Visual Servoing Platform  version 3.6.1 under development (2024-12-17)
Public Member Functions | Static Public Attributes | Friends | List of all members
vpAdaptiveGain Class Reference
vs: Visual servoing module » Task and control laws

#include <visp3/vs/vpAdaptiveGain.h>

Public Member Functions

 vpAdaptiveGain ()
 
VP_EXPLICIT vpAdaptiveGain (double c)
 
 vpAdaptiveGain (double gain_at_zero, double gain_at_infinity, double slope_at_zero)
 
void initFromConstant (double c)
 
void initFromVoid (void)
 
void initStandard (double gain_at_zero, double gain_at_infinity, double slope_at_zero)
 
double setConstant (void)
 
double value_const (double x) const
 
double value (double x) const
 
double limitValue_const (void) const
 
double limitValue (void) const
 
double getLastValue (void) const
 
double operator() (double x) const
 
double operator() (const vpColVector &x) const
 
double operator() (void) const
 

Static Public Attributes

static const double DEFAULT_LAMBDA_ZERO = 1.666
 
static const double DEFAULT_LAMBDA_INFINITY = 0.1666
 
static const double DEFAULT_LAMBDA_SLOPE = 1.666
 

Friends

VISP_EXPORT std::ostream & operator<< (std::ostream &os, const vpAdaptiveGain &lambda)
 

Detailed Description

Adaptive gain computation.

As described in [23], a varying gain $ \lambda $ could be used in the visual servoing control law

\[{\bf v}_c = -\lambda {\bf L}^{+}_{e} {\bf e}\]

with

\[ \lambda (|| {\bf e}||) = (\lambda_0 - \lambda_\infty) e^{ -\frac{ \lambda'_0}{\lambda_0 - \lambda_\infty}||{\bf e}||} + \lambda_\infty \]

where:

  • $\lambda_0 = \lambda(0)$ is the gain in 0, that is for very small values of $||{\bf e}||$
  • $\lambda_\infty = \lambda_{||{\bf e}|| \rightarrow \infty}\lambda(||{\bf e}||)$ is the gain to infinity, that is for very high values of $||{\bf e}||$
  • $\lambda'_0$ is the slope of $\lambda$ at $||{\bf e}|| = 0$

As described in Tutorial: How to boost your visual servo control law, the interest of Using an adaptive gain is to reduce the time to convergence in order to speed up the servo.

The following example shows how to use this class in order to use an adaptive gain with the following parameters $\lambda_0 = 4$, $\lambda_\infty = 0.4 $ and $\lambda'_0 = 30$.

#include <visp3/vs/vpAdaptiveGain.h>
#include <visp3/vs/vpServo.h>
#ifdef ENABLE_VISP_NAMESPACE
using namespace VISP_NAMESPACE_NAME;
#endif
int main()
{
vpAdaptiveGain lambda(4, 0.4, 30); // lambda(0)=4, lambda(oo)=0.4 and lambda'(0)=30
vpServo servo;
servo.setLambda(lambda);
while(1) {
vpColVector v = servo.computeControlLaw();
}
}
vpAdaptiveGain
Adaptive gain computation.
Definition: vpAdaptiveGain.h:122
vpColVector
Implementation of column vector and the associated operations.
Definition: vpColVector.h:191
vpServo::setLambda
void setLambda(double c)
Definition: vpServo.h:986
vpServo::computeControlLaw
vpColVector computeControlLaw()
Definition: vpServo.cpp:705
VISP_NAMESPACE_NAME
Definition: vpEigenConversion.h:44

This other example shows how to use this class in order to set a constant gain $\lambda = 0.5$ that will ensure an exponential decrease of the task error.

#include <visp3/vs/vpAdaptiveGain.h>
#include <visp3/vs/vpServo.h>
int main()
{
vpAdaptiveGain lambda(0.5);
vpServo servo;
servo.setLambda(lambda);
while(1) {
vpColVector v = servo.computeControlLaw();
}
}
Examples
mbot-apriltag-2D-half-vs.cpp, mbot-apriltag-ibvs.cpp, mbot-apriltag-pbvs.cpp, servoAfma4Point2DCamVelocityKalman.cpp, servoAfma62DhalfCamVelocity.cpp, servoAfma6AprilTagIBVS.cpp, servoAfma6AprilTagPBVS.cpp, servoAfma6Cylinder2DCamVelocity.cpp, servoAfma6Cylinder2DCamVelocitySecondaryTask.cpp, servoAfma6FourPoints2DCamVelocityLs_cur_integrator.cpp, servoAfma6Points2DCamVelocityEyeToHand.cpp, servoBebop2.cpp, servoFlirPtuIBVS.cpp, servoFrankaIBVS.cpp, servoFrankaPBVS.cpp, servoPololuPtuPoint2DJointVelocity.cpp, servoUniversalRobotsIBVS.cpp, servoUniversalRobotsPBVS.cpp, servoViper850Point2DCamVelocityKalman.cpp, tutorial-ibvs-4pts-plotter-continuous-gain-adaptive.cpp, and tutorial-ibvs-4pts-plotter-gain-adaptive.cpp.

