Main Page Modules Namespace List Class Hierarchy Alphabetical List Compound List File List Namespace Members Compound Members File Members Related Pages

OSCAR::PerformanceCriteria Class Reference
[PerformanceCriteria]

PerformanceCriteria should be used to select a solution from a set of options for a manipulator. More...

#include <PerformanceCriteria.h>

Inheritance diagram for OSCAR::PerformanceCriteria:

[legend]List of all members.


Public Types
enum	NormalizeScheme { LocalUnit, GlobalUnit, LocalAvg, GlobalAvg, None }
Public Methods
	PerformanceCriteria ()
	PerformanceCriteria (const PerformanceCriteria &crit)
PerformanceCriteria &	operator= (const PerformanceCriteria &crit)
void	SetDirection (MinOrMax minOrMax)
MinOrMax	GetDirection () const
void	SetCriticalLimit (double criticalLimit)
double	GetCriticalLimit () const
void	SetWeight (double weight)
double	GetWeight () const
void	SetExponent (double exponent)
double	GetExponent () const
void	SetNormalizeScheme (NormalizeScheme normalizeScheme=LocalUnit)
NormalizeScheme	GetNormalizeScheme (void) const
void	SetNormalizeValues (double min, double max, double avg)
void	GetNormalizeValues (double &min, double &max, double &avg)
virtual CriteriaPropertiesRepository const *	CalculateProperties (RepositoryList &repoList)
virtual bool	PickSolution (Vector &solution, RepositoryList &repoList)
unsigned int	GetNumOfPickedSolution (void)
virtual	~PerformanceCriteria ()
virtual bool	GetGradient (const Vector &joints, Vector &gradient)
Protected Methods
virtual CriteriaPropertiesRepository const *	calculateProperties (RepositoryList &repoList)
bool	globalUnitNormalize (RepositoryList &repoList)
bool	localUnitNormalize (RepositoryList &repoList)
bool	localAvgNormalize (RepositoryList &repoList)
bool	globalAvgNormalize (RepositoryList &repoList)
virtual bool	computeCriteria (RepositoryList &repoList)=0
Protected Attributes
double	weight
double	criticalLimit
double	exponent
double	normalizeMin
double	normalizeMax
double	normalizeAvg
MinOrMax	minOrMax
String	criteriaName
NormalizeScheme	normalizeScheme
CriteriaPropertiesRepository *	criteriaPropertiesPtr
unsigned int	solutionNum

Detailed Description

Author:: Mark Tisius and Chetan Kapoor

PerformanceCriteria is setup to be as general as possible to allow many different redundancy resolution techniques. It has a criticalLimit argument that can be used to mark a solution valid or invalid such as in a pass/fail scheme. Also using PCFilter a sequential filters scheme can be easily setup. Different fusion techniques can also be implemented by setting the weights and exponents in the individual criteria and then placing them in PCFusion. All criteria derive from this class so it contains the functionality to check validity of solutions, get criteria values for solutions, normalize criteria values, and pick the best solution based upon what decision-making scheme is set.

Todo:: PerformanceCriteria, when sent a set of joint solutions created by IKJGenerateOptions, currently calculates the properties of all joint solutions, regardless of the joint solutions validity. This is an efficiency issue. Also, it is possible for PickSolution to pick an invalid solution. This needs to be changed.

Member Enumeration Documentation

enum OSCAR::PerformanceCriteria::NormalizeScheme

Enumeration values:

LocalUnit

GlobalUnit

LocalAvg

GlobalAvg

None

Constructor & Destructor Documentation

OSCAR::PerformanceCriteria::PerformanceCriteria ( )

Constructor for PerformanceCriteria Use this constructor to create a PerformanceCriteria object. This constructor gives default values normalizeScheme (none), normalizeMin (1000000), normalizeMax (0), nomalizeAvg (1), weight (1), exponent (1), and criticalLimit (1). This is done to prevent problems in specific performance criteria that inherit from PerformanceCriteria.

OSCAR::PerformanceCriteria::PerformanceCriteria ( const PerformanceCriteria & crit )

Copy Constructor for PerformanceCriteria Use this to create a PerformanceCriteria object whose default are set via the constructor argument. Note: The pointer to the criteria properties repository is copied so both criteria will use the same repository.

Parameters:

crit A PerformanceCriteria object that is used to create the new object

virtual OSCAR::PerformanceCriteria::~PerformanceCriteria ( ) [virtual]

Destructor for PerformanceCriteria.

