API  4.5
For C++ developers
OpenSim::PolynomialPathFitter Class Reference

A utility class for fitting a set of FunctionBasedPaths to existing geometry-path in an OpenSim model using MultivariatePolynomialFunctions. More...

+ Inheritance diagram for OpenSim::PolynomialPathFitter:

OpenSim Properties, Sockets, Outputs, Inputs

Properties (single-value)
ModelProcessor model
 "The model containing geometry-based path objects to which " "polynomial-based path objects will be fitted." More...
 
TableProcessor coordinate_values
 "The reference trajectory used to sample coordinate values for " "path fitting." More...
 
std::string output_directory
 "The directory to which the path fitting results are written." More...
 
double moment_arm_threshold
 "The moment arm threshold value that determines whether or not a " "path depends on a model coordinate. In other words, the moment " "arm of a path with respect to a coordinate must be greater than " "this value to be included during path fitting." More...
 
int minimum_polynomial_order
 "The minimum order of the polynomial used to fit each path. The " "order of a polynomial is the highest power of the independent " "variable(s) in the polynomial." More...
 
int maximum_polynomial_order
 "The maximum order of the polynomial used to fit each path. The " "order of a polynomial is the highest power of the independent " "variable(s) in the polynomial." More...
 
SimTK::Vec2 global_coordinate_sampling_bounds
 "The global bounds (in degrees) that determine the minimum and " "maximum coordinate value samples at each time point." More...
 
double moment_arm_tolerance
 "The tolerance on the root-mean-square (RMS) error (in meters) " "between the moment arms computed from an original model path and " "a fitted polynomial-based path, which is used to determine the " "order of the polynomial used in the fitted path (default: 1e-4)." More...
 
double path_length_tolerance
 "The tolerance on the root-mean-square (RMS) error (in meters) " "between the path lengths computed from an original model path and " "a fitted polynomial-based path, which is used to determine the " "order of the polynomial used in the fitted path (default: 1e-4)." More...
 
int num_samples_per_frame
 "The number of samples taken per time frame in the coordinate " "values table used to fit each path (default: 25)." More...
 
int num_parallel_threads
 "The number of threads used to parallelize the path fitting " "process (default: two fewer than the number of available " "hardware threads)." More...
 
std::string latin_hypercube_algorithm
 "The Latin hypercube sampling algorithm used to sample coordinate " "values for path fitting (default: 'random')." More...
 
Properties (list)
PolynomialPathFitterBounds coordinate_sampling_bounds
 "The bounds (in degrees) that determine the minimum and maximum " "coordinate value samples at each time point for specific " "coordinates. These bounds override the default coordinate " "sampling bounds." More...
 

Public Member Functions

 PolynomialPathFitter ()
 
 ~PolynomialPathFitter () noexcept override
 
 PolynomialPathFitter (const PolynomialPathFitter &)
 
 PolynomialPathFitter (PolynomialPathFitter &&)
 
PolynomialPathFitteroperator= (const PolynomialPathFitter &)
 
PolynomialPathFitteroperator= (PolynomialPathFitter &&)
 
void setModel (ModelProcessor model)
 The model containing geometry-based path objects to which polynomial-based path objects will be fitted. More...
 
void setCoordinateValues (TableProcessor coordinateValues)
 The reference trajectory used to sample coordinate values for path fitting. More...
 
void run ()
 Run the path fitting process. More...
 
void setOutputDirectory (std::string directory)
 The directory to which the path fitting results are written. More...
 
std::string getOutputDirectory () const
 
void setMomentArmThreshold (double threshold)
 The moment arm threshold value that determines whether or not a path depends on a model coordinate. More...
 
double getMomentArmThreshold () const
 The moment arm threshold value that determines whether or not a path depends on a model coordinate. More...
 
void setMinimumPolynomialOrder (int order)
 The minimum order of the polynomial used to fit each path. More...
 
int getMinimumPolynomialOrder () const
 The minimum order of the polynomial used to fit each path. More...
 
void setMaximumPolynomialOrder (int order)
 The maximum order of the polynomial used to fit each path. More...
 
int getMaximumPolynomialOrder () const
 The maximum order of the polynomial used to fit each path. More...
 
void setGlobalCoordinateSamplingBounds (SimTK::Vec2 bounds)
 The global bounds that determine the minimum and maximum coordinate value samples at each time point. More...
 
SimTK::Vec2 getGlobalCoordinateSamplingBounds () const
 The global bounds that determine the minimum and maximum coordinate value samples at each time point. More...
 
void appendCoordinateSamplingBounds (const std::string &coordinatePath, const SimTK::Vec2 &bounds)
 The bounds (in degrees) that determine the minimum and maximum coordinate value samples at each time point for the coordinate at coordinatePath. More...
 
void setMomentArmTolerance (double tolerance)
 The tolerance on the root-mean-square (RMS) error (in meters) between the moment arms computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path. More...
 
double getMomentArmTolerance () const
 The tolerance on the root-mean-square (RMS) error (in meters) between the moment arms computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path. More...
 
void setPathLengthTolerance (double tolerance)
 The tolerance on the root-mean-square (RMS) error (in meters) between the path lengths computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path. More...
 
double getPathLengthTolerance () const
 The tolerance on the root-mean-square (RMS) error (in meters) between the path lengths computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path. More...
 
void setNumSamplesPerFrame (int numSamples)
 The number of samples taken per time frame in the coordinate values table used to fit each path. More...
 
int getNumSamplesPerFrame () const
 The number of samples taken per time frame in the coordinate values table used to fit each path. More...
 
void setNumParallelThreads (int numThreads)
 The number of threads used to parallelize the path fitting process. More...
 
int getNumParallelThreads () const
 The number of threads used to parallelize the path fitting process. More...
 
void setLatinHypercubeAlgorithm (std::string algorithm)
 The Latin hypercube sampling algorithm used to sample coordinate values for path fitting. More...
 
std::string getLatinHypercubeAlgorithm () const
 The Latin hypercube sampling algorithm used to sample coordinate values for path fitting. More...
 
