Analytical is a base class that should be inherited by methods that are made to solve the IK analytically instead of iteratively. More...
#include <InverseKinematics.hpp>
Classes | |
| struct | Properties |
| struct | Solution |
| struct | UniqueProperties |
Public Types | |
| enum | Validity_t { VALID = 0, OUT_OF_REACH = 1 << 0, LIMIT_VIOLATED = 1 << 1 } |
| Bitwise enumerations that are used to describe some properties of each solution produced by the analytical IK. More... | |
| enum | ExtraDofUtilization { UNUSED = 0, PRE_ANALYTICAL, POST_ANALYTICAL, PRE_AND_POST_ANALYTICAL } |
| If there are extra DOFs in the IK module which your Analytical solver implementation does not make use of, those DOFs can be used to supplement the analytical solver using Jacobian transpose iteration. More... | |
| typedef std::function< bool(const Eigen::VectorXd &_better, const Eigen::VectorXd &_worse, const InverseKinematics *_ik)> | QualityComparison |
Public Member Functions | |
| Analytical (InverseKinematics *_ik, const std::string &_methodName, const Properties &_properties) | |
| Constructor. | |
| virtual | ~Analytical ()=default |
| Virtual destructor. | |
| const std::vector< Solution > & | getSolutions () |
| Get the solutions for this IK module, along with a tag indicating whether each solution is valid. More... | |
| const std::vector< Solution > & | getSolutions (const Eigen::Isometry3d &_desiredTf) |
| Get the solutions for this IK module, along with a tag indicating whether each solution is valid. More... | |
| void | computeGradient (const Eigen::Vector6d &_error, Eigen::VectorXd &_grad) override |
| You should not need to override this function. More... | |
| virtual const std::vector< Solution > & | computeSolutions (const Eigen::Isometry3d &_desiredBodyTf)=0 |
| Use this function to fill the entries of the mSolutions variable. More... | |
| virtual const std::vector< std::size_t > & | getDofs () const =0 |
| Get a list of the DOFs that will be included in the entries of the solutions returned by getSolutions(). More... | |
| void | setPositions (const Eigen::VectorXd &_config) |
| Set the configuration of the DOFs. More... | |
| Eigen::VectorXd | getPositions () const |
| Get the configuration of the DOFs. More... | |
| void | setExtraDofUtilization (ExtraDofUtilization _utilization) |
| Set how you want extra DOFs to be utilized by the IK module. | |
| ExtraDofUtilization | getExtraDofUtilization () const |
| Get how extra DOFs are being utilized by the IK module. | |
| void | setExtraErrorLengthClamp (double _clamp) |
| Set how much to clamp the error vector that gets applied to extra DOFs. | |
| double | getExtraErrorLengthClamp () const |
| Get how much we will clamp the error vector that gets applied to extra DOFs. | |
| void | setQualityComparisonFunction (const QualityComparison &_func) |
| Set the function that will be used to compare the qualities of two solutions. More... | |
| void | resetQualityComparisonFunction () |
| Reset the quality comparison function to the default method. | |
| Properties | getAnalyticalProperties () const |
| Get the Properties for this Analytical class. | |
| void | constructDofMap () |
| Construct a mapping from the DOFs of getDofs() to their indices within the Node's list of dependent DOFs. More... | |
 Public Member Functions inherited from dart::dynamics::InverseKinematics::GradientMethod | |
| GradientMethod (InverseKinematics *_ik, const std::string &_methodName, const Properties &_properties) | |
| Constructor. | |
| virtual | ~GradientMethod ()=default |
| Virtual destructor. | |
| virtual std::unique_ptr< GradientMethod > | clone (InverseKinematics *_newIK) const =0 |
| Enable this GradientMethod to be cloned to a new IK module. | |
| void | evalGradient (const Eigen::VectorXd &_q, Eigen::Map< Eigen::VectorXd > _grad) |
| This function is used to handle caching the gradient vector and interfacing with the solver. More... | |
| const std::string & | getMethodName () const |
| Get the name of this GradientMethod. | |
| void | clampGradient (Eigen::VectorXd &_grad) const |
| Clamp the gradient based on the clamp settings of this GradientMethod. | |
| void | setComponentWiseClamp (double _clamp=DefaultIKGradientComponentClamp) |
| Set the component-wise clamp for this GradientMethod. More... | |
| double | getComponentWiseClamp () const |
| Get the component-wise clamp for this GradientMethod. | |
| void | applyWeights (Eigen::VectorXd &_grad) const |
| Apply weights to the gradient based on the weight settings of this GradientMethod. More... | |
| void | setComponentWeights (const Eigen::VectorXd &_weights) |
| Set the weights that will be applied to each component of the gradient. More... | |
