KalmanFilterBLAS Class Reference

BLAS/LAPACK optimized operations for Kalman filtering. More...

Detailed Description

BLAS/LAPACK optimized operations for Kalman filtering.

This class provides high-performance implementations of core Kalman filter operations using CBLAS and LapackWrapper while maintaining Eigen matrix interfaces for compatibility with existing code.

Design principle: Single Responsibility - handles only mathematical computations

Collaboration diagram for KalmanFilterBLAS:

Static Public Member Functions
static void	computeInnovationCovariance (const MatrixXd &H, const MatrixXd &P, const MatrixXd &R, MatrixXd &Q)
	Compute innovation covariance Q = H*P*H' + R using BLAS.
static bool	computeKalmanGain (const MatrixXd &H, const MatrixXd &P, MatrixXd &Q, MatrixXd &K, MatrixXd *Qinv, E_Inverter inverter)
	Compute Kalman gain K = P*H'*Q^-1 using BLAS/LAPACKE.
static void	computeHP (const MatrixXd &H, const MatrixXd &P, MatrixXd &HP)
	Compute H*P matrix product using BLAS.
static void	updateStateVector (const VectorXd &x, const MatrixXd &K, const VectorXd &v, VectorXd &xp, VectorXd &dx, int begX, int numX, int begH, int numH)
	Update state vector: xp = x + K*v using BLAS.
static void	updateCovarianceStandard (const MatrixXd &P, const MatrixXd &K, const MatrixXd &H, const MatrixXd &HP, MatrixXd &Pp)
	Update covariance: Pp = P - K*H*P (standard form) using BLAS.
static void	updateCovarianceJoseph (const MatrixXd &P, const MatrixXd &K, const MatrixXd &H, const MatrixXd &R, MatrixXd &Pp)
	Update covariance: Pp = (I-K*H)*P*(I-K*H)' + K*R*K' (Joseph form) using BLAS.
static bool	solveLinearSystem (MatrixXd &Q, MatrixXd &B, E_Inverter inverter)
	Solve Q*X = B for multiple right-hand sides using LAPACKE.

Member Function Documentation

computeHP()

void KalmanFilterBLAS::computeHP ( const MatrixXd & H, const MatrixXd & P, MatrixXd & HP )

static

Compute H*P matrix product using BLAS.

Parameters

H	Design matrix (numH x numX)
P	State covariance (numX x numX)
HP	Output matrix (numH x numX)

Here is the call graph for this function:

Here is the caller graph for this function:

computeInnovationCovariance()

void KalmanFilterBLAS::computeInnovationCovariance ( const MatrixXd & H, const MatrixXd & P, const MatrixXd & R, MatrixXd & Q )

static

Compute innovation covariance Q = H*P*H' + R using BLAS.

Parameters

H	Design matrix (numH x numX)
P	State covariance (numX x numX)
R	Measurement noise (numH x numH)
Q	Output innovation covariance (numH x numH)

Here is the call graph for this function:

computeKalmanGain()

bool KalmanFilterBLAS::computeKalmanGain ( const MatrixXd & H, const MatrixXd & P, MatrixXd & Q, MatrixXd & K, MatrixXd * Qinv, E_Inverter inverter )

static

Compute Kalman gain K = P*H'*Q^-1 using BLAS/LAPACKE.

Parameters

H	Design matrix (numH x numX)
P	State covariance (numX x numX)
Q	Innovation covariance (numH x numH) - will be modified!
K	Output Kalman gain (numX x numH)
Qinv	Optional output for Q inverse
inverter	Inversion method to use

Returns

true if successful, false if inversion failed

Here is the call graph for this function:

solveLinearSystem()

bool KalmanFilterBLAS::solveLinearSystem ( MatrixXd & Q, MatrixXd & B, E_Inverter inverter )

static

Solve Q*X = B for multiple right-hand sides using LAPACKE.

Parameters

Q	Coefficient matrix (n x n) - will be modified!
B	Right-hand side / solution matrix (n x m) - will be modified!
inverter	Solver method to use

Returns

true if successful, false otherwise

Here is the call graph for this function:

Here is the caller graph for this function:

updateCovarianceJoseph()

void KalmanFilterBLAS::updateCovarianceJoseph ( const MatrixXd & P, const MatrixXd & K, const MatrixXd & H, const MatrixXd & R, MatrixXd & Pp )

static

Update covariance: Pp = (I-K*H)*P*(I-K*H)' + K*R*K' (Joseph form) using BLAS.

Parameters

P	Prior covariance (numX x numX)
K	Kalman gain (numX x numH)
H	Design matrix (numH x numH)
R	Measurement noise (numH x numH)
Pp	Output posterior covariance (numX x numX)

Here is the call graph for this function:

updateCovarianceStandard()

void KalmanFilterBLAS::updateCovarianceStandard ( const MatrixXd & P, const MatrixXd & K, const MatrixXd & H, const MatrixXd & HP, MatrixXd & Pp )

static

Update covariance: Pp = P - K*H*P (standard form) using BLAS.

Parameters

P	Prior covariance (numX x numX)
K	Kalman gain (numX x numH)
H	Design matrix (numH x numX)
HP	Pre-computed H*P (numH x numX)
Pp	Output posterior covariance (numX x numX)

Here is the call graph for this function:

updateStateVector()

void KalmanFilterBLAS::updateStateVector ( const VectorXd & x, const MatrixXd & K, const VectorXd & v, VectorXd & xp, VectorXd & dx, int begX, int numX, int begH, int numH )

static

Update state vector: xp = x + K*v using BLAS.

Parameters

x	Prior state (numX)
K	Kalman gain (numX x numH)
v	Innovation vector (numH)
xp	Output posterior state (numX)
dx	Output state change (numX)
begX	Starting index in full state vector
numX	Number of states to update
begH	Starting index in measurement vector
numH	Number of measurements

Here is the call graph for this function:

---

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

• /ginan/src/cpp/common/kalmanBlas.hpp
• /ginan/src/cpp/common/kalmanBlas.cpp