[[TOC]]

title: promp::ProMP summary: Class that implements a multi-dimensional Probabilistic Motion Primitive. References:

promp::ProMP

Class that implements a multi-dimensional Probabilistic Motion Primitive. References: More...

#include <promp.hpp>

Public Functions

	Name
	ProMP(const std::vector< Trajectory > & data, int num_bf, double std_bf =-1) constructor: The constructor will parameterize a phase vector, and compute the basis function matrix ( \Psi ) for all phase steps. Then, for each demonstration within the std::vector `data`, it will estimate a. vector of basis functions' weights ( w_i ). Lastly, it fits a gaussian over all the weight vectors to obtain ( \mu_w ), and ( \Sigma_w )
	ProMP(const TrajectoryGroup & data, int num_bf, double std_bf =-1) constructor: The constructor will parameterize a phase vector, and compute the basis function matrix ( \Psi ) for all phase steps. Then, for each demonstration within the std::vector `data`, it will estimate a. vector of basis functions' weights ( w_i ). Lastly, it fits a gaussian over all the weight vectors to obtain ( \mu_w ), and ( \Sigma_w )
	ProMP(const Eigen::VectorXd & w, const Eigen::MatrixXd & cov_w, double std_bf, int n_sample, size_t dims, double time_mod =1.0) This is an alternate constructor that uses prelearned weights, covariance and number of samples in the trajectory.
const Eigen::MatrixXd &	get_basis_function() const Gets the basis function matrix for the current trained ProMP.
Eigen::MatrixXd	generate_basis_function(const Eigen::VectorXd & phase) const generates basis functions
Eigen::VectorXd	compute_phase(size_t timesteps) const maps time vector into a phase vector give a desired number of timesteps
const Eigen::VectorXd &	get_phase() const Gets the phase vector.
const Eigen::VectorXd &	get_weights() const Gets the vector of the mean of all weights ( \mu_w ).
const Eigen::MatrixXd &	get_covariance() const Gets the co-variance matrix ( \Sigma_w ).
int	get_n_samples() const Gets the number of samples for the trajectory ( \s_ ).
int	get_traj_length() const Gets the trajectory length.
double	get_std_bf() const Gets the std deviation ( \std_bf ) of the ProMP.
size_t	get_dims() const return the number of dimensions represented and generated by the promp
double	get_mean_demo_time_mod() const Gets the average time modulation in the demonstrations used to train the ProMP.
void	set_ridge_factor(double ridge_factor) Set the ridge factor value which condition the pphi inverse. Helps against singularities.
Eigen::MatrixXd	generate_trajectory() const Generates MEAN trajectory based on current weights distribution and rbf.
Eigen::MatrixXd	generate_trajectory(size_t req_num_steps) const Generates MEAN trajectory based on current weights distribution and a required number of steps.
Eigen::MatrixXd	generate_trajectory_with_speed(double req_phase_speed) const Generates MEAN trajectory based on current weights distributions and a required phase speed ( \dot{z}_t ).
Eigen::MatrixXd	generate_trajectory_at(const Eigen::VectorXd & phase) const
void	condition_goal(const Eigen::VectorXd & goal, const Eigen::MatrixXd & std) set desired goal/end point for trajectory
void	condition_start(const Eigen::VectorXd & start, const Eigen::MatrixXd & std) set desired start/initial point for trajectory
Eigen::MatrixXd	gen_traj_std_dev(size_t req_num_steps =0) const Generates standard deviation vector with the standard deviation for every time step, with a certain number of time steps.
std::vector< Eigen::MatrixXd >	generate_trajectory_covariance(size_t req_num_steps =0) const Generates step covariance matrices for for each step of the trajectory, with a certain number of time steps.
void	condition_via_points(const std::vector< std::tuple< int, Eigen::VectorXd, Eigen::MatrixXd > > & via_points) Conditions all via points registered in 'viaPoints_'. It updates _mean_w and _cov_w and clear _via_points vector.
void	condition_via_point(int t, const Eigen::VectorXd & via_point, const Eigen::MatrixXd & std) Conditions all via points registered in 'viaPoints_'. It updates _mean_w and _cov_w and clear _via_points vector.
Eigen::VectorXd	get_upper_weights(double K)
Eigen::VectorXd	get_lower_weights(double K)
double	phase_speed_from_steps(int steps) generate the speed from a number of steps

Friends

	Name
std::ostream &	operator<<(std::ostream & out, const ProMP & mp) Overloads << operator in order to print the weights of a given promp.

