Virtual Class representing all particle filters. More...

#include <particlefilter.h>

Inheritance diagram for BFL::ParticleFilter< StateVar, MeasVar >:

Public Member Functions
	ParticleFilter (MCPdf< StateVar > prior, ConditionalPdf< StateVar, StateVar > proposal, int resampleperiod=0, double resamplethreshold=0, int resamplescheme=DEFAULT_RS)
	Constructor.
	ParticleFilter (MCPdf< StateVar > prior, MCPdf< StateVar > post, ConditionalPdf< StateVar, StateVar > *proposal, int resampleperiod=0, double resamplethreshold=0, int resamplescheme=DEFAULT_RS)
	Constructor.
	ParticleFilter (const ParticleFilter< StateVar, MeasVar > &filt)
	Copy Constructor.
virtual MCPdf< StateVar > *	PostGet ()
	Get Posterior density.
ConditionalPdf< StateVar, StateVar > *	ProposalGet ()
	Get a pointer to the proposal density.
virtual void	ProposalSet (ConditionalPdf< StateVar, StateVar > *const cpdf)
	Set the proposal density.
virtual	~ParticleFilter ()
	Destructor.
Protected Member Functions
virtual bool	DynamicResampleStep ()
	Resample if necessary.
virtual bool	ProposalStepInternal (SystemModel< StateVar > const sysmodel, const StateVar &u, MeasurementModel< MeasVar, StateVar > const measmodel, const MeasVar &z, const StateVar &s)
	Proposal step.
virtual bool	Resample ()
	Actual Resampling happens here;.
virtual bool	StaticResampleStep ()
	Resample if wanted.
virtual bool	UpdateInternal (SystemModel< StateVar > const sysmodel, const StateVar &u, MeasurementModel< MeasVar, StateVar > const measmodel, const MeasVar &z, const StateVar &s)
	Actual implementation of Update, varies along filters.
virtual bool	UpdateWeightsInternal (SystemModel< StateVar > const sysmodel, const StateVar &u, MeasurementModel< MeasVar, StateVar > const measmodel, const MeasVar &z, const StateVar &s)
	Update Weights.
Protected Attributes
bool	_created_post
	created own post
bool	_dynamicResampling
	Dynamic resampling or fixed period resampling?
vector< WeightedSample < StateVar > >	_new_samples
	While updating store list of new samples.
vector< Sample< StateVar > >	_new_samples_unweighted
	While resampling.
vector< WeightedSample < StateVar > >::iterator	_ns_it
	Iterator for new list of samples.
vector< WeightedSample < StateVar > >	_old_samples
	While updating store list of old samples.
vector< WeightedSample < StateVar > >::iterator	_os_it
	Iterator for old list of samples.
ConditionalPdf< StateVar, StateVar > *	_proposal
	Pointer to the Proposal Density.
bool	_proposal_depends_on_meas
	Proposal depends on last measurement?
int	_resamplePeriod
	Number of timestep between resampling from the Posterior Pdf.
int	_resampleScheme
	Which resample algorithm (see top of particle.h for #defines)
double	_resampleThreshold
	Threshold used when dynamic resampling.
WeightedSample< StateVar >	_sample
	While updating use sample<StateVar>

Detailed Description

template<typename StateVar, typename MeasVar>
class BFL::ParticleFilter< StateVar, MeasVar >

Virtual Class representing all particle filters.

This is a virtual class representing the family of all particle filters. Particle filters are filters in which the Posterior density is represented by a set of particles (aka (weighted) samples). In other words, the posterior density is a Monte Carlo Pdf (class MCPdf)

However, the updating of the Posterior density can still be done in several ways, that's why the System and Measurement update members are still pure virtual functions.

This class is the base class for all sorts of particle filters.

Todo:: : Actually all particle filters represented by this class are of the "Sequential importance sampling methods" type. Typical of those methods is the so called Proposal density. In theory it would be possible to create Filters using a recursive version of other Monte Carlo methods (eg. MCMC methods), although I am not aware of any of these (due to the increased complexity).

See also:: MCPdf; Sample; WeightedSample

Bug:: Resampling is not implemented generically enough yet. There's only the possibility to choose between static period resampling and dynamic resampling as proposed by Jun Liu. The correct way of implementing this would be to create a virtual function that has to be implemented by the user, but this creates more hassle for the user (a different particle filter for each scheme).

Definition at line 159 of file particlefilter.h.

Constructor & Destructor Documentation

template<typename SV, typename MV >

BFL::ParticleFilter< SV, MV >::ParticleFilter	(	MCPdf< SV > *	prior,
		ConditionalPdf< SV, SV > *	proposal,
		int	resampleperiod = `0`,
		double	resamplethreshold = `0`,
		int	resamplescheme = `DEFAULT_RS`
	)

Constructor.

Precondition:: you created the necessary models and the prior

Parameters:

prior	pointer to the Monte Carlo Pdf prior density
proposal	pointer to the proposal density to use
resampleperiod	fixed resampling period (if desired)
resamplethreshold	threshold used when dynamic resampling
resamplescheme	resampling scheme, see header file for different defines and their meaning

Bug:

prior should be of type pdf and not mcpdf. See also notes with implementation

let the user implement her/his own resamplescheme

Definition at line 29 of file particlefilter.h.

template<typename SV, typename MV >

BFL::ParticleFilter< SV, MV >::ParticleFilter	(	MCPdf< SV > *	prior,
		MCPdf< SV > *	post,
		ConditionalPdf< SV, SV > *	proposal,
		int	resampleperiod = `0`,
		double	resamplethreshold = `0`,
		int	resamplescheme = `DEFAULT_RS`
	)

Constructor.

