Scripting environments like Matlab, Python and the OpenSim GUI shell allow users to interact with the classes of the OpenSim API (see 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 is resources of some of the common calls that you will make in the scripting environment of your choice.
These are inline commands that help you find the methods available for a class
|methodsview('Model') or methodsview(osimModel)||GUI, Matlab||Examine the methods available for a class (e.g. Model) or for an existing object that you've created (e.g. osimModel)|
|type(aObject)||GUI, Python||Prints the full qualified type of aObject|
|Up/Down Arrows at the command prompt||GUI, Interactive Python||Recalls past commands entered at the prompt|
|Tab completion||Matlab, Interactive Python||Define an OpenSim model or object then use tab completion in Matlab to see the available methods|
|dir(aObject)||Python||Returns a list of methods available on aObject|
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 simulations. The commands below are common packages or libraries you will find useful.
|java.swing||GUI||Swing is the generic GUI kit for Java, allows for I/O and creating windows if needed. Examples of usage can be found here.|
|java.lang||GUI||Lang is the Java language core libraries e.g String, Math, etc. To use particular libraries (for example Math), do "from java.lang import Math"|
|java.io||GUI||IO is the package in Java that's used to perform input/output operations including file reading/writing.|
|org.opensim.modeling||GUI and Matlab||Modeling classes from the OpenSim API. This package is automatically imported in the GUI Shell. In Matlab, excute command import org.opensim.modeling.* to avoid having to type org.opensim.modeling.'Class'|
|org.opensim.utils||GUI||Provides some convenient file browsing options|
>>> import java.io as io # Import Read/Write package >>> trcFolder = io.File(trcDataFolder) # Use package to read TRC file
Creating New Objects
Creation of objects is performed by the constructor method of the Class. We can create an object and populate its properties, or pass the class a correctly formatted xml file.
Note: 'path' indicates your system path to the file (i.e., C:/Users/<username>/Documents/OpenSim/4.0/Models/...).
Note: Delete "modeling." from code for Matlab.
|Action||Class Information||Default construction (Scripting shell)||XML construction (Scripting shell)|
|Marker set object||markerSet = modeling.MarkerSet()||markerSet = modeling.MarkerSet([path 'markerSet.xml'])|
|Marker data object||markerData = modeling.MarkerData()||markerData = modeling.MarkerData([path 'walkingData.trc'])|
|Scaling tool object||scaleTool = modeling.ScaleTool()||scaleTool = modeling.ScaleTool([path 'scale_setup.xml'])|
|Inverse Dynamics analysis object||idTool = modeling.InverseDynamicsTool()||idTool = modeling.InverseDynamicsTool([path 'ID_setup.xml'])|
|Control Set||controlSet = modeling.ControlSet()||controlSet = modeling.ControlSet([path 'controls.mot'])|
>>> myModel = modeling.Model("gait2354.osim") # Load a Model from file >>> markerSetFile = "gait2354_MarkerSet.xml" # Define the full path to the MarkerSet file >>> newMarkers = modeling.MarkerSet(myModel, markerSetFile) # Construct a MarkerSet Object >>> myModel.updateMarkerSet(newMarkers) # Append newMarkers to the Model's MarkerSet, replacing markers with the same name
>>> myModel = modeling.Model("gait2354.osim") # Create a Model Object >>> pelvisbody = myModel.getBodySet().get("pelvis") # Get a handle to the pelvis body >>> newMarker = modeling.Marker("LASI", pelvisbody, modeling.Vec3(1,1,1)) # Create a Marker called 'LASI' that is attached to the pelvis at (1,1,1) >>> myModel.addMarker(newMarker) # Add newMarker to the Model
>> 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 >> newBody.setMass(11); % Set the mass of the body >> 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 >> bodyMass = 11; % Set the mass of the body >> newBody = Body(bodyName, bodyMass, massCenter, inertia); % Creates a body from properties sent to constructor
>> pBody = Body('pelvis', bodyMass, massCenter, inertia); % Create the parent body >> cBody = Body('femur_r', bodyMass, massCenter, inertia); % Create the child body >> name = 'hip_r'; % Joint name >> locInParent = Vec3(-0.0707,-0.0661,0.0835); % Location of the joint origin expressed in the parent frame >> oriInParent = Vec3(0,0,0); % Orientation of the joint frame in the parent frame >> locInChild = Vec3(0,0,0); % Location of the joint origin expressed in the child frame >> oriInChild = Vec3(0,0,0); % 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 models and dealing with model states is very common. Below are the methods for loading, copying and initializing a model
|Action||GUI Command||Matlab/Python Command|
|Loads the specified model in the GUI||loadModel(modelFileName)|
|Creates a handle to the current model in the GUI||myModel = getCurrentModel()|
|Load a model from file (Create Model Object)||myModel = modeling.Model(modelFileName)||myModel = Model('gait2354.osim')|
|Creates a copy of myModel.||myModelCopy = modeling.Model(myModel)||myModelCopy = Model(myModel)|
|Initialize the model and get the default state||myState = myModel.initSystem()||myState = myModel.initSystem()|
>>> 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
You can use the functionality of the OpenSim API to access the properties of a model and change their values. For examples, refer to the distributed scripts muscleScaler and alterTendonSlackLength. The OpenSim API Doxygen lists all of the available functions. For example, executing the following commands in the GUI scripting shell, sets and gets the name of the current model.
