Page History

Scripting environments like Matlab, Python and the OpenSim GUI shell allow users to interact with the classes of the OpenSim API (see The Introduction to the OpenSim API). There are many example scripts that are located in the OpenSim scripting folder, available with the distribution to help you get started. This page summarizes is resources of some of the common calls that you will make in the scripting environment of your choice. 

...

Beginner Scripting Resources  

• Run the example scripts found in the distribution.
• As you run the example scripts, find the related Classes and Methods in 

  New window link
  Body Doxygen Link https://simtk.org/api_docs/opensim/api_docs/

• The structure for classes can often be found in OpenSim's files (.xml or .osim). For example, Setup_Scale.xml shows the hierarchy of the ScaleTool.
• Use the

  New window link
  Body OpenSim Search bar Link http://opensim.stanford.edu/support/index.html

   to search for a Class or Method in the documentation, on the forum or on the Doxygen pages. 
• If you are having trouble with a Class or Method, ask a question on the 

  New window link
  Body OpenSim forum Link https://simtk.org/forums/viewforum.php?f=91&sid=c0e8a20a44fe5ca1524b3bb78fe03dcf&sid=26e02750f974c25f796217f05616c06a

  .

Inline Method and Type getting

...

Packages and libraries are collections of classes and methods (see Introduction to the OpenSim API for background) that have a well-defined interface and can be imported into your programming environment to utilize. These can be packages to browse files, read and write files, do mathematics operations, and run  run simulations. The commands below are common packages or libraries you will find useful.

...

>> newBody  = Body();                                        % Creates a body
>> massCenter = Vec3(-0.0707,0.0,0.0);                       % Create a Vec3 Object with mass center information
>> inertia = Inertia(0.128, 0.0871, 0.0579, 0, 0, 0);        % Create an Inertia object with only principal axes
>> newBody.setName('pelvis');                                % Set the name of the body using a string
>> newBody.setMass(11);                                      % Set the mass of the body using a double
>> newBody.setMassCenter(massCenter);                        % Set the mass center
>> newBody.setInertia(inertia);                              % Set the body's inertia 
 
%% Alternatively, a body can be created by sending all of these properties directly to the constructor
>> massCenter = Vec3(-0.0707,0.0,0.0);                       % Create a Vec3 Object with Mass center information
>> inertia = Inertia(0.128, 0.0871, 0.0579, 0, 0, 0);         % Create an Inertia object with only principal axes
>> bodyName = 'pelvis' ;                                      % Set the name of the body using a string
>> bodyMass = 11 ;                                            % Set the mass of the body using a double
>> newBody  = Body(bodyName, bodyMass, massCenter, inertia);  % Creates ana body from properties sent to constructor

...

>> pBody  = Body('pelvis', bodyMass, massCenter,inertiaMat inertia); 	% Either create or get a handle% toCreate the parent body
>> cBody  = Body('femur_r', bodyMass, massCenter,inertiaMat inertia);		% Either create or get a handle% toCreate the child body
>> name = 'hip_r'												% Assign the joint;                                              % Joint name
>> locInParent   = Vec3(-0.0707,-0.0661,0.0835);				% A Vec3 location in the parent frame
>> orienInParent = Vec3(0,0,0);									% A Vec3 orientation                  % Location of the joint origin expressed in the parent frame
>> locInChildoriInParent = Vec3(0,0.0,0);									% A Vec3 location                                   % Orientation of the joint frame in the childparent frame
>> orienInChildlocInChild = Vec3(0,0,0);									% A Vec3 orientation                                    % Location of the joint origin expressed in the child frame
>> rrHiporiInChild = BallJointVec3(name0,nBody,locInParent,orienInParent,cBody,locInChild,orienInChild)

>> myModel = Model('gait2354.osim');				# Create a Model Object
>> femur_r = myModel.getBodySet().get('femur_r');	# Get a Handle to the Body A
>> tibia_r  = myModel.getBodySet().get('tibia_r');	# Get a Handle to the Body B
>> zAxisTorque = Vec3(0,0,1);					# A Vec3 of the z-axis
>> torqueActuator = TorqueActuator();			# Create a Torque Actuator Object
>> torqueActuator.setBodyA(femur_r); 			# Set Body A
>> torqueActuator.setBodyB(tibia_r);			# Set Body B
>> torqueActuator.setAxis(zAxisTorque);			# Set the axis about the torque to act
>> torqueActuator.setOptimalForce(10);			# Set the optimal force (gain) of the actuator