Detailed Description

class promp::ProMP;

Class that implements a multi-dimensional Probabilistic Motion Primitive. References:

Paraschos A, Daniel C, Peters J, Neumann G. Probabilistic movement primitives. Advances in neural information processing systems. 2013. [pdf]
Paraschos A, Daniel C, Peters J, Neumann G. Using probabilistic movement primitives in robotics. Autonomous Robots. 2018 Mar;42(3):529-51. [pdf].

Public Functions Documentation

function ProMP

inline ProMP(
    const std::vector< Trajectory > & data,
    int num_bf,
    double std_bf =-1
)

constructor: The constructor will parameterize a phase vector, and compute the basis function matrix ( \Psi ) for all phase steps. Then, for each demonstration within the std::vector data, it will estimate a. vector of basis functions' weights ( w_i ). Lastly, it fits a gaussian over all the weight vectors to obtain ( \mu_w ), and ( \Sigma_w )

Parameters:

data vector of trajectories. All the trajectories MUST have the same length and same dimension.
num_bf number of basis functions
std_bf standard deviation; this is set automatically to 1.0 / (n_rbf*n_rbf) if std_bf <= 0

function ProMP

inline ProMP(
    const TrajectoryGroup & data,
    int num_bf,
    double std_bf =-1
)

Parameters:

data vector of trajectories. All the trajectories MUST have the same length and same dimension.
num_bf number of basis functions
std_bf standard deviation; this is set automatically to 1.0 / (n_rbf*n_rbf) if std_bf <= 0

function ProMP

ProMP(
    const Eigen::VectorXd & w,
    const Eigen::MatrixXd & cov_w,
    double std_bf,
    int n_sample,
    size_t dims,
    double time_mod =1.0
)

This is an alternate constructor that uses prelearned weights, covariance and number of samples in the trajectory.

Parameters:

w The mean of weights' distribution
cov_w The co-variance of weights' distribution
std_bf The standard deviation of each basis function
n_sample The number of samples required for the trajectory

function get_basis_function

const Eigen::MatrixXd & get_basis_function() const

Gets the basis function matrix for the current trained ProMP.

Return: The basis function ( \Psi ).

function generate_basis_function

Eigen::MatrixXd generate_basis_function(
    const Eigen::VectorXd & phase
) const

generates basis functions

Parameters:

phase

Return: matrix of basis functions

function compute_phase

Eigen::VectorXd compute_phase(
    size_t timesteps
) const

maps time vector into a phase vector give a desired number of timesteps

function get_phase

const Eigen::VectorXd & get_phase() const

Gets the phase vector.

Return: The phase vector.

function get_weights

const Eigen::VectorXd & get_weights() const

Gets the vector of the mean of all weights ( \mu_w ).

Return: The weights.

function get_covariance

const Eigen::MatrixXd & get_covariance() const

Gets the co-variance matrix ( \Sigma_w ).

Return: The co-variance matrix.

function get_n_samples

int get_n_samples() const

Gets the number of samples for the trajectory ( \s_ ).

Return: The number of samples for the trajectory

function get_traj_length

int get_traj_length() const

Gets the trajectory length.

Return: The trajectory length

function get_std_bf

double get_std_bf() const

Gets the std deviation ( \std_bf ) of the ProMP.

Return: The std deviation.

function get_dims

size_t get_dims() const

return the number of dimensions represented and generated by the promp

Return: Number of dimensions of the generated function

function get_mean_demo_time_mod

double get_mean_demo_time_mod() const

Gets the average time modulation in the demonstrations used to train the ProMP.

Return: m_alpha_ The avergae demo time modulation

function set_ridge_factor

void set_ridge_factor(
    double ridge_factor
)

Set the ridge factor value which condition the pphi inverse. Helps against singularities.

Parameters:

ridge_factor

function generate_trajectory

Eigen::MatrixXd generate_trajectory() const

Generates MEAN trajectory based on current weights distribution and rbf.

Return: Mean trajectory

function generate_trajectory

Eigen::MatrixXd generate_trajectory(
    size_t req_num_steps
) const

Generates MEAN trajectory based on current weights distribution and a required number of steps.

