Kalman Class Reference

A one-dimensional Kalman filter, which has as its state variables position, measured in radians from some implied reference point, and a reference speed, in radians per unit time, which is the actual maximum speed when the telescope motors are full on (ie, corresponding to commanded speed 1.0). More...

#include <kalman.h>

Public Member Functions
	Kalman ()
	No-arg constructor.
	Kalman (double theta, double maxthetadot, double dz, ms_t initial_time)
	Create a one-dimensional Kalman filter for angular position. More...
void	update (ms_t t, double speed)
	Update the position. More...
void	click (ms_t t, double speed, double click_angle)
	Update the position with a position measurement. More...
void	reset_angle (double ang)
	Set the current estimate of the angle. More...
double	angle (void) const
	The current estimate of the angle.
double	fullspeed_pos (void) const
	The current estimate of the reference/maximum speed, when driven in the positive direction.
double	fullspeed_neg (void) const
	The current estimate of the reference/maximum speed, when driven in the negative direction.
double	angle_error (void) const
	An estimate of the error in angular position, in units of radians-squared.
double	fullspeed_error (void) const
	An estimate of the error in reference speed.
void	verbose (char verbose_char)
	Enable tracing of the behaviour of the Kalman filter. More...

Detailed Description

A one-dimensional Kalman filter, which has as its state variables position, measured in radians from some implied reference point, and a reference speed, in radians per unit time, which is the actual maximum speed when the telescope motors are full on (ie, corresponding to commanded speed 1.0).

The 'speeds' provided to this class, in the range [-1.0,1.0], are relative to this reference speed.

Constructor & Destructor Documentation

Kalman::Kalman	(	double	theta,
		double	maxthetadot,
		double	dz,
		ms_t	initial_time
	)

Create a one-dimensional Kalman filter for angular position.

The filter manages estimates for the position and the angular reference speed. This speed is the speed in rad/s which corresponds to relative speed 1.0.

The filter is reasonably insensitive to the initial values chosen here; indeed part of the point of it is to improve these and their estimated errors. The filter does not (currently) maintain an estimate of the error dz. Note that if this dz error is estimated too small, the filter becomes rather unstable.

The angle is not normalised to stay in any particular range.

Parameters

theta	the initial estimate of the angle (rad)
maxthetadot	the initial estimate of the reference speed (rad/s), corresponding to commanded speed 1.0
dz	an estimate of the error in the actual positions of the reed switches (rad)
initial_time	the start time of the filter (ms)

Member Function Documentation

void Kalman::click	(	ms_t	t,
		double	speed,
		double	ref_angle
	)

Update the position with a position measurement.

The provided speed is taken to have been constant over the interval from the last update to now.

Parameters

t	the current time (ms)
speed	the current speed (fractional speed)
ref_angle	the current actual angle (rad)

See also

update(ms_t,double)

void Kalman::reset_angle

(

double

ang

)

Set the current estimate of the angle.

This is not an update of the position estimate, but an unequivocal reset of the value.

Parameters

ang	the current angle

void Kalman::update	(	ms_t	t,
		double	speed
	)

Update the position.

The provided speed is taken to have been constant over the interval from the last update to now. This is equivalent to the click method but with an infinite measurement variance, and hence a trivial (effectively absent) corrector step.

Parameters

t	the current time (ms)
speed	the current speed (fractional speed)

void Kalman::verbose ( char  verbose_char )

inline

Enable tracing of the behaviour of the Kalman filter.

After this method is called with a character argument, updates to the filter result in a tracing line written to the serial port, which starts with that character. The format of these is not currently documented here: see the code for details.

---

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

• kalman.h
• kalman.cpp