>>> myModel = opensim.Model("arm26.osim")
>>> coordActuator = opensim.CoordinateActuator()
>>> coordActuator.setName('r_elbow_actuator')
>>> coordActuator.setCoordinate(myModel.getComponent('r_shoulder/r_shoulder_elev'))

Loading, Creating, and Initializing Models

Loading models and dealing with model states is very common. Below are the methods for loading, copying and initializing a model

>>> modelFileName = modelFolder+"/gait2354_simbody.osim" # Define the full path to the model file
>>> loadModel(modelFileName)                             0,0);                                    % Orientation of the joint frame in the child frame
>> rHip = BallJoint(name, pBody, locInParent, oriInParent, cBody, locInChild, oriInChild); % Construct the hip joint

>> myModel = Model('gait2354.osim');                % Create a Model Object from file
>> femur_r = myModel.getBodySet().get('femur_r');   % Get a handle to the femur
>> tibia_r  = myModel.getBodySet().get('tibia_r');  % Get a handle to the tibia
>> zAxis = Vec3(0,0,1);                             % A Vec3 of the z-axis
>> torqueActuator = TorqueActuator();               % Create a TorqueActuator Object
>> torqueActuator.setBodyA(femur_r);                % Set BodyA
>> torqueActuator.setBodyB(tibia_r);                % Set BodyB
>> torqueActuator.setAxis(zAxis);                   % Set the axis about which the torque will act
>> torqueActuator.setOptimalForce(10);              % Set the optimal force (gain) of the actuator

>>> myModel = opensim.Model("arm26.osim")
>>> coordActuator = opensim.CoordinateActuator()
>>> coordActuator.setName('r_elbow_actuator')
>>> coordActuator.setCoordinate(myModel.getComponent('r_shoulder/r_shoulder_elev'))

Loading, Creating, and Initializing Models

Loading models and dealing with model states is very common. Below are the methods for loading, copying and initializing a model

>>> modelFileName = modelFolder+"/gait2354_simbody.osim" # Define the full path to the model file
>>> loadModel(modelFileName)                             # Load the model into the GUI
>>> myModel = getCurrentModel()                          # Create a handle to the current model
>>> myState = myModel.initSystem()                       # Initialize the model and obtain the default state

...

...

>>> ECRB = myModel.getMuscles().get("ECRB")               # Get a handle to the ECRB
>>> backupTendonSlackLength = ECRB.getTendonSlackLength() # Back up the original tendon slack length (just in case)
>>> ECRB.setTendonSlackLength(0.2105)                     # Prescribe a new Tendon slack length
>>> myModel.initSystem()                                  # Re-initialize the model

You may need to downcast an object from an abstract class (e.g., Muscle)  to a derived class (e.g., Thelen2003Muscle) in order to gain access to its properties and methods. Here is an example:

>> import org.opensim.modeling.*
>> myModel = Model('arm26.osim');
>> mcl_TRIlong = Thelen2003Muscle.safeDownCast( myModel.getMuscles().get('TRIlong') );
>> mcl_TRIlong.setFmaxTendonStrain( 0.5*mcl_TRIlong.getFmaxTendonStrain() );

Using the Tools

Tools contain a number of grouped Methods that allow you to run a study. For example, to scale a Model to match experimental data the ScaleTool groups GenericModelMaker(), ModelScaler() and MarkerPlacer() together. The AnalyzeTool() can group StaticOptimization() and MuscleAnalysis() together to output muscle states of a Static Optimization.  

Working with Files and Paths

Below are examples of the OpenSim GUI shells native commands with files and file paths. Matlab and Python environments have their own methods for dealing with files and file paths. 

...

Parent and Concrete Classes

If you are calling a method or function (e.g., getting or setting properties) that you are pretty sure should work, but you are getting an error that the method doesn't exist, this may mean that you need to downcast. In the C++ programming language, programmers use a concept called "inheritance" to build up complexity without re-writing the same code multiple times. For example, in OpenSim Thelen2003Muscle and MillardMuscle both rely on code in the common parent Muscle class that they "inherit" from.