Parameters:

req_num_steps The requested number of steps for trajectory

Return: Mean trajectory

function generate_trajectory_with_speed

Eigen::MatrixXd generate_trajectory_with_speed(
    double req_phase_speed
) const

Generates MEAN trajectory based on current weights distributions and a required phase speed ( \dot{z}_t ).

Parameters:

req_phase_speed The request phase speed. ** To play trajectory at orignal speed set ( \dot{z}_t = 1.0 )**.

Return: { description_of_the_return_value }

function generate_trajectory_at

Eigen::MatrixXd generate_trajectory_at(
    const Eigen::VectorXd & phase
) const

function condition_goal

void condition_goal(
    const Eigen::VectorXd & goal,
    const Eigen::MatrixXd & std
)

set desired goal/end point for trajectory

Parameters:

goal desired value at end
std desired standard deviation. ** typically around ( 10^{-6}) for accuracy **

Return:

function condition_start

void condition_start(
    const Eigen::VectorXd & start,
    const Eigen::MatrixXd & std
)

set desired start/initial point for trajectory

Parameters:

start desired value at start
std desired standard deviation.** typically around ( 10^{-6}) for accuracy **

Return:

function gen_traj_std_dev

Eigen::MatrixXd gen_traj_std_dev(
    size_t req_num_steps =0
) const

Generates standard deviation vector with the standard deviation for every time step, with a certain number of time steps.

Parameters:

req_num_steps The requested number steps. if <= 0 (default) use the internal phase parametrization

Return: Standard Deviation Vector ( DIAG( \Sigma ) )

function generate_trajectory_covariance

std::vector< Eigen::MatrixXd > generate_trajectory_covariance(
    size_t req_num_steps =0
) const

Generates step covariance matrices for for each step of the trajectory, with a certain number of time steps.

Parameters:

req_num_steps The requested number steps. if <= 0 (default) use the internal phase parametrization

Return: std::vector containg and entry for each step covariance matrix

equivalent to (a little slower)

function condition_via_points

void condition_via_points(
    const std::vector< std::tuple< int, Eigen::VectorXd, Eigen::MatrixXd > > & via_points
)

Conditions all via points registered in 'viaPoints_'. It updates _mean_w and _cov_w and clear _via_points vector.

function condition_via_point

void condition_via_point(
    int t,
    const Eigen::VectorXd & via_point,
    const Eigen::MatrixXd & std
)

Conditions all via points registered in 'viaPoints_'. It updates _mean_w and _cov_w and clear _via_points vector.

Parameters:

t time at which via point is to be added (between 0 and LAST TIME STEP)
via_point desired value at via point
std desired standard deviation.** typically around ( 10^{-6}) for accuracy **

Return:

Todo: Use phase ( z_t) instead of time step t.

set via point for trajectory

function get_upper_weights

Eigen::VectorXd get_upper_weights(
    double K
)

function get_lower_weights

Eigen::VectorXd get_lower_weights(
    double K
)

function phase_speed_from_steps

inline double phase_speed_from_steps(
    int steps
)

generate the speed from a number of steps

Parameters:

steps number of steps required in the trajectory

Return: double the phase speed to obtain the desired number of steps in the trajectory

Friends

friend operator<<

friend std::ostream & operator<<(
    std::ostream & out,

    const ProMP & mp
);

Overloads << operator in order to print the weights of a given promp.

Parameters:

out The out
mp THIS promp

Updated on 2022-01-20 at 11:06:17 +0100--- title: promp::Trajectory summary: Class that represents a multidimensional trajectory. A trajectory is described by the values at each timestep and the speed parameter. speed indicates how the trajectory has been modulated, for example speed=2 means that the original trajectory had twice the timesteps.

promp::Trajectory

Class that represents a multidimensional trajectory. A trajectory is described by the values at each timestep and the speed parameter. speed indicates how the trajectory has been modulated, for example speed=2 means that the original trajectory had twice the timesteps.