>>> myModel = getCurrentModel() >>> myModel.setName("My Model") >>> mymodel.getName() >>> "my Model"
You should generally avoid adding and removing objects from a model that is "live" in the OpenSim GUI. Instead you should make a copy of the model, make additions and deletions, then reload in the GUI.
The API also allows you to access and edit the components of an OpenSim model, like its bodies, muscles, and joints. Some properties have custom "get" and "set" functions - see the respective classes for details.
|Action||Class Information||Getting a Handle to a Set||Reference an Object by index||Get object from a name||Alternative 'long path'|
|Body Set||bodySet= myModel.getBodySet()||rightFemur = bodySet().get(1)||rightFemur = bodySet().get("femur_r")||rightFemur =myModel.getBodySet().get("femur_r")|
|Joint Set||jointSet = myModel.getJointSet()||rightHip = jointSet ().get(7)||rightHip = jointSet().get("hip_r")||rightHip = myModel.getJointSet().get("hip_r")|
|Coordinate Set||cordSet= myModel.getCoordinateSet()||hip_coord = cordSet().get(4)||hip_coord = cordSet().get("hip_flexion_r")||hip_coord = cordSet().get("hip_flexion_r")|
|Muscle Set||muscleSet= myModel.getMuscles()||recFemR = muscleSet().get(3)||rec_fem_r = recFemRt().get("'recFem_r")||rec_fem_r= myModel.getMuscles().get("'recFem_r")|
|Path Point of a Muscle||pathPoints = myModel.getMuscles().get("'recFem_r").getGeometryPath().getPathPointSet()||recFemPathPoint1 = pathPoints.get(0)||recFemPathPoint1 = pathPoints.get("rect_fem_r-P1")||recFemPathPoint1 = myModel.getMuscles().get("'recFem_r").getGeometryPath().getPathPointSet()|
Once a Handle to the component has been created you can edit its properties and methods.
|Get the optimal fiber length||recFemFiberLength = RectusFemoris.getOptimalFiberLength()|
|Set the optimal fiber length||RectusFemoris.setOptimalFiberLength(0.23)|
|Set the tendon slack length||RectusFemoris.setTendonSlackLength(0.2105)|
|Get the muscle maximum isometric force||recFemMaxForce = RectusFemoris.getMaxIsometricForce()|
>>> 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() );
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
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();
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();
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.
|Command or Class||Platform||Action|
|scale = ScaleTool('Scale_Setup_file.xml')||Matlab||Returns a Scaletool object with properties defined in the Scale_Setup_File.xml|
ik = InverseKinematicsTool()
Returns an InverseKinematicsTool object with default properties. You must set the properties of the tool.
Begin the Inverse Kinematics simulation by calling the run() method.
rra = RRATool'RRA_Setup.xml')
Returns an RRATool object with properties defined in the RRA.
Allow RRA to alter the trunk COM by using the setAdjustCOMToReduceResiduals() method.
Begin the RRA simulation by calling the run() method.
|so = AnalyzeTool('SO_Setup.xml')||Matlab||Returns an AnalzeTool that has a Static Optimization analysis included.|
We've exposed the most commonly used SimTK classes. In particular:
modeling.Vec3(double e0, double e1, double e2)
|All||Creates a Simtk Vec3 object (a vector of length 3). If passed only one argument (double e) all elements will be set to e.|
|modeling.Vector(int length, double e)||All||Creates a Simtk Vector with specified length. All elements are set to e.|
modeling.Mat33(double e0, ... , double e8)
|All||Creates a Simtk Mat33 object (a 3x3 matrix). If passed only one argument (double e), the diagonal elements will be set to e and other elements set to zero.|
See doxygen link at right for additional constructors
|All||Creates a Simtk Inertia object. All constructors are available except symmetric matrix constructors.|
|modeling.State()||All||Creates a Simtk State object. See the doxygen link for more information.|
Creates a Simtk Stage object, optionally realized to level l. See the doxygen link for more information.
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.
For Matlab, leave off "modeling". For Python, change "modeling" to "opensim".
|modeling.ArrayDouble.createVec3([0.0,0.05,0.35])||All||Creates a SimTK::Vec3|
|All||returns SimTK::Vec3 populated from ArrayDouble of size 3.|
|modeling.ArrayDouble.getAsVector()||All||return SimTK::Vector populated from ArrayDouble|
|modeling.ArrayDouble.populateFromVector(SimTK::Vector aVector)||All||populate an ArrayDouble from the passed in SimTK Vector|
|modeling.ArrayDouble.getValuesFromVec3(SimTK::Vec3 vec3)||All||returns an ArrayDouble populated from the passed in SimTK Vec3|
When referring to indexed elements remember that Matlab begins indexing at 1 while OpenSim data structures begin at 0.
In 4.0, you can perform advanced multibody calculations in scripting via the SimbodyMatterSubsystem inside the OpenSim Model (model.getMatterSubsystem()). The SimbodyMatterSubsystem class allows you to compute the mass matrix, Jacobians, inverse dynamics moments, etc.
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);
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.
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.
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);
Applies to OpenSim 4.2 (unreleased) and above.
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
Make sure you have permission to write to the logs_folder.
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:
Frequently Asked Questions
Next: Scripting in the GUI