Member Function Documentation

virtual CriteriaPropertiesRepository const* OSCAR::PerformanceCriteria::calculateProperties ( RepositoryList & repoList ) [protected, virtual]

Reimplemented in OSCAR::PCFusion.

virtual CriteriaPropertiesRepository const* OSCAR::PerformanceCriteria::CalculateProperties ( RepositoryList & repoList ) [virtual]

Gets the criteria properties for the entire solution set. Use this method to calculate the criteria properties for the solution set. Note: The validity is determined by the critical limit and the direction. If a local normalization scheme is chosen only valid solutions will be used to normalize. Global normalization is based on the values set by the user. GlobalUnit normalization values are updated if a valid solution has a greater/lower min/max. The nature of GlobalAvg does not allow automatic updates.

Returns:
CriteriaPropertiesRepository* Pointer to a properties repository where data for the criteria is stored.

Parameters:

repoList RepositoryList object that contains the solutions, information calculated by the criteria criteria, and the criteria values.

Exceptions:

noSolutionPassed Error if all solutions are invalid and a local normalization scheme is set.

Reimplemented in OSCAR::PCFilter, and OSCAR::PCFusion.

virtual bool OSCAR::PerformanceCriteria::computeCriteria ( RepositoryList & repoList ) [protected, pure virtual]

Implemented in OSCAR::PCAverageDistReciprocal, OSCAR::PCEEWrench, OSCAR::PCEEWrenchD1, OSCAR::PCJointTorque, OSCAR::PCLinkForce, OSCAR::PCSmallestMinDist, OSCAR::PCAccelerationLimitsAvoidance, OSCAR::PCDexterity, OSCAR::PCDirectionalEEFStiffness, OSCAR::PCEEFKineticEnergy, OSCAR::PCFilter, OSCAR::PCForceTransmissibility, OSCAR::PCFusion, OSCAR::PCGeneralizedDynamicEfficiency, OSCAR::PCGeneralizedInertialEfficiency, OSCAR::PCGeneralizedStiffness, OSCAR::PCGeneralizedTorqueMinimization, OSCAR::PCGeneralizedVelocityMinimization, OSCAR::PCInertialFrobeniusNorm, OSCAR::PCInverseMeasureOfTransmissibility, OSCAR::PCJacobianFrobeniusNorm, OSCAR::PCJointRangeAvailability, OSCAR::PCKineticEnergy, OSCAR::PCLoadTorqueMinimization, OSCAR::PCMagneticEEF, OSCAR::PCMeasureOfTransmissibility, OSCAR::PCPotentialEnergy, OSCAR::PCSingularityAvoidance, OSCAR::PCTorqueLimitsAvoidance, OSCAR::PCVelocityLimitsAvoidance, and OSCAR::PCVelocityTransmissibility.

double OSCAR::PerformanceCriteria::GetCriticalLimit ( ) const

Gets the critical limit of the criteria. Use this method to get the critical limit of the PerformanceCriteria object. It will either be a maximum limit or a minimum limit depending on the optimization direction. It is also used to normalize the criteria value.

Returns:
double Value of the critical limit.

MinOrMax OSCAR::PerformanceCriteria::GetDirection ( ) const

Gets the direction which the criteria is being optimized. Use this method to get the optimization direction for a PerformanceCriteria object.

Returns:
MinOrMax The optimzation direction, either Minimize or Maximize

double OSCAR::PerformanceCriteria::GetExponent ( ) const

Gets the exponent of the criteria. Use this method to get the exponent of the PerformanceCriteria object. The exponent is used to keep criteria values low until the critical limit is being approached.

Returns:
double Exponent value

virtual bool OSCAR::PerformanceCriteria::GetGradient ( const Vector & joints,

Vector & gradient

) [virtual]

Get the gradient of the performance criterion.

Parameters:

joints The current configuration at which the gradients will be determined.

gradients A vector containing the gradient upon return.

Returns:
True if successful. False if the criterion does not have a gradient.

Reimplemented in OSCAR::PCFusion, and OSCAR::PCJointRangeAvailability.

NormalizeScheme OSCAR::PerformanceCriteria::GetNormalizeScheme ( void ) const

Get the normalization scheme for the criterion.

Returns:
NormalizeScheme Enum representing the type of normalization.

void OSCAR::PerformanceCriteria::GetNormalizeValues ( double & min,

double & max,

double & avg

)

Gets the values used for global normalization.

Returns:
void

Parameters:

min The minimum criteria value. Gets updated during motion.

max The maximum criteria value. Gets updated during motion.

avg The average criteria value.

unsigned int OSCAR::PerformanceCriteria::GetNumOfPickedSolution ( void )

double OSCAR::PerformanceCriteria::GetWeight ( ) const

Gets the weight of the criteria. Use this method to get the weight of the PerformanceCriteria object. The weight is used to rank importance when multiple criteria are used in a fusion object. Weight has no effect unless multiple criteria are utilized.