If you have a handle to a base class object (e.g. Muscle)  you may need to downcast the object to one of its derived (or concrete) classes, like the Thelen2003Muscle, in order to gain access to properties and methods specific to the concrete class. In the below example, we get a reference to a muscle in the model and return the class name and concrete class name;

>>> model = Model(path2model)
>>> muscle = model.getMuscles().get(0);
>>> muscle.getClassName()
muscle
>>> muscle.getConcreteClassName()
Thelen2003Muscle

Then to get a reference to the concrete class you use the safeDownCast() method. 

>>> muscle = model.getMuscles().get(0);    # The object you get here is of base class Muscle
>>> thelenMuscle = Thelen2003Muscle.safeDownCast(muscle)  # To use a method specific to Thelen2003Muscle you need to safeDownCast
>>> timeConstant = thelenMuscle.getActivationTimeConstant() # Get property that is present is specific to Thelen2003Muscle 

Outputs

OpenSim 4.0 uses component outputs and reporters to collect variables of interest and print them to file. To display the output names for a component, use the method getOutputNames();

>>> muscle.getOutputNames();

Reading Data from TRC Files into TimeSeriesTable

To read marker trajectories from a trc file into a TimeSeriesTableVec3, construct a table by passing the filename into the TimeSeriesTableVec3 constructor. Afterwards, you can query the TimeSeriesTable for the data as shown in the code snippet below.

>>> markerData  = TimeSeriesTableVec3('markerdata.trc');
>>> numDataRows = markerData.getNumRows();
>>> numMarkersInTable = markerData.getNumColumns();

Using the Tools

Tools contain a number of grouped Methods that allow you to run a study. For example, to scale a Model to match experimental data the ScaleTool groups GenericModelMaker(), ModelScaler() and MarkerPlacer() together. The AnalyzeTool() can group StaticOptimization() and MuscleAnalysis() together to output muscle states of a Static Optimization. Tools can be initialized from a setup file (.xml) that has stored settings. They also have methods that methods can be called to change the input models, data files, and some settings. Use the API documentation or methodsview() in Matlab (described above) to explore the methods that are available for the Tool you are using. 

Working with Vectors, Matrices, and Other SimTK Classes

We've exposed the most commonly used SimTK classes. In particular:

...

• For Simbody doxygen links above Vec3P corresponds to a modeling.Vec3 object and RealP corresponds to a double value.
• To see the available methods for these objects, use methodsview() or tab completion (Matlab only).
• You can find more information in the SimTK Basics section of the User's Guide. Note that only the SimTK classes listed above are available through scripting.
• You can also find more information in the section below on Handling C++ Templates.

ArrayDouble

In many cases the function you're trying to call takes an argument type different from the object you have already. This is true even moving data between Matlab and Java objects, and between objects in the SimTK namespace and those in the OpenSim namespace. In particular, many OpenSim methods return an ArrayDouble (Array<double) and you would prefer to convert the array to a different type. The following set of convenience methods are "Adaptors" intended to help you pass data around between OpenSim objects and low level SimTK objects.

...

model = Model('arm26.osim');
s = model.initSystem();
% For the given inputs, we will use the inverse dynamics operator
% (calcResidualForce()) to compute the first column
% of the mass matrix. We accomplish this by setting all inputs to 0
% except for the acceleration of the first coordinate.
%   f_residual = M udot + f_inertial + f_applied 
%              = M ~[1, 0, ...] + 0 + 0
model.realizeVelocity(s);
appliedMobilityForces = Vector();
appliedBodyForces = VectorOfSpatialVec();
knownUdot = Vector(s.getNU(), 0.0);
knownUdot.set(0, 1.0);
knownLambda = Vector();
residualMobilityForces = Vector();
smss = model.getMatterSubsystem();
smss.calcResidualForce(s, appliedMobilityForces, appliedBodyForces, ...
                       knownUdot, knownLambda, residualMobilityForces);

Handling C++ Templates (e.g. Array<double>)

Templates are advanced C++ constructs that are used extensively throughout the OpenSim API and Simbody API. If you see notation like Array<double> in the doxygen or C++ code that you are trying to replicate, this means you're working with a templatized class and will need to find its appropriate mapping in the scripting environment.

