raisim::SparseJacobian

class · #include <raisim/math.hpp>

Sparse Jacobian with three rows (e.g. of a point's position or of a body's angular velocity with respect to the generalized velocity) that stores only its non-zero columns.

v holds the values as a dense 3 x size block (column-major); column i of v is column idx[i] of the full 3 x dof Jacobian. All other columns of the full Jacobian are zero.

Constructors & destructor

SparseJacobian

SparseJacobian()

~SparseJacobian

~SparseJacobian()

Public member functions

resize

void resize(size_t cols)

Sets the number of stored columns to cols. The storage is reallocated only when cols exceeds the current capacity (the old values are then lost). All stored values are set to zero; the entries of idx are not reset.

Parameters

Parameter	Description
cols (in)	number of stored columns

operator[]

double operator[](size_t i) const

Parameters

Parameter	Description
i (in)	index into the column-major value block (row i % 3 of stored column i / 3)

Returns the value.

double &operator[](size_t i)

Parameters

Parameter	Description
i (in)	index into the column-major value block (row i % 3 of stored column i / 3)

Returns reference to the value.

operator()

double operator()(size_t i, size_t j) const

Parameters

Parameter	Description
i (in)	row (0 to 2)
j (in)	stored column (full-Jacobian column idx[j])

Returns the value.

double &operator()(size_t i, size_t j)

Parameters

Parameter	Description
i (in)	row (0 to 2)
j (in)	stored column (full-Jacobian column idx[j])

Returns reference to the value.

operator*=

void operator*=(double val)

Multiplies all stored values by val.

Parameters

Parameter	Description
val (in)	scale factor

ptr

double *ptr()

Returns pointer to the first stored value.

const double *ptr() const

Returns pointer to the first stored value.

data

double *data()

Returns pointer to the first stored value (Eigen-compatible name).

const double *data() const

Returns pointer to the first stored value (Eigen-compatible name).

e

Eigen::Map<Eigen::Matrix<double, -1, -1>> e()

Returns Eigen map of the 3 x size value block.

const Eigen::Map<const Eigen::Matrix<double, -1, -1>> e() const

Returns read-only Eigen map of the 3 x size value block.

fillSub

template <size_t j, size_t k>
void fillSub(size_t startRow, size_t startColumn, const Mat<j, k> &in)

Copies in into the value block. Bounds are checked only in debug builds.

Parameters

Parameter	Description
startRow (in)	row of the block's top-left element
startColumn (in)	stored column of the block's top-left element
in (in)	block values

operator=

SparseJacobian &operator=(const SparseJacobian &rhs)

Deep copy of the size, values and indices. The storage only grows, to at least rhs.capacity columns: the articulated system sizes its Jacobians for every degree of freedom and fills them in place, so a copy (a checkpoint restore, for one) must keep that room. Grown storage is zeroed, as resize() leaves it.

Parameters

Parameter	Description
rhs (in)	Jacobian to copy

Returns *this

Public attributes

size

size_t size

Number of stored (non-zero) columns.

capacity

size_t capacity

Number of allocated columns.

v

MatDyn v

Value storage, 3 x capacity and column-major; only the first size columns are meaningful.

idx

std::vector<size_t> idx

idx[i] is the full-Jacobian column (generalized-velocity index) of stored column i.