Precondition:: you created the necessary models and the prior

Parameters:

prior	pointer to the Monte Carlo Pdf prior density
post	pointer to the Monte Carlo Pdf post density
proposal	pointer to the proposal density to use
resampleperiod	fixed resampling period (if desired)
resamplethreshold	threshold used when dynamic resampling
resamplescheme	resampling scheme, see header file for different defines and their meaning

Bug:

prior should be of type pdf and not mcpdf. See also notes with implementation

let the user implement her/his own resamplescheme

Definition at line 74 of file particlefilter.h.

template<typename SV , typename MV >

BFL::ParticleFilter< SV, MV >::~ParticleFilter ( ) [virtual]

Destructor.

Definition at line 116 of file particlefilter.cpp.

template<typename SV, typename MV>

BFL::ParticleFilter< SV, MV >::ParticleFilter ( const ParticleFilter< SV, MV > & filt )

Copy Constructor.

Bug:: implementation probably contains a bug

Definition at line 123 of file particlefilter.h.

Member Function Documentation

template<typename SV , typename MV >

bool BFL::ParticleFilter< SV, MV >::DynamicResampleStep ( ) [protected, virtual]

Resample if necessary.

Bug:: let the user implement her/his own resamplescheme

Definition at line 272 of file particlefilter.cpp.

template<typename SV , typename MV >

MCPdf< SV > * BFL::ParticleFilter< SV, MV >::PostGet ( ) [virtual]

Get Posterior density.

Get the current Posterior density

Returns:: a pointer to the current posterior

Reimplemented from BFL::Filter< StateVar, MeasVar >.

Definition at line 390 of file particlefilter.cpp.

template<typename SV , typename MV >

ConditionalPdf< SV, SV > * BFL::ParticleFilter< SV, MV >::ProposalGet ( )

Get a pointer to the proposal density.

Returns:: a pointer to the proposal density

Definition at line 139 of file particlefilter.cpp.

template<typename SV, typename MV >

void BFL::ParticleFilter< SV, MV >::ProposalSet ( ConditionalPdf< SV, SV > *const cpdf ) [virtual]

Set the proposal density.

Parameters:

cpdf	the new proposal density. The order of the conditional arguments is fixed and should be: x (state), u (input), z (measurement), s (sensor param). Off course all of them are optional

Definition at line 133 of file particlefilter.cpp.

template<typename SV, typename MV>

bool BFL::ParticleFilter< SV, MV >::ProposalStepInternal	(	SystemModel< SV > *const	sysmodel,
		const SV &	u,
		MeasurementModel< MV, SV > *const	measmodel,
		const MV &	z,
		const SV &	s
	)		`[protected, virtual]`

Proposal step.

Implementation of proposal step

Parameters:

sysmodel	pointer to the used system model
u	input param for proposal density
measmodel	pointer to the used measurementmodel
z	measurement param for proposal density
s	sensor param for proposal density

Bug:: Make sampling method variable. See implementation.

Reimplemented in BFL::EKParticleFilter.

Definition at line 145 of file particlefilter.cpp.

template<typename SV , typename MV >

bool BFL::ParticleFilter< SV, MV >::Resample ( ) [protected, virtual]

Actual Resampling happens here;.

Reimplemented in BFL::EKParticleFilter.

Definition at line 345 of file particlefilter.cpp.

template<typename SV , typename MV >

bool BFL::ParticleFilter< SV, MV >::StaticResampleStep ( ) [protected, virtual]

Resample if wanted.

Bug:: let the user implement her/his own resamplescheme

Definition at line 308 of file particlefilter.cpp.

template<typename StateVar, typename MeasVar>

bool BFL::ParticleFilter< SV, MV >::UpdateInternal	(	SystemModel< StateVar > *const	sysmodel,
		const StateVar &	u,
		MeasurementModel< MeasVar, StateVar > *const	measmodel,
		const MeasVar &	z,
		const StateVar &	s
	)		`[protected, virtual]`

Actual implementation of Update, varies along filters.

Parameters:

sysmodel	pointer to the used system model
u	input param for proposal density
measmodel	pointer to the used measurementmodel
z	measurement param for proposal density
s	sensor param for proposal density

Implements BFL::Filter< StateVar, MeasVar >.

Reimplemented in BFL::ASIRFilter< StateVar, MeasVar >, BFL::BootstrapFilter< StateVar, MeasVar >, and BFL::EKParticleFilter.

Definition at line 318 of file particlefilter.cpp.

template<typename SV, typename MV>

bool BFL::ParticleFilter< SV, MV >::UpdateWeightsInternal	(	SystemModel< SV > *const	sysmodel,
		const SV &	u,
		MeasurementModel< MV, SV > *const	measmodel,
		const MV &	z,
		const SV &	s
	)		`[protected, virtual]`

Update Weights.

Parameters:

sysmodel	pointer to the used system model
u	input param for proposal density
measmodel	pointer to the used measurementmodel
z	measurement param for proposal density
s	sensor param for proposal density