- Public Member Functions inherited from OpenSim::Object
virtual ~Object ()
 Virtual destructor for cleanup. More...
 
bool isEqualTo (const Object &aObject) const
 Equality operator wrapper for use from languages not supporting operator overloading. More...
 
Objectoperator= (const Object &aObject)
 Copy assignment copies he base class fields, including the properties. More...
 
virtual bool operator== (const Object &aObject) const
 Determine if two objects are equal. More...
 
virtual bool operator< (const Object &aObject) const
 Provide an ordering for objects so they can be put in sorted containers. More...
 
void setName (const std::string &name)
 Set the name of the Object. More...
 
const std::string & getName () const
 Get the name of this Object. More...
 
void setDescription (const std::string &description)
 Set description, a one-liner summary. More...
 
const std::string & getDescription () const
 Get description, a one-liner summary. More...
 
const std::string & getAuthors () const
 Get Authors of this Object. More...
 
void setAuthors (const std::string &authors)
 Set Authors of this object. More...
 
const std::string & getReferences () const
 Get references or publications to cite if using this object. More...
 
void setReferences (const std::string &references)
 Set references or publications to cite if using this object. More...
 
int getNumProperties () const
 Determine how many properties are stored with this Object. More...
 
const AbstractPropertygetPropertyByIndex (int propertyIndex) const
 Get a const reference to a property by its index number, returned as an AbstractProperty. More...
 
AbstractPropertyupdPropertyByIndex (int propertyIndex)
 Get a writable reference to a property by its index number, returned as an AbstractProperty. More...
 
bool hasProperty (const std::string &name) const
 Return true if this Object has a property of any type with the given name, which must not be empty. More...
 
const AbstractPropertygetPropertyByName (const std::string &name) const
 Get a const reference to a property by its name, returned as an AbstractProperty. More...
 
AbstractPropertyupdPropertyByName (const std::string &name)
 Get a writable reference to a property by its name, returned as an AbstractProperty. More...
 
template<class T >
bool hasProperty () const
 Return true if this Object contains an unnamed, one-object property that contains objects of the given template type T. More...
 
template<class T >
const Property< T > & getProperty (const PropertyIndex &index) const
 Get property of known type Property<T> as a const reference; the property must be present and have the right type. More...
 
template<class T >
const Property< T > & getPropertyByName (const std::string &name) const
 Get property of known type Property<T> as a const reference; the property must be present and have the right type. More...
 
template<class T >
Property< T > & updProperty (const PropertyIndex &index)
 Get property of known type Property<T> as a writable reference; the property must be present and have the right type. More...
 
template<class T >
Property< T > & updPropertyByName (const std::string &name)
 Get property of known type Property<T> as a writable reference; the property must be present and have the right type.
More...
 
bool isObjectUpToDateWithProperties () const
 Returns true if no property's value has changed since the last time setObjectIsUpToDateWithProperties() was called. More...
 
void readObjectFromXMLNodeOrFile (SimTK::Xml::Element &objectElement, int versionNumber)
 We're given an XML element from which we are to populate this Object. More...
 
virtual void updateFromXMLNode (SimTK::Xml::Element &objectElement, int versionNumber)
 Use this method to deserialize an object from a SimTK::Xml::Element. More...
 
void updateXMLNode (SimTK::Xml::Element &parent, const AbstractProperty *prop=nullptr) const
 Serialize this object into the XML node that represents it. More...
 
bool getInlined () const
 Inlined means an in-memory Object that is not associated with an XMLDocument. More...
 
void setInlined (bool aInlined, const std::string &aFileName="")
 Mark this as inlined or not and optionally provide a file name to associate with the new XMLDocument for the non-inline case. More...
 
std::string getDocumentFileName () const
 If there is a document associated with this object then return the file name maintained by the document. More...
 
int getDocumentFileVersion () const
 If there is a document associated with this object then return its version number. More...
 
void setAllPropertiesUseDefault (bool aUseDefault)
 
bool print (const std::string &fileName) const
 Write this Object into an XML file of the given name; conventionally the suffix to use is ".osim". More...
 
std::string dump () const
 dump the XML representation of this Object into an std::string and return it. More...
 
virtual bool isA (const char *type) const
 The default implementation returns true only if the supplied string is "Object"; each Object-derived class overrides this to match its own class name. More...
 
const std::string & toString () const
 Wrapper to be used on Java side to display objects in tree; this returns just the object's name. More...
 
PropertySetgetPropertySet ()
 OBSOLETE: Get a reference to the PropertySet maintained by the Object. More...
 
const PropertySetgetPropertySet () const
 

Public Types

typedef PolynomialPathFitter Self
 This typedef might be useful within the member functions of this class. More...
 
typedef Object Super
 Use this typedef to refer to the superclass of this class. More...
 

Static Public Member Functions

static void evaluateFunctionBasedPaths (Model model, TableProcessor trajectory, const std::string &functionBasedPathsFile, double pathLengthTolerance=1e-4, double momentArmTolerance=1e-4)
 Print out a summary of the path fitting results, including information about the fitted polynomial functions and root-mean-square (RMS) errors between the original and fitted paths. More...
 
