lds #

Linear Dynamical Systems (LDS) namespace.

Namespaces #

Name
lds::gaussian Linear Dynamical Systems with Gaussian observations.
lds::poisson Linear Dynamical Systems with Poisson observations.

Classes #

	Name
class	lds::Controller
class	lds::EM
class	lds::Fit LDS Fit Type.
class	lds::SSID
class	lds::SwitchedController SwitchedController Type.
class	lds::System Linear Dynamical System Type.
class	lds::UniformMatrixList
class	lds::UniformSystemList
class	lds::UniformVectorList

Types #

	Name
enum	SSIDWt { kSSIDNone, kSSIDMOESP, kSSIDCVA} weighting options for SSID
enum	MatrixListFreeDim { kMatFreeDimNone, kMatFreeDim1, kMatFreeDim2}
using double	data_t
using arma::Col< data_t >	Vector
using arma::Mat< data_t >	Matrix
using arma::Cube< data_t >	Cube
using arma::subview< data_t >	View

Functions #

	Name
void	Limit(std::vector< data_t > & x, data_t lb, data_t ub)
void	Limit(Vector & x, data_t lb, data_t ub)
void	Limit(Matrix & x, data_t lb, data_t ub)
void	Reassign(Vector & some, const Vector & other, const std::string & parenthetical =“Reassign”) reassigns contents of some Vector in place
void	Reassign(Matrix & some, const Matrix & other, const std::string & parenthetical =“Reassign”) reassigns contents of some Matrix in place
void	ForceSymPD(Matrix & X) forces matrix to be symmetric positive-definite
void	ForceSymMinEig(Matrix & X, data_t eig_min =0) forces matrix to be symmetric and have a minimum eigenvalue
void	lq(Matrix & L, Matrix & Qt, const Matrix & X) LQ decomposition.
Matrix	calcCov(const Matrix & A, const Matrix & B) Calculate covariance matrix.

Attributes #

	Name
const std::size_t	kControlTypeDeltaU control designed to penalize change in input
const std::size_t	kControlTypeIntY control using integral action
const std::size_t	kControlTypeAdaptM adapt control setpoint with re-estimated disturbance `m`
const data_t	kInf Some useful numbers.
const data_t	kPi
const data_t	kDefaultP0 default state estimate covar
const data_t	kDefaultQ0 default process noise covar
const data_t	kDefaultR0 default output noise covar

Type Details #

SSIDWt #

Enumerator	Value	Description
kSSIDNone		None.
kSSIDMOESP		MOESP (AKA “robust method” in van Overschee 1996)
kSSIDCVA		CVA “Canonical Variate Analysis”.

Weighting options for singular value decomposition performed during subspace identification (SSID)

Reference:

van Overschee, de Moor. 1996. Subspace Identification for Linear Systems.

MatrixListFreeDim #

Enumerator	Value	Description
kMatFreeDimNone		neither dim free to be hetero in mat list
kMatFreeDim1		allow 1st dim of mats in list to be hetero
kMatFreeDim2		allow 2nd dim of mats in list to be hetero

data_t #

using lds::data_t = typedef double;

Type of all data in library. If need 32b, change double to float. This could be potentially useful for large scale problems where there are memory constraints.

Vector #

using lds::Vector = typedef arma::Col<data_t>;

Matrix #

using lds::Matrix = typedef arma::Mat<data_t>;

Cube #

using lds::Cube = typedef arma::Cube<data_t>;

View #

using lds::View = typedef arma::subview<data_t>;

Function Details #

Limit #

inline void Limit(
    std::vector< data_t > & x,
    data_t lb,
    data_t ub
)

Limit #

inline void Limit(
    Vector & x,
    data_t lb,
    data_t ub
)

Limit #

inline void Limit(
    Matrix & x,
    data_t lb,
    data_t ub
)

Reassign #

inline void Reassign(
    Vector & some,
    const Vector & other,
    const std::string & parenthetical ="Reassign"
)

Parameters:

some some Vector
other other Vector
parenthetical optional description provided by caller to ease debugging

Reassign #

inline void Reassign(
    Matrix & some,
    const Matrix & other,
    const std::string & parenthetical ="Reassign"
)

Parameters:

some some Matrix
other other Matrix
parenthetical optional description provided by caller to ease debugging

ForceSymPD #

void ForceSymPD(
    Matrix & X
)

Parameters:

X mutated matrix

ForceSymMinEig #

void ForceSymMinEig(
    Matrix & X,
    data_t eig_min =0
)

Parameters:

X mutated matrix
eig_min [optional] minimum eigen value

lq #

void lq(
    Matrix & L,
    Matrix & Qt,
    const Matrix & X
)

Parameters:

L lower triangle matrix
Qt orthonormal matrix (transposed cf QR decomp)
X matrix being decomposed

calcCov #

Matrix calcCov(
    const Matrix & A,
    const Matrix & B
)

Parameters:

A some matrix
B some other matrix

Return: covariance

Attribute Details #

kControlTypeDeltaU #

static const std::size_t kControlTypeDeltaU = 0x1;

Control was designed to penalize change in input (i.e., the state was augmented with input u)

kControlTypeIntY #

static const std::size_t kControlTypeIntY = kControlTypeDeltaU << 1;

Control using integral action (i.e., the state was augmented with output y during design)

kControlTypeAdaptM #

static const std::size_t kControlTypeAdaptM = kControlTypeDeltaU << 2;

Adapt control setpoint adapted with re-estimated process disturbance m.

kInf #

static const data_t kInf = std::numeric_limits<data_t>::infinity();

kPi #

static const data_t kPi = arma::datum::pi;

kDefaultP0 #

static const data_t kDefaultP0 = 1e-6;

kDefaultQ0 #

static const data_t kDefaultQ0 = 1e-6;

kDefaultR0 #

static const data_t kDefaultR0 = 1e-2;

Updated on 19 May 2022 at 17:16:03 Eastern Daylight Time