Returns:
double Value for the weight.

bool OSCAR::PerformanceCriteria::globalAvgNormalize ( RepositoryList & repoList ) [protected]

bool OSCAR::PerformanceCriteria::globalUnitNormalize ( RepositoryList & repoList ) [protected]

bool OSCAR::PerformanceCriteria::localAvgNormalize ( RepositoryList & repoList ) [protected]

bool OSCAR::PerformanceCriteria::localUnitNormalize ( RepositoryList & repoList ) [protected]

PerformanceCriteria& OSCAR::PerformanceCriteria::operator= ( const PerformanceCriteria & crit )

Copy operator for PerformanceCriteria Use this to copy one PerformanceCriteria object into another. Note: The pointer to the criteria properties repository is copied so both criteria will use the same repository.

Returns:
A reference to itself

Parameters:

crit A PerformanceCriteria object that is copied into the calling object.

virtual bool OSCAR::PerformanceCriteria::PickSolution ( Vector & solution,

RepositoryList & repoList

) [virtual]

Picks the optimal solution based upon the decision method set by the operator.

Returns:
bool True if successful, False if no solution passes critical limit. The Vector solution parameter will not get overwritten. Call GetError for error information.

Parameters:

repoList RepositoryList object that contains the solutions, intermediate calculations, and the calculated criteria values.

bestSolution Vector containing the joint angles in radians of the optimal solution for the given criterion and direction.

Exceptions:

noSolutionPassed All solutions violated the critical limit set for the criteria. Try reseting the critical limit.

Reimplemented in OSCAR::PCFilter, and OSCAR::PCFusion.

void OSCAR::PerformanceCriteria::SetCriticalLimit ( double criticalLimit )

Set the critical limit for the criteria. Use this method to set the critical limit for a PerformanceCriteria object. It will either be a maximum limit or a minimum limit depending on the optimization direction. It is also used to normalize the criteria value if a fusion scheme is chosen. Set to -1 if limit checking is not desired.

Returns:
void

Parameters:

criticalLimit Value for the critical limit which also is the normalizing factor.

void OSCAR::PerformanceCriteria::SetDirection ( MinOrMax minOrMax )

Set the direction which the criteria should be optimized. Use this method to set the direction for opitimization of a PerformanceCriteria object.

Returns:
void

Parameters:

minOrMax The directions to optimize. Choose either Minimize if you want the criteria values minimized and Maximize if you want the criteria values maximized. This depends on each individual criteria formulation

void OSCAR::PerformanceCriteria::SetExponent ( double exponent )

Set the exponent for the criteria. Use this method to set the exponent for a PerformanceCriteria object. The exponent will be used to keep criteria values low until the critical limit is being approached.

Returns:
void

Parameters:

exponent Exponent value.

void OSCAR::PerformanceCriteria::SetNormalizeScheme ( NormalizeScheme normalizeScheme = LocalUnit )

Set the normalization scheme for the criterion. If this method is not called no normalization will occur. Use this method to set the normalization type for a PerformanceCriteria object. The different types are: LocalUnit-Normalizes between 0 and 1 based on the range of the solution set. Only solutions which do not violate the critical limit are used in normalization. GlobalUnit-Normalizes between 0 and 1 based on the range set by the user - call SetNormalizeValues to set these. Note: New min and max are set if a criteria value is found to be larger/smaller as long as it is valid. LocalAverage-Scales by the average of the solution set. Only solutions which are valid are used to compute the local average. GlobalAverage-Scales by the average set by the user - call SetNormalizeValues to set this value.

Returns:
void

Parameters:

normalizeScheme Enum used to specify the type of normalization scheme. The default is GlobalUnit.

void OSCAR::PerformanceCriteria::SetNormalizeValues ( double min,

double max,

double avg

)

Set the values used for global normalization. Note: If GlobalUnit is set then avg will not be used, and for GlobalAvg min and max will not be used. Note: GlobalUnit will update these values during motion if different valid minimums or maximums are found.

Returns:
void

Parameters:

min The minimum criteria value.

max The maximum criteria value.

avg The average criteria value.

void OSCAR::PerformanceCriteria::SetWeight ( double weight )

Set the weight for the criteria. Use this method to set the weight for a PerformanceCriteria object. The weight will be used to rank importance when multiple criteria used in decision making. Weight has no effect unless multiple criteria are utilized.

Returns:
void

Parameters:

weight Value for the weight.