- Static Public Member Functions inherited from OpenSim::Object
static void registerType (const Object &defaultObject)
 Register an instance of a class; if the class is already registered it will be replaced. More...
 
static void renameType (const std::string &oldTypeName, const std::string &newTypeName)
 Support versioning by associating the current Object type with an old name. More...
 
static const ObjectgetDefaultInstanceOfType (const std::string &concreteClassName)
 Return a pointer to the default instance of the registered (concrete) Object whose class name is given, or NULL if the type is not registered. More...
 
template<class T >
static bool isObjectTypeDerivedFrom (const std::string &concreteClassName)
 Return true if the given concrete object type represents a subclass of the template object type T, and thus could be referenced with a T*. More...
 
static ObjectnewInstanceOfType (const std::string &concreteClassName)
 Create a new instance of the concrete Object type whose class name is given as concreteClassName. More...
 
static void getRegisteredTypenames (Array< std::string > &typeNames)
 Retrieve all the typenames registered so far. More...
 
template<class T >
static void getRegisteredObjectsOfGivenType (ArrayPtrs< T > &rArray)
 Return an array of pointers to the default instances of all registered (concrete) Object types that derive from a given Object-derived type that does not have to be concrete. More...
 
static bool PrintPropertyInfo (std::ostream &os, const std::string &classNameDotPropertyName, bool printFlagInfo=true)
 Dump formatted property information to a given output stream, useful for creating a "help" facility for registered objects. More...
 
static bool PrintPropertyInfo (std::ostream &os, const std::string &className, const std::string &propertyName, bool printFlagInfo=true)
 Same as the other signature but the class name and property name are provided as two separate strings. More...
 
static ObjectmakeObjectFromFile (const std::string &fileName)
 Create an OpenSim object whose type is based on the tag at the root node of the XML file passed in. More...
 
static const std::string & getClassName ()
 Return the name of this class as a string; i.e., "Object". More...
 
static void setSerializeAllDefaults (bool shouldSerializeDefaults)
 Static function to control whether all registered objects and their properties are written to the defaults section of output files rather than only those values for which the default was explicitly overwritten when read in from an input file or set programmatically. More...
 
static bool getSerializeAllDefaults ()
 Report the value of the "serialize all defaults" flag. More...
 
static bool isKindOf (const char *type)
 Returns true if the passed-in string is "Object"; each Object-derived class defines a method of this name for its own class name. More...
 
static void setDebugLevel (int newLevel)
 Set the amount of logging output. More...
 
static int getDebugLevel ()
 Get the current setting of debug level. More...
 
static ObjectSafeCopy (const Object *aObject)
 Use the clone() method to duplicate the given object unless the pointer is null in which case null is returned. More...
 
static void RegisterType (const Object &defaultObject)
 OBSOLETE alternate name for registerType(). More...
 
static void RenameType (const std::string &oldName, const std::string &newName)
 OBSOLETE alternate name for renameType(). More...
 

Auto-generated functions

static PolynomialPathFittersafeDownCast (OpenSim::Object *obj)
 For use in MATLAB and Python to access the concrete class. More...
 
static const std::string & getClassName ()
 This returns "PolynomialPathFitter". More...
 
void assign (Object &aObject) override
 This allows copy assignment in the Java GUI. More...
 
PolynomialPathFitterclone () const override
 Create a new heap-allocated copy of the concrete object to which this Object refers. More...
 
const std::string & getConcreteClassName () const override
 Returns the class name of the concrete Object-derived class of the actual object referenced by this Object, as a string. More...
 

Additional Inherited Members

- Static Public Attributes inherited from OpenSim::Object
static const std::string DEFAULT_NAME
 Name used for default objects when they are serialized. More...
 
- Protected Member Functions inherited from OpenSim::Object
 Object ()
 The default constructor is only for use by constructors of derived types. More...
 
 Object (const std::string &fileName, bool aUpdateFromXMLNode=true)
 Constructor from a file, to be called from other constructors that take a file as input. More...
 
 Object (const Object &source)
 Copy constructor is invoked automatically by derived classes with default copy constructors; otherwise it must be invoked explicitly. More...
 
 Object (SimTK::Xml::Element &aElement)
 Construct the base class portion of an Object from a given Xml element that describes this Object. More...
 
template<class T >
PropertyIndex addProperty (const std::string &name, const std::string &comment, const T &value)
 Define a new single-value property of known type T, with the given name, associated comment, and initial value. More...
 
template<class T >
PropertyIndex addOptionalProperty (const std::string &name, const std::string &comment)
 Add an optional property, meaning it can contain either no value or a single value. More...
 
template<class T >
PropertyIndex addOptionalProperty (const std::string &name, const std::string &comment, const T &value)
 Add an optional property, meaning it can contain either no value or a single value. More...
 
template<class T >
PropertyIndex addListProperty (const std::string &name, const std::string &comment, int minSize, int maxSize)
 Define a new list-valued property of known type T, with the given name, associated comment, minimum (==0) and maximum (>0) allowable list lengths, and a zero-length initial value. More...
 
template<class T , template< class > class Container>
PropertyIndex addListProperty (const std::string &name, const std::string &comment, int minSize, int maxSize, const Container< T > &valueList)
 Define a new list-valued property as above, but assigning an initial value via some templatized container class that supports size() and indexing. More...
 
