[SimpleOptimizationTask Inherit] More...

#include <simple_optimization_task.h>

Inheritance diagram for stomp_core_examples::SimpleOptimizationTask:

Public Member Functions
bool	computeCosts (const Eigen::MatrixXd &parameters, std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, Eigen::VectorXd &costs, bool &validity) override
	computes the state costs as a function of the distance from the bias parameters More...

bool	computeNoisyCosts (const Eigen::MatrixXd &parameters, std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, int rollout_number, Eigen::VectorXd &costs, bool &validity) override
	computes the state costs as a function of the distance from the bias parameters More...

bool	filterParameterUpdates (std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, const Eigen::MatrixXd &parameters, Eigen::MatrixXd &updates) override
	Filters the given parameters which is applied after the update. It could be used for clipping of joint limits or projecting into the null space of the Jacobian. More...

bool	generateNoisyParameters (const Eigen::MatrixXd &parameters, std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, int rollout_number, Eigen::MatrixXd &parameters_noise, Eigen::MatrixXd &noise) override
	[SimpleOptimizationTask Inherit] More...

	SimpleOptimizationTask (const Eigen::MatrixXd &parameters_bias, const std::vector< double > &bias_thresholds, const std::vector< double > &std_dev)
	A simple task for demonstrating how to use Stomp. More...

Public Member Functions inherited from stomp_core::Task
virtual void	done (bool success, int total_iterations, double final_cost, const Eigen::MatrixXd &parameters)
	Called by Stomp at the end of the optimization process. More...

virtual bool	filterNoisyParameters (std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, int rollout_number, Eigen::MatrixXd &parameters, bool &filtered)
	Filters the given noisy parameters which is applied after noisy trajectory generation. It could be used for clipping of joint limits or projecting into the null space of the Jacobian. More...

virtual void	postIteration (std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, double cost, const Eigen::MatrixXd &parameters)
	Called by STOMP at the end of each iteration. More...

Protected Member Functions
bool	smoothParameterUpdates (std::size_t start_timestep, std::size_t num_timesteps, int iteration_number, Eigen::MatrixXd &updates)
	Perform a smooth update given a noisy update. More...

Protected Attributes
std::vector< double >	bias_thresholds_

Eigen::MatrixXd	parameters_bias_

Eigen::MatrixXd	smoothing_M_

std::vector< double >	std_dev_

Detailed Description

[SimpleOptimizationTask Inherit]

A dummy task for testing STOMP

Definition at line 37 of file simple_optimization_task.h.

Constructor & Destructor Documentation

stomp_core_examples::SimpleOptimizationTask::SimpleOptimizationTask	(	const Eigen::MatrixXd &	parameters_bias,
		const std::vector< double > &	bias_thresholds,
		const std::vector< double > &	std_dev
	)

inline

A simple task for demonstrating how to use Stomp.

Parameters

parameters_bias	default parameter bias used for computing cost.
bias_thresholds	threshold to determine whether two trajectories are equal
std_dev	standard deviation used for generating noisy parameters

Definition at line 46 of file simple_optimization_task.h.

Member Function Documentation

bool stomp_core_examples::SimpleOptimizationTask::computeCosts	(	const Eigen::MatrixXd &	parameters,
		std::size_t	start_timestep,
		std::size_t	num_timesteps,
		int	iteration_number,
		Eigen::VectorXd &	costs,
		bool &	validity
	)

inlineoverridevirtual

computes the state costs as a function of the distance from the bias parameters

Parameters

parameters	A matrix [num_dimensions][num_parameters] of the policy parameters to execute
start_timestep	The start index into the 'parameters' array, usually 0.
num_timesteps	The number of elements to use from 'parameters' starting from 'start_timestep'
iteration_number	The current iteration count in the optimization loop
costs	A vector containing the state costs per timestep.
validity	Whether or not the trajectory is valid

Returns: True if cost were properly computed, otherwise false

Implements stomp_core::Task.

Definition at line 107 of file simple_optimization_task.h.

bool stomp_core_examples::SimpleOptimizationTask::computeNoisyCosts	(	const Eigen::MatrixXd &	parameters,
		std::size_t	start_timestep,
		std::size_t	num_timesteps,
		int	iteration_number,
		int	rollout_number,
		Eigen::VectorXd &	costs,
		bool &	validity
	)

inlineoverridevirtual

computes the state costs as a function of the distance from the bias parameters

Parameters

parameters	A matrix [num_dimensions][num_parameters] of the policy parameters to execute
start_timestep	The start index into the 'parameters' array, usually 0.
num_timesteps	The number of elements to use from 'parameters' starting from 'start_timestep'
iteration_number	The current iteration count in the optimization loop
rollout_number	The index of the noisy trajectory.
costs	A vector containing the state costs per timestep.
validity	Whether or not the trajectory is valid

Returns: True if cost were properly computed, otherwise false

Implements stomp_core::Task.

Definition at line 128 of file simple_optimization_task.h.

bool stomp_core_examples::SimpleOptimizationTask::filterParameterUpdates	(	std::size_t	start_timestep,
		std::size_t	num_timesteps,
		int	iteration_number,
		const Eigen::MatrixXd &	parameters,
		Eigen::MatrixXd &	updates
	)

inlineoverridevirtual

Filters the given parameters which is applied after the update. It could be used for clipping of joint limits or projecting into the null space of the Jacobian.

Parameters

start_timestep	The start index into the 'parameters' array, usually 0.
num_timesteps	The number of elements to use from 'parameters' starting from 'start_timestep'
iteration_number	The current iteration count in the optimization loop
parameters	The optimized parameters
updates	The updates to the parameters

Returns: True if successful, otherwise false

Reimplemented from stomp_core::Task.

Definition at line 172 of file simple_optimization_task.h.

bool stomp_core_examples::SimpleOptimizationTask::generateNoisyParameters	(	const Eigen::MatrixXd &	parameters,
		std::size_t	start_timestep,
		std::size_t	num_timesteps,
		int	iteration_number,
		int	rollout_number,
		Eigen::MatrixXd &	parameters_noise,
		Eigen::MatrixXd &	noise
	)

inlineoverridevirtual

[SimpleOptimizationTask Inherit]

Generates a noisy trajectory from the parameters.

Parameters

parameters	A matrix [num_dimensions][num_parameters] of the current optimized parameters
start_timestep	The start index into the 'parameters' array, usually 0.
num_timesteps	The number of elements to use from 'parameters' starting from 'start_timestep'
iteration_number	The current iteration count in the optimization loop
rollout_number	The index of the noisy trajectory.
parameters_noise	The parameters + noise
noise	The noise applied to the parameters

Returns: True if cost were properly computed, otherwise false

Implements stomp_core::Task.

Definition at line 73 of file simple_optimization_task.h.

bool stomp_core_examples::SimpleOptimizationTask::smoothParameterUpdates	(	std::size_t	start_timestep,
		std::size_t	num_timesteps,
		int	iteration_number,
		Eigen::MatrixXd &	updates
	)