Position, orientation and frame tasks#
PositionTask and OrientationTask can be respectively
use to control the position and orientation of a particular frame on the robot.
The FrameTask is a combination of both, wrapping the two tasks in a single one.
Position task#
Position tasks can be initialized and updated this way:
# Creating a position task, arguments are the frame name and goal (world) position
effector_position = solver.add_position_task("effector", np.array([0., 0., 0.]))
effector_position.configure("effector_position", "soft", 1.0)
# Updating the effector target in the world
effector_position.target_world = np.array([0.1, -0.5, 0.])
Orientation task#
Orientation tasks can be initialized and updated this way:
# Creating an orientation task, argumanets are the frame name and goal (world) orientation (rotation matrix)
effector_orientation = solver.add_orientation_task("effector", np.eye(3))
effector_orientation.configure("effector_orientation", "soft", 1.0)
# Updating the effector orientation in the world
effector_orientation.R_world_frame = np.eye(3)
Frame task#
Initialization#
Frame task are lumping together a position and an orientation task, and can be initialized this way:
# Creating a frame task, arguments are the frame name and goal (world) pose (transformation matrix)
effector_frame = solver.add_frame_task("effector", np.eye(4))
# Configuring the frame task, the two weights are for the position and orientation tasks respectively
effector_frame.configure("effector_frame", "soft", 1.0, 1.0)
The underlying position and orientation tasks can be accessed with the position()
and orientation() methods:
effector_frame.position() # Access the position task
effector_frame.orientation() # Access the orientation task
Update#
You can update the frame task by setting the target pose:
# Updating the effector target in the world (transformation matrix)
effector_frame.T_world_frame = np.eye(4)
Relative position and orientation tasks#
The above mentionned tasks also exists in a relative version, where two frames have to be specified.
# Relative position
camera_task = solver.add_relative_position_task("trunk", "camera", np.array([0., 0., 0.5]))
# Setting the target (here, for the camera position in the trunk)
camera_task.target = np.array([0., 0., 0.4])
# Relative orientation
camera_task = solver.add_relative_orientation_task("trunk", "camera", np.eye(3))
# Setting the target (here, for the camera to trunk rotation)
camera_task.R_a_b = np.eye(3)
# Relative frame
camera_task = solver.add_relative_frame_task("trunk", "camera", np.eye(4))
# Setting the target (here, for the camera to trunk transformation)
camera_task.T_a_b = np.eye(4)
Masking axises#
In some case, you only want to assign a task for one or two axises. To that end, you can use the
axises mask for position and orientation tasks:
# The position task will only affect the z-axis (x and y will be ignored)
effector_position.mask.set_axises("z")
By default, this masking will occur in the “task” frame (the world frame for absolute tasks, and the first frame for
relative tasks). Youc can set the second argument of set_axises() to
"local" to enforce the masking to happen in the local frame.
Alternatively, you can also specify "custom" as the second argument, and provide a custom rotation matrix to
specify the axises in which the task will be applied in the R_local_world
attribute of the mask.
Example#
Here is an example of a 6-axis robot following a target trajectory, expressed as a FrameTask:
6-axis trajectory