#include <trajectory.hpp>

Public Functions

	Name
	Trajectory() =default
	Trajectory(const Eigen::MatrixXd & data, double speed =1.0)
virtual	~Trajectory() =default
size_t	dims() const
size_t	timesteps() const
double	speed() const
const Eigen::MatrixXd &	matrix() const
Trajectory	sub_trajectory(size_t dim) const
Trajectory	sub_trajectory(const std::vector< size_t > & dims) const
void	modulate_in_place(size_t timesteps, bool fast =true)
Trajectory	modulate(size_t steps, bool fast =true) const
double	distance(const Trajectory & other, bool modulate =false) const
double	infer_speed(const Eigen::MatrixXd & obs_traj, double lb, double ub, size_t steps) const

Public Functions Documentation

function Trajectory

Trajectory() =default

\brief default constructor. Build empty trajectory.

function Trajectory

explicit Trajectory(
    const Eigen::MatrixXd & data,
    double speed =1.0
)

\brief constructor that build a trajectory starting from data and speed \param data Eigen::Matrix containing the raw data, each column is a different dof \param speed speed of the original trajectory (time-scale factor: e.g., 2.0 to go from 200 time-steps to 100 time-steps)

function ~Trajectory

virtual ~Trajectory() =default

function dims

inline size_t dims() const

\brief return number of dimensions of the trajectory

function timesteps

inline size_t timesteps() const

\brief return number of timesteps in the trajectory

function speed

inline double speed() const

\brief return the trajectory' speed

function matrix

inline const Eigen::MatrixXd & matrix() const

\brief return the raw data as Eigen::Matrix

function sub_trajectory

Trajectory sub_trajectory(
    size_t dim
) const

\brief return monodimensional trajectory from the selected dimension \param dim dimension used to create the returned trajectory

function sub_trajectory

Trajectory sub_trajectory(
    const std::vector< size_t > & dims
) const

\brief return trajectory using data from the selected dimensions \param dim list of dimensions used to create the returned trajectory

function modulate_in_place

void modulate_in_place(
    size_t timesteps,
    bool fast =true
)

\brief modulate the trajectory to the desired number of timesteps Adjust speed according to speed = this->speed() * this->timesteps() / timesteps \param timesteps desired number of steps in the trajectory

function modulate

Trajectory modulate(
    size_t steps,
    bool fast =true
) const

\brief create a new modulated trajectory with the desired number of timesteps Adjust its speed according to speed = this->speed() * this->timesteps() / timesteps \param timesteps desired number of steps in the trajectory

function distance

double distance(
    const Trajectory & other,
    bool modulate =false
) const

\brief compute the Euclidean distance between this and a second trajectory \param other trajectory used to compute the distance with \param modulate if false the distance is computed using data until the smaller trajectory lenght, if true the other trajectory is modulated to this trajectory length before computing the distance

function infer_speed

double infer_speed(
    const Eigen::MatrixXd & obs_traj,
    double lb,
    double ub,
    size_t steps
) const

\brief infer the speed of a trajectory starting from the raw data \param obs_traj data from whoch speed is inferred, comparing it to this trajectory \param lb lower bound for inferred speed \param ub upper bound for inferred speed \param steps number of speeds to be tested (linspace(lb, ub, steps))

Updated on 2022-01-20 at 11:06:17 +0100--- title: promp::TrajectoryGroup summary: Class for handling data from multiple trajectories. Data from each trajectory should be stored in an individual file, files can be generated from serializer or be in .csv format. Each column within .csv file represents a trajectory. All .csv files must have trajectories(columns) in the same sequence (column-wise)

promp::TrajectoryGroup

Class for handling data from multiple trajectories. Data from each trajectory should be stored in an individual file, files can be generated from serializer or be in .csv format. Each column within .csv file represents a trajectory. All .csv files must have trajectories(columns) in the same sequence (column-wise)

#include <trajectory_group.hpp>

Public Functions

	Name
	TrajectoryGroup() =default
void	load_trajectories(const std::vector< std::string > & files, const std::vector< size_t > & index)
void	load_csv_trajectories(const std::vector< std::string > & files, const std::vector< size_t > & index, char sep =',', bool skip_header =false)
void	load_csv_trajectories(const std::vector< std::string > & files, const std::vector< std::string > & cols, char sep =',')
size_t	normalize_length()
void	normalize_length(size_t len)
std::pair< Eigen::VectorXd, Eigen::VectorXd >	standardize_dims()
const std::vector< Trajectory > &	trajectories() const

Public Functions Documentation

function TrajectoryGroup

TrajectoryGroup() =default

\brief default constructor: create an empty trajectory group

function load_trajectories

void load_trajectories(
    const std::vector< std::string > & files,
    const std::vector< size_t > & index
)

\brief load trajectories from list of files (formatted as generated from io/serializer). \param files of files. \param index list of indexes representing dofs to keep.