C++ templates allow the compiler to automatically generate classes from a common code base. This approach is useful primarily for containers (e.g. Array<double>, Array<int>, Array<string>) since it avoids code duplication making the code easier to maintain; however, these templates exist only in C++ and do not map easily to scripting languages.

One way to work around this is to specialize templates into classes, give those classes unique names, and then use them in scripting. The attached file lists template specializations used in OpenSim API (C++) & Doxygen and the corresponding named classes to be used in scripting (the list is current as of OpenSim 3.2):

Mapping of Templates to Class Names for Scripting

Obtaining Position and Velocity Information

For more information regarding multibody system states, refer to the SimTK Simulation Concepts documentation in the Developer's Guide.

In order to obtain simulation position or velocity state information you must have a State object in hand. 

>>> state = myModel.initSystem();
>>> myModel.equilibrateMuscles(state);

Now you can use methods on Body objects (actually, from the Frame class) to obtain the location or velocity of a point in the ground frame.

>>> body = myModel.getBodySet().get('r_ulna_radius_hand')
>>> massCenter = body.getMassCenter(massCenter)
>>> body.getStationVelocityInGround(state, massCenter)
>>> dir(body) # List other methods available on Body.

Using the Simbody Visualizer

You can use the visualizer from Simbody in Matlab and Python. To do so call the "setUseVisualizer" method and pass in the parameter "true", and when you run the simulation the Simbody Visualizer GUI will pop up. The example of usage is described in "TugOfWar_CompleteRunVisualizer.m"

osimModel = Model('tug_of_war_muscles_controller.osim');
osimModel.setUseVisualizer(true);
smss.calcResidualForce(s, appliedMobilityForces, appliedBodyForces, ...
                       knownUdot, knownLambda, residualMobilityForces);

Class Templates (Vec3(), Array<double>, Vector() )

A summary of Templatized Class use in scripting can be found on the Scripting Versions of OpenSim C++ API Calls page. 

Templates are advanced C++ constructs that are used extensively throughout the OpenSim API and Simbody API. If you see notation like Array<double> in the doxygen or C++ code that you are trying to replicate, this means you're working with a templatized class and, to use that class in scripting, will need to find its appropriate mapping in the scripting environment. For more information, and a easy to use ummary of the C++ to scripting mappings, see the Scripting Versions of OpenSim C++ API Calls page. 

Obtaining Position and Velocity Information

For more information regarding multibody system states, refer to the SimTK Simulation Concepts documentation in the Developer's Guide.

In order to obtain simulation position or velocity state information you must have a State object in hand. 

>>> state = myModel.initSystem();
>>> myModel.equilibrateMuscles(state);

Now you can use methods on Body objects (actually, from the Frame class) to obtain the location or velocity of a point in the ground frame.

>>> body = myModel.getBodySet().get('r_ulna_radius_hand')
>>> massCenter = body.getMassCenter(massCenter)
>>> body.getStationVelocityInGround(state, massCenter)
>>> dir(body) # List other methods available on Body.

Using the Simbody Visualizer

You can use the visualizer from Simbody in Matlab and Python. To do so call the "setUseVisualizer" method and pass in the parameter "true", and when you run the simulation the Simbody Visualizer GUI will pop up. The example of usage is described in "TugOfWar_CompleteRunVisualizer.m"

osimModel = Model('tug_of_war_muscles_controller.osim');
osimModel.setUseVisualizer(true);

Set verbosity or logging level

You can control how much detail is printed to the console using the following:

Logger.setLevel("Off");
Logger.setLevel("Critical");
Logger.setLevel("Error");
Logger.setLevel("Warn");
Logger.setLevel("Info"); % default
Logger.setLevel("Debug");
Logger.setLevel("Trace");

When performing batch processing or optimization you can turn off logging completely by calling Logger.setLevel("Off").

You can always add a new file during runtime to start logging into (in addition to whatever the current logging behavior is). This can be done by calling

Logger.addFileSink('logs_folder/full_path.log');

Make sure you have permission to write to the logs_folder.

Batch Processing

There are several examples in the Matlab scripts and GUI scripts that show how to perform batch processing by calling the OpenSim API (e.g. Analyze, IK). We encourage you to use this approach rather than using Matlab's xml parsing tools. To read more about why this is the case, please see the scripting FAQ:

...