Neuromuscular-Transfemoral-Prosthesis-Model

The repository contains a 3D neuromuscular model of

1. a healthy person
2. a trans-femoral amputee wearing the Otto Bock 3R60 knee prosthesis
3. a trans-femoral amputee wearing the Otto Bock 3R60 knee prosthesis with CMG embedded in the prosthetic shank

Details can be found in the thesis on the TU Delft repository.

The model is based on the work of:

• Song, S., & Geyer, H. (2015). A neural circuitry that emphasizes spinal feedback generates diverse behaviours of human locomotion. The Journal of physiology, 593(16), 3493-3511.
• Thatte, Nitish, and Hartmut Geyer. "Toward balance recovery with leg prostheses using neuromuscular model control." IEEE Transactions on Biomedical Engineering 63.5 (2016): 904-913.

Demo

You can run DEMO.m. This demo script allows you to plot and/or animate results. The results can be obtained through simulation, or by just loading the data included in the repository. Note: the simulation will take some time (45-90 min). Also the model is numerically sensitive, especially the trip prevention simulation. This means that when simulating the model on your computer it can obtain a different, maybe even unsuccessful, result.

Running a model

1. Run the setup_paths.m script.
2. Go to the folder for the model you wish to run.
3. Open the model
4. Load one of the optimized gains .mat files in the Results folder
5. Select 'rapid-accelerator' mode in Simulink ('normal' mode can work as well, however, nicer visualization with 'rapid-accelerator' mode in combination with animPost()
6. Run evaluateCost.m
7. If 'normal' simulation mode is selected, the simulation is visualized. If 'rapid-accelerator' mode is selected, the simulation can be animated using animPost().
8. Simulation data can be plotted using the plotData() function.

Optimizing the model

1. Run the setup_paths.m script.
2. Go to the folder for the model you wish to optimize.
3. Open optimize.m
4. Select the desired initial gains
5. Check if the settings are as you wish in setInnerOptSettings
6. Select which data you want to have plotted during optimization in plotProgressOptimization
7. Run optimize.m

Animating the data

• animPost function. See the animPost3D function for all the options and specifics.
• Example of running animPost:
  • animPost(simout(1).animData3D,'intact',false)
    • Shows an animation of the simulated amputee model
  • animPost(simout(1).animData,'intact',false,'speed',1,'obstacle',true,'view','side','CMG',true,'showFigure',true,'createVideo',true,'info','prosthetic1.2ms','saveLocation','Videos');
    • Shows an animation of the simulated amputee model walking with CMG and tripping over obstacle. A mp4 file is created and saved. The filename contains prosthetic1.2ms and it is saved in a subfolder called Videos located in the current directory

Plotting the data:

• plotData function. Check the function for all the options and specifics
• GaitPhaseData and stepTimes are required. Other data sets are optional for plotting
• Example of running plotData:
  • plotData(simout(1).GaitPhaseData,simout(1).stepTimes,simout, 'angularData',simout(1).angularData, 'musculoData',simout(1).musculoData, 'GRFData',simout(1).GRFData, 'jointTorquesData',simout(1).jointTorquesData, 'CMGData',simout(1).CMGData,'prosthetic3D_1.2ms_yaw',[7 10],0,1,1,0);
    • Shows the data between 7 and 10 seconds
  • plotData(simout(1).GaitPhaseData,simout(1).stepTimes,simout, 'angularData',simout(1).angularData, 'musculoData',simout(1).musculoData, 'GRFData',simout(1).GRFData, 'jointTorquesData',simout(1).jointTorquesData, 'CMGData',simout(1).CMGData,'prosthetic3D_1.2ms_yaw',[],1,1,1);
    • Shows and saves the data. Presented as average data and standard deviation per stride. Saved filename contains prosthetic3D_1.2ms_yaw

Notes:

• Model developed in Matlab 2019b. When running in Matlab 2020a a problem arose for parsim. The option TransferWorkspaceVariables should be added and set to on.
• A compiler needs to be installed
  • https://www.mathworks.com/help/matlab/matlab_external/choose-c-or-c-compilers.html
  • MinGW compiler should work, however,
  • Compilers used:
    • MEX configured to use 'Microsoft Visual C++ 2015 (C)' for C language compilation.
    • CompilerConfiguration with properties:
      • Name: 'Microsoft Visual C++ 2015 (C)'
      • Manufacturer: 'Microsoft'
      • Language: 'C'
      • Version: '14.0'
    • MEX configured to use 'Microsoft Visual C++ 2017' for C++ language compilation.
    • CompilerConfiguration with properties:
      • Name: 'Microsoft Visual C++ 2017'
      • Manufacturer: 'Microsoft'
      • Language: 'C++'
      • Version: '15.0'
• Model is numerically sensitive.
  • Gains were obtained while sampling the data at 30 Hz. Changing this will change the evaluations.
  • On flat terrain this does not differ too much, however, on rougher terrain it might result in an unsuccesful gait.
  • Adding integrator blocks will also affect the outcome, since the solver variable step takes different amount of steps.
• If certain parameters need to be changed in between runs during optimization, don't forget: https://www.mathworks.com/help/physmod/simscape/ug/set-the-run-time-parameters-visibility-preference.html

This model is available for Academic or Non-Profit Organization Noncommercial research use only.