| const Eigen::VectorXd & | getComponentWeights () const |
| Get the weights of this GradientMethod. | |
| void | convertJacobianMethodOutputToGradient (Eigen::VectorXd &grad, const std::vector< std::size_t > &dofs) |
| Convert the gradient that gets generated by Jacobian methods into a gradient that can be used by a GradientDescentSolver. More... | |
| Properties | getGradientMethodProperties () const |
| Get the Properties of this GradientMethod. | |
| void | clearCache () |
| Clear the cache to force the gradient to be recomputed. More... | |
| InverseKinematics * | getIK () |
| Returns the IK module that this GradientMethod belongs to. | |
| const InverseKinematics * | getIK () const |
| Returns the IK module that this GradientMethod belongs to. | |
| Public Member Functions inherited from dart::common::Subject | |
| virtual | ~Subject () |
| Destructor will notify all Observers that it is destructing. | |
Protected Member Functions | |
| virtual void | addExtraDofGradient (Eigen::VectorXd &grad, const Eigen::Vector6d &error, ExtraDofUtilization utilization) |
| This function will compute a gradient which utilizes the extra DOFs that go unused by the Analytical solution and then it will add the components of that gradient to the output parameter: grad. More... | |
| void | checkSolutionJointLimits () |
| Go through the mSolutions vector and tag entries with LIMIT_VIOLATED if any components of their configuration are outside of their position limits. More... | |
| Protected Member Functions inherited from dart::common::Subject | |
| void | sendDestructionNotification () const |
| Send a destruction notification to all Observers. More... | |
| void | addObserver (Observer *_observer) const |
| Add an Observer to the list of Observers. | |
| void | removeObserver (Observer *_observer) const |
| Remove an Observer from the list of Observers. | |
Protected Attributes | |
| std::vector< Solution > | mSolutions |
| Vector of solutions. | |
| UniqueProperties | mAnalyticalP |
| Properties for this Analytical IK solver. | |
| Protected Attributes inherited from dart::dynamics::InverseKinematics::GradientMethod | |
| common::sub_ptr< InverseKinematics > | mIK |
| The IK module that this GradientMethod belongs to. | |
| std::string | mMethodName |
| The name of this method. | |
| Eigen::VectorXd | mLastPositions |
| The last positions that was passed to this GradientMethod. | |
| Eigen::VectorXd | mLastGradient |
| The last gradient that was computed by this GradientMethod. | |
| Properties | mGradientP |
| Properties for this GradientMethod. | |
| Protected Attributes inherited from dart::common::Subject | |
| std::set< Observer * > | mObservers |
| List of current Observers. | |
Detailed Description
Analytical is a base class that should be inherited by methods that are made to solve the IK analytically instead of iteratively.
This provides an extended API that is relevant to Analytical solvers but not iterative solvers.
Creating an Analytical solver will have the side effect of removing the error clamp and error weights from your ErrorMethod. If you still want your error computations to be clamped and weighted, you should set it again after creating the Analytical solver. Clamping and weighting the error vector often helps iterative methods to converge smoothly, but it is counter-productive for analytical methods which do not typically rely on convergence; analytical methods can usually solve the entire error vector directly.
Member Enumeration Documentation
◆ ExtraDofUtilization
If there are extra DOFs in the IK module which your Analytical solver implementation does not make use of, those DOFs can be used to supplement the analytical solver using Jacobian transpose iteration.
This enumeration is used to indicate whether you want those DOFs to be used before applying the analytical solution, after applying the analytical solution, or not be used at all.
Jacobian transpose is used for the extra DOFs because it is inexpensive and robust to degenerate Jacobians which are common in low dimensional joint spaces. The primary advantage of pseudoinverse methods over Jacobian transpose methods is their precision, but analytical methods are even more precise than pseudoinverse methods, so that precision is not needed in this case.
If you want the extra DOFs to use a different method than Jacobian transpose, you can create two seperate IK modules (one which is analytical and one with the iterative method of your choice) and combine them in a HierarchicalIK.
◆ Validity_t
Bitwise enumerations that are used to describe some properties of each solution produced by the analytical IK.