function load_csv_trajectories

void load_csv_trajectories(
    const std::vector< std::string > & files,
    const std::vector< size_t > & index,
    char sep =',',
    bool skip_header =false
)

\brief load trajectories from list of .csv files. \param files of files. \param index list of indexes representing dofs to keep. \param sep values separator. \param skip_header if true skip first line.

function load_csv_trajectories

void load_csv_trajectories(
    const std::vector< std::string > & files,
    const std::vector< std::string > & cols,
    char sep =','
)

\brief load trajectories from list of .csv files. \param files of files. \param cols list of columns (dofs) to keep. \param sep values separator.

function normalize_length

size_t normalize_length()

\brief Normalize all trajectories to the mean length (number of timesteps)

function normalize_length

void normalize_length(
    size_t len
)

\brief Normalize all trajectories to the same desired length \param len desired length

function standardize_dims

std::pair< Eigen::VectorXd, Eigen::VectorXd > standardize_dims()

\brief standardize each dof among the trajectories

function trajectories

inline const std::vector< Trajectory > & trajectories() const

\brief return the vector of trajectories

Updated on 2022-01-20 at 11:06:17 +0100--- title: promp::io::CSVReader summary: Class reading data from .csv files.

promp::io::CSVReader

Class reading data from .csv files.

#include <csv_reader.hpp>

Public Functions

	Name
	CSVReader(const std::string & filename, char delm =',', bool skip_header =false)
Eigen::MatrixXd	get_data()
Eigen::VectorXd	get_data_dof(const std::string & dof, const Eigen::MatrixXd & data)
Eigen::MatrixXd	get_data_dofs(const std::vector< std::string > & dofs, const Eigen::MatrixXd & data)
Eigen::VectorXd	get_data_dof(size_t dof, const Eigen::MatrixXd & data)
Eigen::MatrixXd	get_data_dofs(const std::vector< size_t > & dofs, const Eigen::MatrixXd & data)

Public Functions Documentation

function CSVReader

CSVReader(
    const std::string & filename,
    char delm =',',
    bool skip_header =false
)

function get_data

Eigen::MatrixXd get_data()

function get_data_dof

Eigen::VectorXd get_data_dof(
    const std::string & dof,
    const Eigen::MatrixXd & data
)

function get_data_dofs

Eigen::MatrixXd get_data_dofs(
    const std::vector< std::string > & dofs,
    const Eigen::MatrixXd & data
)

function get_data_dof

static Eigen::VectorXd get_data_dof(
    size_t dof,
    const Eigen::MatrixXd & data
)

function get_data_dofs

static Eigen::MatrixXd get_data_dofs(
    const std::vector< size_t > & dofs,
    const Eigen::MatrixXd & data
)

Updated on 2022-01-20 at 11:06:17 +0100

Uh oh!

FilesExpand file tree

api.md

Latest commit

History

api.md

File metadata and controls

[[TOC]]

promp::ProMP

Public Functions

Friends

Detailed Description

Public Functions Documentation

function ProMP

function ProMP

function ProMP

function get_basis_function

function generate_basis_function

function compute_phase

function get_phase

function get_weights

function get_covariance

function get_n_samples

function get_traj_length

function get_std_bf

function get_dims

function get_mean_demo_time_mod

function set_ridge_factor

function generate_trajectory

function generate_trajectory

function generate_trajectory_with_speed

function generate_trajectory_at

function condition_goal

function condition_start

function gen_traj_std_dev

function generate_trajectory_covariance

function condition_via_points

function condition_via_point

function get_upper_weights

function get_lower_weights

function phase_speed_from_steps

Friends

friend operator<<

promp::Trajectory

Public Functions

Public Functions Documentation

function Trajectory

function Trajectory

function ~Trajectory

function dims

function timesteps

function speed

function matrix

function sub_trajectory

function sub_trajectory

function modulate_in_place

function modulate

function distance

function infer_speed

promp::TrajectoryGroup

Public Functions

Public Functions Documentation

function TrajectoryGroup

function load_trajectories

function load_csv_trajectories

function load_csv_trajectories

function normalize_length

function normalize_length

function standardize_dims

function trajectories

promp::io::CSVReader

Public Functions

Public Functions Documentation

function CSVReader

function get_data

function get_data_dof

function get_data_dofs

function get_data_dof

function get_data_dofs