PropertyIndex getPropertyIndex (const std::string &name) const
 Look up a property by name and return its PropertyIndex if it is found. More...
 
template<class T >
PropertyIndex getPropertyIndex () const
 Look up an unnamed property by the type of object it contains, and return its PropertyIndex if it is found. More...
 
template<class T >
void checkPropertyValueIsPositive (const Property< T > &p) const
 Throw an exception if any of the property's values are not positive. More...
 
template<class T >
void checkPropertyValueIsInSet (const Property< T > &p, const std::set< T > &set) const
 Throw an exception if any of the property's values are not in the provided set. More...
 
template<class T >
void checkPropertyValueIsInRangeOrSet (const Property< T > &p, const T &lower, const T &upper, const std::set< T > &set) const
 Throw an exception if any of the property's values are neither in the provided range nor in the provided set. More...
 
void setObjectIsUpToDateWithProperties ()
 When an object is initialized using the current values of its properties, it can set a flag indicating that it is up to date. More...
 
void clearObjectIsUpToDateWithProperties ()
 For testing or debugging purposes, manually clear the "object is up to date with respect to properties" flag. More...
 
void makeObjectNamesConsistentWithProperties ()
 Make sure the name of an object is consistent with its property type. More...
 
void updateFromXMLDocument ()
 Use this method only if you're deserializing from a file and the object is at the top level; that is, primarily in constructors that take a file name as input. More...
 
void setDocument (XMLDocument *doc)
 Unconditionally set the XMLDocument associated with this object. More...
 
const XMLDocumentgetDocument () const
 Get a const pointer to the document (if any) associated with this object. More...
 
XMLDocumentupdDocument ()
 Get a writable pointer to the document (if any) associated with this object. More...
 
- Protected Attributes inherited from OpenSim::Object
PropertySet _propertySet
 OBSOLETE: Property_Deprecated set for serializable member variables of this and derived classes. More...
 

Detailed Description

A utility class for fitting a set of FunctionBasedPaths to existing geometry-path in an OpenSim model using MultivariatePolynomialFunctions.

The primary inputs to this class include a model containing path objects derived from AbstractGeometryPath (e.g., GeometryPath) and a reference trajectory containing coordinate values for all Coordinates in the model. The path fitting process samples coordinate values around the reference trajectory, computes path lengths and moment arms from the geometry-based paths in the model, and fits polynomial coefficients for MultivariatePolynomialFunction objects based on the path length and moment arm samples. The fitted paths are written to an XML file, along with the modified coordinate values, sampled coordinate values, path lengths, and moment arms for both the original and fitted paths.

Note
Each file name is prefixed with the name of the model, and the directory to which the results are written can be specified using the setOutputDirectory method.

Settings

Various settings can be adjusted to control the path fitting process. The setMomentArmsThreshold method determines whether or not a path depends on a model coordinate. In other words, the absolute value the moment arm of a with respect to a particular coordinate must be greater than this value to be included during path fitting. The setMinimumPolynomialOrder and setMaximumPolynomialOrder methods specify the minimum and maximum order of the polynomial used to fit each path. The setGlobalCoordinateSamplingBounds property specifies the global bounds (in degrees) that determine the minimum and maximum coordinate values sampled at each time point. The method appendCoordinateSamplingBounds can be used to override the global bounds for a specific coordinate. The setMomentArmTolerance and setPathLengthTolerance methods specify the tolerance on the root-mean-square (RMS) error (in meters) between the moment arms and path lengths computed from the original model paths and the fitted polynomial paths. The setNumSamplesPerFrame method specifies the number of samples taken per time frame in the coordinate values table used to fit each path. The setNumParallelThreads method specifies the number of threads used to parallelize the path fitting process. The setLatinHypercubeAlgorithm method specifies the Latin hypercube sampling algorithm used to sample coordinate values for path fitting.

The default settings are as follows:

  • Moment arm threshold: 1e-3 meters
  • Minimum polynomial order: 2
  • Maximum polynomial order: 6
  • Global coordinate sampling bounds: [-10, 10] degrees
  • Moment arm tolerance: 1e-4 meters
  • Path length tolerance: 1e-4 meters
  • Number of samples per frame: 25
  • Number of threads: (# of available hardware threads) - 2
  • Latin hypercube sampling algorithm: "random"
Note
The default settings were chosen based on testing with a human lower-extremity model. Different settings may be required for other models with larger or smaller anatomical measures (e.g., dinosaur models).

Usage

The most basic usage of PolynomialPathFitter requires the user to provide a model and reference trajectory. The model should contain at least one path object derived from AbstractGeometryPath and should not contain any FunctionBasedPath objects. The reference trajectory must contain coordinate values for all Coordinates in the model:

fitter.setModel(ModelProcessor("model.osim"));
fitter.setCoordinateValues(TableProcessor("values.sto"));

The additional settings can be adjusted using the various set methods described above. For example, the global coordinate sampling bounds, bounds for the coordinate at "/jointset/slider/position", and the number of samples per frame can be set as follows:

fitter.setGlobalCoordinateSamplingBounds(SimTK::Vec2(-20.0, 20.0));
fitter.appendCoordinateSamplingBounds(
"/jointset/slider/position", SimTK::Vec2(-5.0, 15.0));
fitter.setNumSamplesPerFrame(50);

The path fitting process can be run using the run() method:

fitter.run();

Recommendations

Information from each step of the path fitting process is logged to the console, provided that you have set the OpenSim::Logger to level "info" or greater. Warnings are printed if the number of samples is likely insufficient for the fitting process, or if the fit for a particular path did not meet the specified tolerances.

It is highly recommended to use the files printed to the output directory to evaluate the quality of the fitted paths (see setOutputDirectory() for more details). Depending on the quality of the original model, it may not be possible to achieve a good fit for all paths (e.g., due to kinks or other discontinuities in the path). Finally, the fitted paths should only be used in simulations for which the coordinate values represent the expected range of motion for the model. If you are unsure if a simulation you have created with the fitted paths is valid, you can use the evaluateFunctionBasedPaths static method to compare the fitted paths to the original model paths given a new kinematic trajectory.

