Code associated with manuscript "Exploration-based learning of a stabilizing controller predicts locomotor adaptation" If you use or build on this code in any way, please cite the following:
Authors: N. Seethapathi, B. L. Clark, and M. Srinivasan.
Title: Exploration-based learning of a stabilizing controller predicts locomotor adaptation.
Journal: To be announced.
Volume and pages: To be announced.
Year: 2024.
The following assumes that you have downloaded the zip file onto your computer, and have un-zipped the .zip file to create a folder. The unzipped folder will be programsForManuscriptSeethapathiEtAl22 by default.
If you have MATLAB already installed on your computer, please directly run the following program within MATLAB: rootSimulateLearningWhileWalking.m This program calls other functions as necessary to perform the simulation, and will generate the key figures.
To run the program in MATLAB, navigate to the folder containing the code, and type rootSimulateLearningWhileWalking into the command prompt and press 'enter' or 'return':
rootSimulateLearningWhileWalking <enter/return>
If don't have MATLAB on your computer or if the above step does not work, see the next section and come back to this section.
The program is written from scratch in the MATLAB programming environment.
The program is open source in the sense you can view the entire code here.
The code is self-contained and does not use any special functions. The only toolbox that may possibly need to be installed is the MATLAB Optimization toolbox, but this may not be necessary depending on your MATLAB installation.
The programs were tested in MATLAB Version: 9.12.0.1927505 (R2022a). However, the programs will run in any older version of MATLAB, as the programs do not require any special functionality available only in the newer versions.
MATLAB is available from Mathworks: https://www.mathworks.com
One may use the program either on a personal computer (eg., Desktop or Laptop) or entirely on the cloud through a browser via MATLAB online: https://www.mathworks.com/products/matlab-online.html#license-types
Octave is a free clone of MATLAB and is available here under the GNU General Public License (GPL). https://octave.org MATLAB programs that don't use special toolboxes, as here, can be run on Octave essentially verbatim. You should be able to run our programs on Octave.
-
The simulation parameters and conditions are encoded within the various loadParameters*** files.
-
See the Methods section and Supplementary Appendix of the manuscript for the mathematical equations underlying the simulation.
-
Because the simulation is stochastic, running the program two times will give very slightly different curves, while having broad statistical similarity as the default noise is low.
-
If your program hangs for whatever reason, please exit the program via crtl-c or exit MATLAB through a force quit, and then re-start the program. The qualitative results are invariant to substantial changes to the learning parameters. But if the parameters are outside of the stability region (as discussed in the manuscript), the biped or the learner may go unstable, resulting in numerical challenges that may result in the program hanging, as in case-H below.
-
The following are some things to try to examine different aspects of the results, as a quick demonstration. This list is by no means comprehensive, as the code allows infinite numerical experiments; these are suggested for qualitative understanding of the results and the structure of the code.
-
In the main program rootSimulateLearningWhileWalking.m, the section with the header "%% define the different adaptation protocols and figures"
You will notice that all lines are commented out except the line: paramFixed.runType = 'figure2aRegularSplitBelt';
Running the code with nothing changed will simulate a protocol corresponding to figure 2a, which is a split-belt walking protocol with a tied belt condition before and after.
Now if you comment out paramFixed.runType = 'figure2aRegularSplitBelt'; but uncomment any other line in that section, the program will run a simulation corresponding to that condition. For instance, uncommenting the following: paramFixed.runType = 'figure3dDifferentSensoryNoiseLevelsHigher0p2'; will run the default split-belt simulation but now with much higher sensory noise. By the uncommenting appropriate sentences in this section, the reader can run 40 different simulation settings and adaptation protocols.