`bolero.representation`.DMPBehavior¶

class bolero.representation.DMPBehavior(execution_time=1.0, dt=0.01, n_features=50, configuration_file=None)[source]¶

Dynamical Movement Primitive.

Can be used to optimize the weights of a DMP with a black box optimizer. This is a wrapper for the optional DMP module of bolero. Only the weights of the DMP will be optimized. To optimize meta-parameters like the goal or the goal velocity, you have to implement your own wrapper. This can be a subclass of this wrapper that only overrides the methods that provide access to the parameters.

An object can be created either by passing a configuration file or the specification of a DMP. A DMP configuration file describes all parameters of the DMP model and it is not recommended to generate it manually.

Parameters:

Parameters:	execution_time : float, optional (default: 1) Execution time of the DMP in seconds. dt : float, optional (default: 0.01) Time between successive steps in seconds. n_features : int, optional (default: 50) Number of RBF features for each dimension of the DMP. configuration_file : string, optional (default: None) Name of a configuration file that should be used to initialize the DMP. If it is set all other arguments will be ignored.

execution_time : float, optional (default: 1): Execution time of the DMP in seconds.
dt : float, optional (default: 0.01): Time between successive steps in seconds.
n_features : int, optional (default: 50): Number of RBF features for each dimension of the DMP.
configuration_file : string, optional (default: None): Name of a configuration file that should be used to initialize the DMP. If it is set all other arguments will be ignored.

__init__(execution_time=1.0, dt=0.01, n_features=50, configuration_file=None)[source]¶

can_step()[source]¶

Returns if step() can be called again.

Note that calling step() after this function returns False will not result in an error. The velocity and acceleration will be set to 0 and we hold the last position instead.

Returns:	can_step : bool Can we call step() again?

get_args()¶

Get parameters for this estimator.

Returns:	params : mapping of string to any Parameter names mapped to their values.

get_n_params()[source]¶

Get number of weights.

Returns:	n_params : int Number of DMP weights

get_outputs(outputs)[source]¶

Get outputs of the last step.

Parameters:	outputs : array-like, shape = (3 * n_task_dims,) Contains positions, velocities and accelerations in that order. Each type is stored contiguously, i.e. for n_task_dims=2 the order would be: xxvvaa (x: position, v: velocity, a: acceleration).

get_params()[source]¶

Get current weights.

Returns:	params : array-like, shape = (n_params,) Current weights

imitate(X, Xd=None, Xdd=None, alpha=0.0, allow_final_velocity=True)[source]¶

Learn weights of the DMP from demonstrations.

Parameters:

Parameters:	X : array, shape (n_task_dims, n_steps, n_demos) The demonstrated trajectories to be imitated. Xd : array, shape (n_task_dims, n_steps, n_demos), optional Velocities of the demonstrated trajectories. Xdd : array, shape (n_task_dims, n_steps, n_demos), optional Accelerations of the demonstrated trajectories. alpha : float >= 0, optional (default: 0) The ridge parameter of linear regression. Small positive values of alpha improve the conditioning of the problem and reduce the variance of the estimates. allow_final_velocity : bool, optional (default: True) Allow the final velocity to be greater than 0

X : array, shape (n_task_dims, n_steps, n_demos): The demonstrated trajectories to be imitated.
Xd : array, shape (n_task_dims, n_steps, n_demos), optional: Velocities of the demonstrated trajectories.
Xdd : array, shape (n_task_dims, n_steps, n_demos), optional: Accelerations of the demonstrated trajectories.
alpha : float >= 0, optional (default: 0): The ridge parameter of linear regression. Small positive values of alpha improve the conditioning of the problem and reduce the variance of the estimates.
allow_final_velocity : bool, optional (default: True): Allow the final velocity to be greater than 0

init(n_inputs, n_outputs)[source]¶

Initialize the behavior.

Parameters:	n_inputs : int number of inputs n_outputs : int number of outputs

load_config(filename)[source]¶

Load DMP configuration.

Parameters:	filename : string Name of YAML file

reset()[source]¶: Reset DMP.

save(dmp, filename)¶

Save DMP model.

Parameters:	dmp : object DMP filename : string Name of YAML file

save_config(filename)[source]¶

Save DMP configuration.

Parameters:	filename : string Name of YAML file

set_inputs(inputs)[source]¶

Set input for the next step.

In case the start position (x0) has not been set as a meta-parameter we take the first position as x0.

Parameters:	inputs : array-like, shape = (3 * n_task_dims,) Contains positions, velocities and accelerations in that order. Each type is stored contiguously, i.e. for n_task_dims=2 the order would be: xxvvaa (x: position, v: velocity, a: acceleration).

set_meta_parameters(keys, meta_parameters)[source]¶

Set DMP meta parameters.

Permitted meta-parameters:

x0 : array: Initial position
g : array: Goal
gd : array: Velocity at the goal
execution_time : float: New execution time

Parameters:	keys : list of string names of meta-parameters meta_parameters : list of float values of meta-parameters

set_params(params)[source]¶

Set new weights.

Parameters:	params : array-like, shape = (n_params,) New weights

step()[source]¶: Compute desired position, velocity and acceleration.

trajectory()[source]¶

Generate trajectory represented by the DMP in open loop.

The function can be used for debugging purposes.

Returns:	X : array, shape (n_steps, n_task_dims) Positions Xd : array, shape (n_steps, n_task_dims) Velocities Xdd : array, shape (n_steps, n_task_dims) Accelerations

bolero.representation.DMPBehavior¶

`bolero.representation`.DMPBehavior¶