Member Function Documentation
◆ addExtraDofGradient()
|
protectedvirtual |
This function will compute a gradient which utilizes the extra DOFs that go unused by the Analytical solution and then it will add the components of that gradient to the output parameter: grad.
You can override this function to customize how the extra DOFs are used. The default behavior is to use a simple Jacobian Transpose method.
The utilization flag will be PRE_ANALYTICAL if the function is being called before the Analytical solution is computed; it will be POST_ANALYTICAL if the function is being called after the Analytical solution is computed.
◆ checkSolutionJointLimits()
|
protected |
Go through the mSolutions vector and tag entries with LIMIT_VIOLATED if any components of their configuration are outside of their position limits.
This will NOT clear the LIMIT_VIOLATED flag from entries of mSolutions which are already tagged with it, even if they do not violate any limits.
◆ computeGradient()
|
overridevirtual |
You should not need to override this function.
Instead, you should override computeSolutions.
Implements dart::dynamics::InverseKinematics::GradientMethod.
◆ computeSolutions()
|
pure virtual |
Use this function to fill the entries of the mSolutions variable.
Be sure to clear the mSolutions vector at the start, and to also return the mSolutions vector at the end. Note that you are not expected to evaluate any of the solutions for their quality. However, you should set the Solution::mValidity flag to OUT_OF_REACH for each solution that does not actually reach the desired transform, and you should call checkSolutionJointLimits() and the end of the function, which will set the LIMIT_VIOLATED flags of any configurations that are outside of the position limits.
Implemented in HuboLegIK, HuboArmIK, and dart::dynamics::IkFast.
◆ constructDofMap()
| void dart::dynamics::InverseKinematics::Analytical::constructDofMap | ( | ) |
Construct a mapping from the DOFs of getDofs() to their indices within the Node's list of dependent DOFs.
This will be called immediately after the Analytical is constructed; this one call is sufficient as long as the DOFs of Analytical::getDofs() is not changed. However, if your Analytical is able to change the DOFs that it operates on, then you will need to call this function each time the DOFs have changed.
◆ getDofs()
|
pure virtual |
Get a list of the DOFs that will be included in the entries of the solutions returned by getSolutions().
Ideally, this should match up with the DOFs being used by the InverseKinematics module, but this might not always be possible, so this function ensures that solutions can be interpreted correctly.
Implemented in HuboLegIK, HuboArmIK, and dart::dynamics::IkFast.
◆ getPositions()
| Eigen::VectorXd dart::dynamics::InverseKinematics::Analytical::getPositions | ( | ) | const |
Get the configuration of the DOFs.
The components of this vector will correspond to the DOFs provided by getDofs().
◆ getSolutions() [1/2]
| const std::vector< IK::Analytical::Solution > & dart::dynamics::IK::Analytical::getSolutions | ( | ) |
Get the solutions for this IK module, along with a tag indicating whether each solution is valid.
This function will assume that you want to use the desired transform given by the IK module's current ErrorMethod.
◆ getSolutions() [2/2]
| const std::vector< IK::Analytical::Solution > & dart::dynamics::IK::Analytical::getSolutions | ( | const Eigen::Isometry3d & | _desiredTf | ) |
Get the solutions for this IK module, along with a tag indicating whether each solution is valid.
This function will compute the configurations using the given desired transform instead of using the IK module's current ErrorMethod.
◆ setPositions()
| void dart::dynamics::InverseKinematics::Analytical::setPositions | ( | const Eigen::VectorXd & | _config | ) |
Set the configuration of the DOFs.
The components of _config must correspond to the DOFs provided by getDofs().
◆ setQualityComparisonFunction()
| void dart::dynamics::InverseKinematics::Analytical::setQualityComparisonFunction | ( | const QualityComparison & | _func | ) |
Set the function that will be used to compare the qualities of two solutions.
This function should return true if the first argument is a better solution than the second argument.
By default, it will prefer the solution which has the smallest size for its largest change in joint angle. In other words, for each configuration that it is given, it will compare the largest change in joint angle for each configuration and pick the one that is smallest.
Note that outside of this comparison function, the Solutions will be split between which are valid, which are out-of-reach, and which are in violation of joint limits. Valid solutions will always be ranked above invalid solutions, and joint limit violations will always be ranked last.
The documentation for this class was generated from the following files:
- dart/dynamics/InverseKinematics.hpp
- dart/dynamics/InverseKinematics.cpp