Note
The evaluateFunctionBasedPaths method can be used independently from the rest of this class, and does not require the FunctionBasedPaths in the model to use MultivariatePolynomialFunctions.

Member Typedef Documentation

◆ Self

This typedef might be useful within the member functions of this class.

This is generated by the OpenSim_DECLARE_*_OBJECT macros.

◆ Super

Use this typedef to refer to the superclass of this class.

Avoid using the explicit type name of the superclass; this would introduce bugs if the superclass is changed.

This is generated by the OpenSim_DECLARE_*_OBJECT macros.

Constructor & Destructor Documentation

◆ PolynomialPathFitter() [1/3]

OpenSim::PolynomialPathFitter::PolynomialPathFitter ( )

◆ ~PolynomialPathFitter()

OpenSim::PolynomialPathFitter::~PolynomialPathFitter ( )
overridenoexcept

◆ PolynomialPathFitter() [2/3]

OpenSim::PolynomialPathFitter::PolynomialPathFitter ( const PolynomialPathFitter )

◆ PolynomialPathFitter() [3/3]

OpenSim::PolynomialPathFitter::PolynomialPathFitter ( PolynomialPathFitter &&  )

Member Function Documentation

◆ appendCoordinateSamplingBounds()

void OpenSim::PolynomialPathFitter::appendCoordinateSamplingBounds ( const std::string &  coordinatePath,
const SimTK::Vec2 &  bounds 
)

The bounds (in degrees) that determine the minimum and maximum coordinate value samples at each time point for the coordinate at coordinatePath.

The bounds are specified as a SimTK::Vec2, where the first element is the minimum bound and the second element is the maximum bound. The maximum sample value at a particular time point is the nominal coordinate value plus the maximum bound, and the minimum sample value is the nominal coordinate value minus the minimum bound. This overrides the global bounds set by setGlobalCoordinateSamplingBounds() for this coordinate.

◆ assign()

void OpenSim::PolynomialPathFitter::assign ( Object aObject)
inlineoverride

This allows copy assignment in the Java GUI.

Exceptions
Exceptionif the argument is not of type PolynomialPathFitter.

◆ clone()

PolynomialPathFitter* OpenSim::PolynomialPathFitter::clone ( ) const
inlineoverridevirtual

Create a new heap-allocated copy of the concrete object to which this Object refers.

It is up to the caller to delete the returned object when no longer needed. Every concrete object deriving from Object implements this pure virtual method automatically, via the declaration macro it invokes (e.g., OpenSim_DECLARE_CONCRETE_OBJECT()). Note that the concrete class overrides modify the return type to be a pointer to the concrete object; that still overrides the base class method because the return type is covariant with (that is, derives from) Object.

Implements OpenSim::Object.

◆ evaluateFunctionBasedPaths()

static void OpenSim::PolynomialPathFitter::evaluateFunctionBasedPaths ( Model  model,
TableProcessor  trajectory,
const std::string &  functionBasedPathsFile,
double  pathLengthTolerance = 1e-4,
double  momentArmTolerance = 1e-4 
)
static

Print out a summary of the path fitting results, including information about the fitted polynomial functions and root-mean-square (RMS) errors between the original and fitted paths.

The trajectory argument is a TableProcessor object containing the simulation trajectory, specifically the set of coordinate values, used to compute path lengths and moment arms. The polynomialPathsFile argument is the path to an XML file containing the set of FunctionBasedPaths fitted to the geometry-based paths in model. These paths can be defined by MultivariatePolynomialFunctions generated by this class or any other Function objects that approximate the original model paths.

◆ getClassName()

static const std::string& OpenSim::PolynomialPathFitter::getClassName ( )
inlinestatic

This returns "PolynomialPathFitter".

See getConcreteClassName() if you want the class name of the underlying concrete object instead.

This is generated by the OpenSim_DECLARE_*_OBJECT macros.

◆ getConcreteClassName()

const std::string& OpenSim::PolynomialPathFitter::getConcreteClassName ( ) const
inlineoverridevirtual

Returns the class name of the concrete Object-derived class of the actual object referenced by this Object, as a string.

This is the string that is used as the tag for this concrete object in an XML file. Every concrete class derived from Object automatically overrides this method via the declaration macro it uses. See getClassName() to get the class name of the referencing (possibly abstract) class rather than the concrete object.

See also
getClassName()

Implements OpenSim::Object.

◆ getGlobalCoordinateSamplingBounds()

SimTK::Vec2 OpenSim::PolynomialPathFitter::getGlobalCoordinateSamplingBounds ( ) const

The global bounds that determine the minimum and maximum coordinate value samples at each time point.

The bounds are specified as a SimTK::Vec2, where the first element is the minimum bound and the second element is the maximum bound. Rotational coordinates are in degrees; translational coordinates in meters. The maximum sample value at a particular time point is the nominal coordinate value plus the maximum bound, and the minimum sample value is the nominal coordinate value minus the minimum bound.

Note
The default global bounds are set to [-10, 10] degrees/meters. If you have a model with paths that cross translational joints, you may to specify smaller bounds for the translational coordinates (see appendCoordinateSamplingBounds()).
To override the default global bounds for a specific coordinate, use the appendCoordinateSamplingBounds() method.

◆ getLatinHypercubeAlgorithm()

std::string OpenSim::PolynomialPathFitter::getLatinHypercubeAlgorithm ( ) const

The Latin hypercube sampling algorithm used to sample coordinate values for path fitting.

The Latin hypercube sampling algorithm is used to sample coordinate values for path fitting. The algorithm can be set to either "random" or "ESEA", which stands for the enhanced stochastic evolutionary algorithm developed by Jin et al. 2005 (see class LatinHypercubeDesign for more details). The "random" algorithm is used by default, and "ESEA" can be used to improve the quality of the sampling at the expense of higher computational cost. For most applications, the "random" algorithm is likely sufficient.

