mc_filter::ExponentialMovingAverage< VectorT > Struct Template Reference

#include <mc_filter/ExponentialMovingAverage.h>

Public Member Functions
	ExponentialMovingAverage (double dt, double timeConstant, const VectorT &initValue=VectorT::Zero())
void	append (const VectorT &value)
const VectorT &	eval () const
void	saturation (double limit)
void	reset (const VectorT &initVal)
double	timeConstant () const
void	timeConstant (double T)

Protected Attributes
VectorT	average_ = VectorT::Zero()
double	alpha_
double	dt_
double	timeConstant_
double	saturation_ = -1.

Detailed Description

template<typename VectorT>
struct mc_filter::ExponentialMovingAverage< VectorT >

Exponential Moving Average.

This filter can be seen as an integrator:

\[ y(t) = \frac{1}{T} \int_{u=0}^t x(u) e^{(u - t) / T} {\rm d}{\tau} \]

with T > 0 a reset period acting as anti-windup. It can also (informally) be interpreted as the average value of the input signal x(t) over the last T seconds. Formally, it represents the amount of time for the smoothed response of a unit input to reach 1-1/e (~63%) of the original signal.

See https://en.wikipedia.org/wiki/Exponential_smoothing. It is equivalent to a low-pass filter https://en.wikipedia.org/wiki/Low-pass_filter applied to the integral of the input signal.

Expects type VectorT to have:

• VectorT::Zero() static function
• VectorT::setZero() Common types: Eigen::Vector3d, Eigen::Vector6d, etc.

Constructor & Destructor Documentation

ExponentialMovingAverage()

template<typename VectorT >

mc_filter::ExponentialMovingAverage< VectorT >::ExponentialMovingAverage ( double  dt, double  timeConstant, const VectorT &  initValue = VectorT::Zero()  )

inline

Constructor.

Parameters

dt	Time in [s] between two readings.
timeConstant	Informally, length of the recent-past window, in [s].
initValue	Initial value of the output average.

Member Function Documentation

append()

template<typename VectorT >

void mc_filter::ExponentialMovingAverage< VectorT >::append ( const VectorT &  value )

inline

Append a new reading to the series.

Parameters

value

New value.

eval()

template<typename VectorT >

const VectorT& mc_filter::ExponentialMovingAverage< VectorT >::eval ( ) const

inline

Evaluate the smoothed statistic.

reset()

template<typename VectorT >

void mc_filter::ExponentialMovingAverage< VectorT >::reset ( const VectorT &  initVal )

inline

Reset average to provided value.

Parameters

initVal

initial value of the average

saturation()

template<typename VectorT >

void mc_filter::ExponentialMovingAverage< VectorT >::saturation ( double  limit )

inline

Set output saturation; disable by providing a negative value.

Parameters

limit

Output will saturate between -limit and +limit.

timeConstant() [1/2]

template<typename VectorT >

double mc_filter::ExponentialMovingAverage< VectorT >::timeConstant ( ) const

inline

Get time constant of the filter.

timeConstant() [2/2]

template<typename VectorT >

void mc_filter::ExponentialMovingAverage< VectorT >::timeConstant ( double  T )

inline

Update time constant.

Parameters

T	New time constant of the filter.

Note

T is explicitely enforced to respect the Nyquist–Shannon sampling theorem, that is T is at least 2*timestep.

Member Data Documentation

alpha_

template<typename VectorT >

double mc_filter::ExponentialMovingAverage< VectorT >::alpha_

protected

average_

template<typename VectorT >

VectorT mc_filter::ExponentialMovingAverage< VectorT >::average_ = VectorT::Zero()

protected

dt_

template<typename VectorT >

double mc_filter::ExponentialMovingAverage< VectorT >::dt_

protected

saturation_

template<typename VectorT >

double mc_filter::ExponentialMovingAverage< VectorT >::saturation_ = -1.

protected

timeConstant_

template<typename VectorT >

double mc_filter::ExponentialMovingAverage< VectorT >::timeConstant_

protected

---

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

• include/mc_filter/ExponentialMovingAverage.h