First, clone this repository. Then, open MatLab, navigate to the folder where the BioMAC-Sim-Toolbox is located and add all folders and subfolders from "src" to the path:
cd(what('BioMAC-Sim-Toolbox').path); % If not already in the BioMAC-Sim-Toolbox folder
addpath(genpath('src'));
savepath;Alternatively, you can manually add the toolbox to the path by going to the "Current Folder Browser" and navigate to the folder where "src" is located. Right click, go to "add to path" and click "selected folders and subfolders".
When calling a model for the first time, it will be compiled automatically. If you would like to compile on different systems on the same machine, you need to manually delete the .o files after building the model files to ensure that you can build on the other system.
Compilers
First, make sure that MatLab Compiler is installed. For Windows, mingw is needed. For instructions, see here: https://nl.mathworks.com/matlabcentral/answers/311290-faq-how-do-i-install-the-mingw-compiler. For Linux/MacOS, gcc should be installed by default.
IPOPT
We recommend using the following version which can be obtained through the Add-On Explorer in MATLAB: https://github.com/ebertolazzi/mexIPOPT. For Mac Users using Apple Silicon native MatLab versions, follow these instructions: ebertolazzi/mexIPOPT#24
OpenSim (optional)
When using the model gait3d, or gait2d_osim, it is recommended to install OpenSim and its MatLab API from https://simtk.org/frs/?group_id=91. We currently support versions 3.3, 4.0, 4.1, 4.3, and 4.5. For Mac Users using Apple Silicon native MatLab versions, OpenSim is not supported currently.
You can use the model gait3d, or gait2d_osim, without an OpenSim installation: We have included .mat files that allow you to use the base models and those used in the ExampleScripts without an OpenSim installation. In that case, you should not recalculate the moment arms, which requires OpenSim.
This is an instruction to start using the BioMAC-Sim-Toolbox. Extended documentation can be found here: https://mad-lab-fau.github.io/BioMAC-Sim-Toolbox/
The BioMAC-Sim-Toolbox contains three standard models: OpenSim-based Gait2d_osim and gait3d, as well as gait2dc, which is defined by a spreadsheet. A model can be instanciated using its definition file:
model_3d = Gait3d('gait3d_pelvis213.osim'); %or
model_2d_osim = Gait2d_osim('gait2d.osim'); %or
model_2dc = Gait2dc('gait2dc_par.xls');In this example, a Collocation object is created for a simulation of 40 Nodes duration and backward Euler integration:
nNodes = 40; Euler = 'BE'; logfile = 'example.log'; plotLog = 1;
problem = Collocation(model_2dc, nNodes, Euler, logfile, plotLog);Next, we define which variables we want to optimize: states and controls. For all, we use the default lower and upper bounds that our model provides. Our target is a simulation at 3.5m/s, therefore, we fix the speed and set the duration loosely to an interval between 0.2s and 2s.
xmin = repmat(model_2dc.states.xmin, 1, nNodes+1);
xmax = repmat(model_2dc.states.xmax, 1, nNodes+1);
xmax(model_2dc.extractState('q', 'pelvis_tx'), 1) = 0;
xmin(model_2dc.extractState('q', 'pelvis_tx'), 1) = 0;
problem.addOptimVar('states', xmin, xmax);
problem.addOptimVar('controls',repmat(model_2dc.controls.xmin,1,nNodes+1), repmat(model_2dc.controls.xmax,1,nNodes+1));
problem.addOptimVar('dur',0.2, 2);
problem.addOptimVar('speed',3.5,3.5);Then, we supply the initial guess to our problem. For simplicity, we set the initial guess to 'mid', which is the middle between upper and lower bound. Setting a good initial guess can speed up the simulation, see the ExampleScripts.
problem.makeinitialguess('mid'); Objectives In this simple example, we only optimize for muscle effort, therefore we set the weight to 1. An objective can be added the following way:
Weffort = 1;
weightsType = 'equal';
exponent = 3;
problem.addObjective(@effortTermMuscles, Weffort, weightsType, exponent);Constraints
In our example, we make use of @dynamicConstraints for dynamically consistent output. We also set @periodicityConstraint for half-gait-cycle periodicity.
problem.addConstraint(@dynamicConstraints,repmat(model_2dc.constraints.fmin,1,nNodes),repmat(model_2dc.constraints.fmax,1,nNodes))
problem.addConstraint(@periodicityConstraint,zeros(model_2dc.nStates+model_2dc.nControls,1),zeros(model_2dc.nStates+model_2dc.nControls,1),1)Calling IPOPT By adding objectives and constrained, the problem is now defined and can be solved using IPOPT in our default settings:
solver = IPOPT();
result = solver.solve(problem);Inspecting Results
To show a video of the resulting motion, call the Collocation.writeMovie function.
result.problem.writeMovie(result.X);Note that the output of this example isn't quite realistic, as muscle effort alone is not sufficient to describe walking or running.
The above example is a simplified version of script2D, and we created introduction examples for all models in the toolbox:
- gait2dc:
IntroductionExamples.script2D - gait2d_osim:
Treadmill.script2D - gait3d:
IntroductionExamples.script3D
To run any of the introduction scripts, make sure that the folder "ExampleScripts" is also on your MATLAB path. The folders that are named starting with a + will then also be part of the path.
If you are using our toolbox in your research, we would be glad for a citation. Our paper is still under review, we will update the citation as soon as it is available. Follow Anne Koelewijn on BlueSky or LinkedIn for updates.