◆ getMaximumPolynomialOrder()

int OpenSim::PolynomialPathFitter::getMaximumPolynomialOrder ( ) const

The maximum order of the polynomial used to fit each path.

The order of a polynomial is the highest power of the independent variable(s) in the polynomial.

Note
The default maximum polynomial order is set to 6.

◆ getMinimumPolynomialOrder()

int OpenSim::PolynomialPathFitter::getMinimumPolynomialOrder ( ) const

The minimum order of the polynomial used to fit each path.

The order of a polynomial is the highest power of the independent variable(s) in the polynomial.

Note
The default minimum polynomial order is set to 2.

◆ getMomentArmThreshold()

double OpenSim::PolynomialPathFitter::getMomentArmThreshold ( ) const

The moment arm threshold value that determines whether or not a path depends on a model coordinate.

In other words, the moment arm of a path with respect to a particular coordinate must be greater than this value to be included during path fitting.

Note
The default moment arm threshold is set to 1e-3 meters.

◆ getMomentArmTolerance()

double OpenSim::PolynomialPathFitter::getMomentArmTolerance ( ) const

The tolerance on the root-mean-square (RMS) error (in meters) between the moment arms computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path.

The moment arm RMS error must be less than the tolerance for the polynomial order to be accepted. If the RMS error is greater than the tolerance, the polynomial order is increased by one and the path is refitted. This process is repeated until the RMS error is less than the tolerance or the maximum polynomial order is reached.

Note
The default moment arm tolerance is set to 1e-4 meters.
The path length RMS error must also be less than the path length tolerance for the polynomial order to be accepted (see setPathLengthTolerance).

◆ getNumParallelThreads()

int OpenSim::PolynomialPathFitter::getNumParallelThreads ( ) const

The number of threads used to parallelize the path fitting process.

This setting is used to divide the coordinate sampling, path length and moment arm computations, and path fitting across multiple threads. The number of threads must be greater than zero.

Note
The default number of threads is set to two fewer than the number of available hardware threads.

◆ getNumSamplesPerFrame()

int OpenSim::PolynomialPathFitter::getNumSamplesPerFrame ( ) const

The number of samples taken per time frame in the coordinate values table used to fit each path.

Note
The default number of samples per frame is set to 25.

◆ getOutputDirectory()

std::string OpenSim::PolynomialPathFitter::getOutputDirectory ( ) const

◆ getPathLengthTolerance()

double OpenSim::PolynomialPathFitter::getPathLengthTolerance ( ) const

The tolerance on the root-mean-square (RMS) error (in meters) between the path lengths computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path.

The path length RMS error must be less than the tolerance for the polynomial order to be accepted. If the RMS error is greater than the tolerance, the polynomial order is increased by one and the path is refitted. This process is repeated until the RMS error is less than the tolerance or the maximum polynomial order is reached.

Note
The default path length tolerance is set to 1e-4 meters.
The moment arm RMS error must also be less than the moment arm tolerance for the polynomial order to be accepted (see setMomentArmTolerance).

◆ operator=() [1/2]

PolynomialPathFitter& OpenSim::PolynomialPathFitter::operator= ( const PolynomialPathFitter )

◆ operator=() [2/2]

PolynomialPathFitter& OpenSim::PolynomialPathFitter::operator= ( PolynomialPathFitter &&  )

◆ run()

void OpenSim::PolynomialPathFitter::run ( )

Run the path fitting process.

The path fitting process consists of the following steps:

1. Load the model and reference coordinate values trajectory. The
   coordinate values table is modified to update the column labels
   based on the model coordinate paths, to update any coordinates
   dependent on `CoordinateCouplerConstraint`s, and to convert the
   coordinate values to radians if the "inDegrees" metadata flag is
   set to "yes".

2. Set sampling bounds for coordinates based on the global and
   coordinate-specific bounds properties.

3. Verify that the remaining user settings are valid.

4. Sample coordinate values around the reference trajectory using
   Latin hypercube sampling. The sampling is defined based on the
   coordinate bounds and range maps, the number of samples per frame,
   and the Latin hypercube sampling algorithm.

5. Compute path lengths and moment arms from the geometry-based paths
   in the input model.

6. Filter out bad coordinate samples and populate a map containing
   the coordinates that path is dependent on.

7. Fit the polynomial coefficients by finding a least-squares fit
   between the path lengths and moment arms computed from the
   geometry-based paths and the path lengths and moment arms
   computed from the fitted polynomial-based paths.

8. Print out a summary of the path fitting results, including
   information about the fitted polynomial functions and
   root-mean-square (RMS) errors between the original and fitted
   paths.

9. Write the fitted paths, modified coordinate values, sampled
   coordinate values, path lengths, and moment arms to files.
Note
Steps 4, 5, and 7 are parallelized using the number of threads specified via the setParallel() method.

◆ safeDownCast()

static PolynomialPathFitter* OpenSim::PolynomialPathFitter::safeDownCast ( OpenSim::Object obj)
inlinestatic

For use in MATLAB and Python to access the concrete class.

Example: cObj = PolynomialPathFitter.safeDownCast(obj). This is equivalent to dynamic_cast<PolynomialPathFitter*>(obj) in C++.

◆ setCoordinateValues()

void OpenSim::PolynomialPathFitter::setCoordinateValues ( TableProcessor  coordinateValues)

The reference trajectory used to sample coordinate values for path fitting.