Definition at line 121 of file vpAdaptiveGain.h.

Constructor & Destructor Documentation

◆ vpAdaptiveGain() [1/3]

vpAdaptiveGain::vpAdaptiveGain ( )

Basic constructor which initializes all the parameters with their default value:

Definition at line 52 of file vpAdaptiveGain.cpp.

References initFromVoid().

◆ vpAdaptiveGain() [2/3]

vpAdaptiveGain::vpAdaptiveGain ( double  c)

Constructor that initializes the gain as constant. In that case $\lambda(||{\bf e}||) = c$.

Parameters
c: Value of the constant gain. A typical value is 0.5.

Definition at line 59 of file vpAdaptiveGain.cpp.

References initFromConstant().

◆ vpAdaptiveGain() [3/3]

vpAdaptiveGain::vpAdaptiveGain ( double  gain_at_zero,
double  gain_at_infinity,
double  slope_at_zero 
)

Constructor that initializes the gain as adaptive.

Parameters
gain_at_zero: the expected gain when $||{\bf e}||=0$: $\lambda_0$.
gain_at_infinity: the expected gain when $||{\bf e}||\rightarrow\infty$: $\lambda_\infty$.
slope_at_zero: the expected slope of $\lambda(||{\bf e}||)$ when $||{\bf e}||=0$: $\lambda'_0$.

Definition at line 61 of file vpAdaptiveGain.cpp.

References initStandard().

Member Function Documentation

◆ getLastValue()

double vpAdaptiveGain::getLastValue ( void  ) const
inline

Gets the last adaptive gain value which was stored in the class.

Returns
It returns the last adaptive gain value which was stored in the class.

Definition at line 261 of file vpAdaptiveGain.h.

◆ initFromConstant()

void vpAdaptiveGain::initFromConstant ( double  c)

Initializes the parameters to have a constant gain. In that case $\lambda(||{\bf e}||) = c$.

Parameters
c: Value of the constant gain. A typical value is 0.5.
Examples
servoAfma4Point2DCamVelocityKalman.cpp.

Definition at line 67 of file vpAdaptiveGain.cpp.

Referenced by vpAdaptiveGain().

◆ initFromVoid()

void vpAdaptiveGain::initFromVoid ( void  )

Initializes the parameters with the default value :

Definition at line 75 of file vpAdaptiveGain.cpp.

References DEFAULT_LAMBDA_INFINITY, DEFAULT_LAMBDA_SLOPE, DEFAULT_LAMBDA_ZERO, and initStandard().

Referenced by vpAdaptiveGain().

◆ initStandard()

void vpAdaptiveGain::initStandard ( double  gain_at_zero,
double  gain_at_infinity,
double  slope_at_zero 
)

Set the parameters $\lambda_0, \lambda_\infty, \lambda'_0$ used to compute $\lambda(||{\bf e}||)$.

Parameters
gain_at_zero: the expected gain when $||{\bf e}||=0$: $\lambda_0$.
gain_at_infinity: the expected gain when $||{\bf e}||\rightarrow\infty$: $\lambda_\infty$.
slope_at_zero: the expected slope of $\lambda(||{\bf e}||)$ when $||{\bf e}||=0$: $\lambda'_0$.
Examples
mbot-apriltag-2D-half-vs.cpp, mbot-apriltag-ibvs.cpp, mbot-apriltag-pbvs.cpp, servoAfma4Point2DCamVelocityKalman.cpp, servoAfma6FourPoints2DCamVelocityLs_cur_integrator.cpp, servoAfma6Points2DCamVelocityEyeToHand.cpp, and servoViper850Point2DCamVelocityKalman.cpp.

Definition at line 82 of file vpAdaptiveGain.cpp.

Referenced by initFromVoid(), and vpAdaptiveGain().

◆ limitValue()

double vpAdaptiveGain::limitValue ( void  ) const

Gets the value of the gain at infinity (ie the value of $\lambda_\infty = c $) and stores it as a parameter of the class.

Returns
It returns the value of the gain at infinity.

Definition at line 129 of file vpAdaptiveGain.cpp.

References limitValue_const().

Referenced by operator()().

◆ limitValue_const()

double vpAdaptiveGain::limitValue_const ( void  ) const

Gets the value of the gain at infinity (ie the value of $ \lambda_\infty = c $). This function is similar to limitValue() except that here the value is not stored as a parameter of the class.

