Applied to Grip Torque and Jaw Angle Estimation of Robotic Surgical Tools" has been provided in this repository as well. This code consists of two script files to train the neural networks (train_roll.m
and train_torque.m
) as well as one script to generate the figures and tables used in the paper (plot_torque.m
). The actual neural nets used in the paper are provided in the Neural_Nets
directory, and as such the plot_torque
script can be run without training and it will use these networks.
Note that the code requres MATLAB, and training requires the Neural Network Toolbox, which may not be included in all MATLAB licences. Evaluating the networks (plot_torque.m
) should be possible without the toolbox.
Code is included to export the run-time part of the Neural Network to C++ code. This function is only programmed to work with a narrow subset of NN objects, which are shown tested test_manual_nn.m
. If you want to do more complicated NN's you will either have to extend this function, or switch to an automated Matlab->C++ tool. This exporting should be able to function without the Neural Network Toolbox.
Please see the following papers for more details on how this dataset has been created and used:
- John J. O'Neill, Trevor K. Stephens, and Timothy M. Kowalewski. Evaluation of Torque Measurement Surrogates as Applied to Grip Torque and Jaw Angle Estimation of Robotic Surgical Tools, in IEEE Robotics and Automation Letters, vol. 3, no. 4, pp. 3027-3034, Oct. 2018. [pdf]
- Nathan J. Kong, Trevor K. Stephens, John J. O'Neill, and Timothy M. Kowalewski. Design of a Portable Dynamic Calibration Instrument for da Vinci Si Tools. In Design of Medical Devices Conference, pages V001T11A023-V001T11A023, Minneapolis, MN, 2017. American Society of Mechanical Engineers. [pdf]
If you would like to cite this dataset in your own work, we would appreciate it if you could cite the 2018 RA-L paper:
@ARTICLE{8392722,
author={John J. O’Neill and Trevor K. Stephens and Timothy M. Kowalewski},
journal={IEEE Robotics and Automation Letters},
title={Evaluation of Torque Measurement Surrogates as Applied to Grip Torque and Jaw Angle Estimation of Robotic Surgical Tools},
year={2018},
volume={3},
number={4},
pages={3027-3034},
doi={10.1109/LRA.2018.2849862},
month={Oct},}
If you have any questions about the data or the code, please feel free to open an issue, or contact @john-j-oneill or @trevor-k-stephens directly and we will be happy to help.