The reference trajectory should be provided using a TableProcessor object. The reference trajectory must contain coordinate values for all Coordinates in the model. We assumed that the coordinate values meet all the kinematic constraints in the model, except for CoordinateCouplerConstraints, since we automatically update the coordinate trajectory to satisfy these constraints. The TimeSeriesTable must contain the "inDegrees" metadata flag; the coordinate values are automatically converted to radians if this flag is set to "yes".

◆ setGlobalCoordinateSamplingBounds()

void OpenSim::PolynomialPathFitter::setGlobalCoordinateSamplingBounds ( SimTK::Vec2  bounds)

The global bounds that determine the minimum and maximum coordinate value samples at each time point.

The bounds are specified as a SimTK::Vec2, where the first element is the minimum bound and the second element is the maximum bound. Rotational coordinates are in degrees; translational coordinates in meters. The maximum sample value at a particular time point is the nominal coordinate value plus the maximum bound, and the minimum sample value is the nominal coordinate value minus the minimum bound.

Note
The default global bounds are set to [-10, 10] degrees/meters. If you have a model with paths that cross translational joints, you may to specify smaller bounds for the translational coordinates (see appendCoordinateSamplingBounds()).
To override the default global bounds for a specific coordinate, use the appendCoordinateSamplingBounds() method.

◆ setLatinHypercubeAlgorithm()

void OpenSim::PolynomialPathFitter::setLatinHypercubeAlgorithm ( std::string  algorithm)

The Latin hypercube sampling algorithm used to sample coordinate values for path fitting.

The Latin hypercube sampling algorithm is used to sample coordinate values for path fitting. The algorithm can be set to either "random" or "ESEA", which stands for the enhanced stochastic evolutionary algorithm developed by Jin et al. 2005 (see class LatinHypercubeDesign for more details). The "random" algorithm is used by default, and "ESEA" can be used to improve the quality of the sampling at the expense of higher computational cost. For most applications, the "random" algorithm is likely sufficient.

◆ setMaximumPolynomialOrder()

void OpenSim::PolynomialPathFitter::setMaximumPolynomialOrder ( int  order)

The maximum order of the polynomial used to fit each path.

The order of a polynomial is the highest power of the independent variable(s) in the polynomial.

Note
The default maximum polynomial order is set to 6.

◆ setMinimumPolynomialOrder()

void OpenSim::PolynomialPathFitter::setMinimumPolynomialOrder ( int  order)

The minimum order of the polynomial used to fit each path.

The order of a polynomial is the highest power of the independent variable(s) in the polynomial.

Note
The default minimum polynomial order is set to 2.

◆ setModel()

void OpenSim::PolynomialPathFitter::setModel ( ModelProcessor  model)

The model containing geometry-based path objects to which polynomial-based path objects will be fitted.

The model should be provided using a ModelProcessor object. We expect the model to contain at least one path object derived from AbstractGeometryPath and does not already contain any FunctionBasedPath objects. The bounds for clamped coordinates are obeyed during the fitting process. Locked coordinates are unlocked if data is provided for them, or replaced with WeldJoints if no data is provided for them.

◆ setMomentArmThreshold()

void OpenSim::PolynomialPathFitter::setMomentArmThreshold ( double  threshold)

The moment arm threshold value that determines whether or not a path depends on a model coordinate.

In other words, the moment arm of a path with respect to a particular coordinate must be greater than this value to be included during path fitting.

Note
The default moment arm threshold is set to 1e-3 meters.

◆ setMomentArmTolerance()

void OpenSim::PolynomialPathFitter::setMomentArmTolerance ( double  tolerance)

The tolerance on the root-mean-square (RMS) error (in meters) between the moment arms computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path.

The moment arm RMS error must be less than the tolerance for the polynomial order to be accepted. If the RMS error is greater than the tolerance, the polynomial order is increased by one and the path is refitted. This process is repeated until the RMS error is less than the tolerance or the maximum polynomial order is reached.

Note
The default moment arm tolerance is set to 1e-4 meters.
The path length RMS error must also be less than the path length tolerance for the polynomial order to be accepted (see setPathLengthTolerance).

◆ setNumParallelThreads()

void OpenSim::PolynomialPathFitter::setNumParallelThreads ( int  numThreads)

The number of threads used to parallelize the path fitting process.

This setting is used to divide the coordinate sampling, path length and moment arm computations, and path fitting across multiple threads. The number of threads must be greater than zero.

Note
The default number of threads is set to two fewer than the number of available hardware threads.

◆ setNumSamplesPerFrame()

void OpenSim::PolynomialPathFitter::setNumSamplesPerFrame ( int  numSamples)

The number of samples taken per time frame in the coordinate values table used to fit each path.

Note
The default number of samples per frame is set to 25.

◆ setOutputDirectory()

void OpenSim::PolynomialPathFitter::setOutputDirectory ( std::string  directory)

The directory to which the path fitting results are written.

If the path fitting is successful, the fitted paths are written as a Set of FunctionBasedPath objects (with path length functions defined using MultivariatePolynomialFunction objects) to an XML file. Files containing the modified coordinate values, sampled coordinate values, path lengths, and moment arms for both the original and fitted paths are also written to the output directory.

Note
By default, results are written to the current working directory.

◆ setPathLengthTolerance()

void OpenSim::PolynomialPathFitter::setPathLengthTolerance ( double  tolerance)

The tolerance on the root-mean-square (RMS) error (in meters) between the path lengths computed from an original model path and a fitted polynomial-based path, which is used to determine the order of the polynomial used in the fitted path.

The path length RMS error must be less than the tolerance for the polynomial order to be accepted. If the RMS error is greater than the tolerance, the polynomial order is increased by one and the path is refitted. This process is repeated until the RMS error is less than the tolerance or the maximum polynomial order is reached.

