Classes
struct	Observation
	Sigma-weighted pose observation pushed by Resync-Steps. More...
struct	LocalizationConfig
class	Localization
	Thread-safe particle-filter localization service. More...
struct	RecorderConfig
class	LocalizationRecorder

Enumerations
enum class	ResamplingAlgorithm : uint8_t { Multinomial = 0 , Residual = 1 , Stratified = 2 , Systematic = 3 }

Functions
void	computeResampling (ResamplingAlgorithm method, const std::vector< double > &logWeights, std::mt19937 &rng, std::vector< size_t > &outIndexes, size_t outCount=0)
double	effectiveSampleRatio (const std::vector< double > &logWeights)

Enumeration Type Documentation

◆ ResamplingAlgorithm

strong

Enumerator
Multinomial	Classical: M independent samples from the CDF.
Residual	Deterministic floor(M*w) part + multinomial remainder.
Stratified	Strata of size 1/M, one uniform draw per stratum.
Systematic	Single uniform offset + fixed-step traversal (lowest variance).

Definition at line 24 of file resampling.hpp.

void libstp::localization::computeResampling	(	ResamplingAlgorithm	method,
		const std::vector< double > &	logWeights,
		std::mt19937 &	rng,
		std::vector< size_t > &	outIndexes,
		size_t	outCount = 0 )

inline

Compute resampling indexes from log-weights.

Parameters

method	One of ResamplingAlgorithm.
logWeights	Per-particle log-weights (unnormalized).
rng	Random source.
outIndexes	Filled with indexes drawn from logWeights. Resized to outCount (or logWeights.size() if outCount == 0).
outCount	Desired output count; 0 means "same as input".

Definition at line 41 of file resampling.hpp.

double libstp::localization::effectiveSampleRatio ( const std::vector< double > & logWeights )

inline

Effective sample size (ESS) computed from log-weights, normalized to [0, 1].

Log-sum-exp trick to dodge underflow on tightly peaked weight distributions. Ported from MRPT's CParticleFilterDataImpl::ESS.

Definition at line 199 of file resampling.hpp.