Returns
It returns the value of the gain at infinity.

Definition at line 115 of file vpAdaptiveGain.cpp.

Referenced by limitValue().

◆ operator()() [1/3]

double vpAdaptiveGain::operator() ( const vpColVector &  x) const

Operator which computes $\lambda({||x||}_{\infty})$ where

\[ \lambda ({||x||}_{\infty}) = (\lambda_0 - \lambda_\infty) e^{ -\frac{ \lambda'_0}{\lambda_0 - \lambda_\infty}{||x||}_{\infty}} + \lambda_\infty \]

Parameters
x: Input vector $ \bf x$ to consider.
Returns
It returns the value of the computed gain.

Definition at line 140 of file vpAdaptiveGain.cpp.

References vpColVector::infinityNorm(), and value().

◆ operator()() [2/3]

double vpAdaptiveGain::operator() ( double  x) const

Operator that computes $\lambda(x)$ where

\[ \lambda (x) = (\lambda_0 - \lambda_\infty) e^{ -\frac{ \lambda'_0}{\lambda_0 - \lambda_\infty}x} + \lambda_\infty \]

Parameters
x: Input value to consider. During a visual servo this value can be the Euclidean norm $||{\bf e}||$ or the infinity norm $||{\bf e}||_{\infty}$ of the task function.
Returns
It returns the value of the computed gain.
See also
value()

Definition at line 136 of file vpAdaptiveGain.cpp.

References value().

◆ operator()() [3/3]

double vpAdaptiveGain::operator() ( void  ) const

Gets the value of the gain at infinity (ie the value of $\lambda_\infty = c $).

Returns
It returns the value of the gain at infinity.
See also
limitValue()

Definition at line 138 of file vpAdaptiveGain.cpp.

References limitValue().

◆ setConstant()

double vpAdaptiveGain::setConstant ( void  )

Sets the internal parameters in order to obtain a constant gain equal to the gain in 0 set through the parameter $\lambda_0$.

Returns
It returns the value of the constant gain $\lambda_0$.

Definition at line 97 of file vpAdaptiveGain.cpp.

◆ value()

double vpAdaptiveGain::value ( double  x) const

Computes the value of the adaptive gain $\lambda(x)$ using:

\[ \lambda (x) = (\lambda_0 - \lambda_\infty) e^{ -\frac{ \lambda'_0}{\lambda_0 - \lambda_\infty}x} + \lambda_\infty \]

This value is stored as a parameter of the class.

Parameters
x: Input value to consider. During a visual servo this value can be the Euclidean norm $||{\bf e}||$ or the infinity norm $||{\bf e}||_{\infty}$ of the task function.
Returns
It returns the value of the computed gain.

Definition at line 122 of file vpAdaptiveGain.cpp.

References value_const().

Referenced by operator()().

◆ value_const()

double vpAdaptiveGain::value_const ( double  x) const

Computes the value of the adaptive gain $\lambda(x)$ using:

\[ \lambda (x) = (\lambda_0 - \lambda_\infty) e^{ -\frac{ \lambda'_0}{\lambda_0 - \lambda_\infty}x} + \lambda_\infty \]

Parameters
x: Input value to consider. During a visual servo this value can be the Euclidean norm $||{\bf e}||$ or the infinity norm $||{\bf e}||_{\infty}$ of the task function.
Returns
It returns the value of the computed gain.

Definition at line 108 of file vpAdaptiveGain.cpp.

Referenced by value().

Friends And Related Function Documentation

◆ operator<<

VISP_EXPORT std::ostream& operator<< ( std::ostream &  os,
const vpAdaptiveGain &  lambda 
)
friend

Prints the adaptive gain parameters $\lambda_0, \lambda_\infty, \lambda'_0$.

Parameters
os: The stream where to print the adaptive gain parameters.
lambda: The adaptive gain containing the parameters to print.

Definition at line 142 of file vpAdaptiveGain.cpp.

Member Data Documentation

◆ DEFAULT_LAMBDA_INFINITY

const double vpAdaptiveGain::DEFAULT_LAMBDA_INFINITY = 0.1666
static

Definition at line 125 of file vpAdaptiveGain.h.

Referenced by initFromVoid().

◆ DEFAULT_LAMBDA_SLOPE

const double vpAdaptiveGain::DEFAULT_LAMBDA_SLOPE = 1.666
static

Definition at line 126 of file vpAdaptiveGain.h.

Referenced by initFromVoid().

◆ DEFAULT_LAMBDA_ZERO

BEGIN_VISP_NAMESPACE const double vpAdaptiveGain::DEFAULT_LAMBDA_ZERO = 1.666
static

Definition at line 124 of file vpAdaptiveGain.h.

Referenced by initFromVoid().