Submodules#

robosuite.controllers.interpolators.base_interpolator module#

class robosuite.controllers.interpolators.base_interpolator.Interpolator#

Bases: object

General interpolator interface.

abstract get_interpolated_goal()#

Provides the next step in interpolation given the remaining steps.

Returns

Next interpolated step

Return type

np.array

robosuite.controllers.interpolators.linear_interpolator module#

class robosuite.controllers.interpolators.linear_interpolator.LinearInterpolator(ndim, controller_freq, policy_freq, ramp_ratio=0.2, use_delta_goal=False, ori_interpolate=None)#

Bases: robosuite.controllers.interpolators.base_interpolator.Interpolator

Simple class for implementing a linear interpolator.

Abstracted to interpolate n-dimensions

Parameters

• ndim (int) – Number of dimensions to interpolate

• controller_freq (float) – Frequency (Hz) of the controller

• policy_freq (float) – Frequency (Hz) of the policy model

• ramp_ratio (float) –

  Percentage of interpolation timesteps across which we will interpolate to a goal position.

  Note

  Num total interpolation steps will be equal to np.floor(ramp_ratio * controller_freq / policy_freq) i.e.: how many controller steps we get per action space update

• ori_interpolate (None or str) –

  If set, assumes that we are interpolating angles (orientation) Specified string determines assumed type of input:

  ’euler’: Euler orientation inputs ‘quat’: Quaternion inputs

get_interpolated_goal()#

Provides the next step in interpolation given the remaining steps.

NOTE: If this interpolator is for orientation, it is assumed to be receiving either euler angles or quaternions

Returns

Next position in the interpolated trajectory

Return type

np.array

set_goal(goal)#

Takes a requested (absolute) goal and updates internal parameters for next interpolation step

Parameters

np.array – Requested goal (absolute value). Should be same dimension as self.dim

set_states(dim=None, ori=None)#

Updates self.dim and self.ori_interpolate.

Initializes self.start and self.goal with correct dimensions.

Parameters

• ndim (None or int) – Number of dimensions to interpolate

• ori_interpolate (None or str) –

  If set, assumes that we are interpolating angles (orientation) Specified string determines assumed type of input:

  ’euler’: Euler orientation inputs ‘quat’: Quaternion inputs

Module contents#