Note
The default path length tolerance is set to 1e-4 meters.
The moment arm RMS error must also be less than the moment arm tolerance for the polynomial order to be accepted (see setMomentArmTolerance).

OpenSim Property, Socket, Output, Input Documentation

◆ coordinate_sampling_bounds

PolynomialPathFitterBounds OpenSim::PolynomialPathFitter::coordinate_sampling_bounds

"The bounds (in degrees) that determine the minimum and maximum " "coordinate value samples at each time point for specific " "coordinates. These bounds override the default coordinate " "sampling bounds."

This property appears in XML files under the tag <coordinate_sampling_bounds>. This property holds a list of objects, and was generated with the OpenSim_DECLARE_LIST_PROPERTY macro; see Property to learn about the property system.

See also
get_coordinate_sampling_bounds(), upd_coordinate_sampling_bounds(), set_coordinate_sampling_bounds(), append_coordinate_sampling_bounds()

◆ coordinate_values

TableProcessor OpenSim::PolynomialPathFitter::coordinate_values

"The reference trajectory used to sample coordinate values for " "path fitting."

This property appears in XML files under the tag <coordinate_values>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_coordinate_values(), upd_coordinate_values(), set_coordinate_values()

◆ global_coordinate_sampling_bounds

SimTK::Vec2 OpenSim::PolynomialPathFitter::global_coordinate_sampling_bounds

"The global bounds (in degrees) that determine the minimum and " "maximum coordinate value samples at each time point."

This property appears in XML files under the tag <global_coordinate_sampling_bounds>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_global_coordinate_sampling_bounds(), upd_global_coordinate_sampling_bounds(), set_global_coordinate_sampling_bounds()

◆ latin_hypercube_algorithm

std::string OpenSim::PolynomialPathFitter::latin_hypercube_algorithm

"The Latin hypercube sampling algorithm used to sample coordinate " "values for path fitting (default: 'random')."

This property appears in XML files under the tag <latin_hypercube_algorithm>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_latin_hypercube_algorithm(), upd_latin_hypercube_algorithm(), set_latin_hypercube_algorithm()

◆ maximum_polynomial_order

int OpenSim::PolynomialPathFitter::maximum_polynomial_order

"The maximum order of the polynomial used to fit each path. The " "order of a polynomial is the highest power of the independent " "variable(s) in the polynomial."

This property appears in XML files under the tag <maximum_polynomial_order>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_maximum_polynomial_order(), upd_maximum_polynomial_order(), set_maximum_polynomial_order()

◆ minimum_polynomial_order

int OpenSim::PolynomialPathFitter::minimum_polynomial_order

"The minimum order of the polynomial used to fit each path. The " "order of a polynomial is the highest power of the independent " "variable(s) in the polynomial."

This property appears in XML files under the tag <minimum_polynomial_order>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_minimum_polynomial_order(), upd_minimum_polynomial_order(), set_minimum_polynomial_order()

◆ model

ModelProcessor OpenSim::PolynomialPathFitter::model

"The model containing geometry-based path objects to which " "polynomial-based path objects will be fitted."

This property appears in XML files under the tag <model>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_model(), upd_model(), set_model()

◆ moment_arm_threshold

double OpenSim::PolynomialPathFitter::moment_arm_threshold

"The moment arm threshold value that determines whether or not a " "path depends on a model coordinate. In other words, the moment " "arm of a path with respect to a coordinate must be greater than " "this value to be included during path fitting."

This property appears in XML files under the tag <moment_arm_threshold>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_moment_arm_threshold(), upd_moment_arm_threshold(), set_moment_arm_threshold()

◆ moment_arm_tolerance

double OpenSim::PolynomialPathFitter::moment_arm_tolerance

"The tolerance on the root-mean-square (RMS) error (in meters) " "between the moment arms computed from an original model path and " "a fitted polynomial-based path, which is used to determine the " "order of the polynomial used in the fitted path (default: 1e-4)."

This property appears in XML files under the tag <moment_arm_tolerance>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_moment_arm_tolerance(), upd_moment_arm_tolerance(), set_moment_arm_tolerance()

◆ num_parallel_threads

int OpenSim::PolynomialPathFitter::num_parallel_threads

"The number of threads used to parallelize the path fitting " "process (default: two fewer than the number of available " "hardware threads)."

This property appears in XML files under the tag <num_parallel_threads>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_num_parallel_threads(), upd_num_parallel_threads(), set_num_parallel_threads()

◆ num_samples_per_frame

int OpenSim::PolynomialPathFitter::num_samples_per_frame

"The number of samples taken per time frame in the coordinate " "values table used to fit each path (default: 25)."

This property appears in XML files under the tag <num_samples_per_frame>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_num_samples_per_frame(), upd_num_samples_per_frame(), set_num_samples_per_frame()

◆ output_directory

std::string OpenSim::PolynomialPathFitter::output_directory

"The directory to which the path fitting results are written."

This property appears in XML files under the tag <output_directory>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_output_directory(), upd_output_directory(), set_output_directory()

◆ path_length_tolerance

double OpenSim::PolynomialPathFitter::path_length_tolerance

"The tolerance on the root-mean-square (RMS) error (in meters) " "between the path lengths computed from an original model path and " "a fitted polynomial-based path, which is used to determine the " "order of the polynomial used in the fitted path (default: 1e-4)."

This property appears in XML files under the tag <path_length_tolerance>. This property was generated with the OpenSim_DECLARE_PROPERTY macro; see Property to learn about the property system.

See also
get_path_length_tolerance(), upd_path_length_tolerance(), set_path_length_tolerance()

The documentation for this class was generated from the following file: