You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What does NOT, in general, work: going to a location described by (desig:a location (reachable-for pr2) (arm ?arm) (location '(some description of the location of an object or the pose of the object))
Why it doesn't work: unless the object is small, there may be a significant distance between its center (which is, its pose) and the pose of whatever handles the object has. Further, the way in which these handles may be graspable may have a different orientation than the object pose itself.
Example: the center of a tray is not reachable for the gripper aligned to the object orientation, but a side handle is reachable from the top with the gripper pointing down.
Apparent alternatives: (to see) seems to have been intended as this, and there may be an (obj) location designator keyword for the same purpose. I didn't get these to work however.
Suggested fix: have a way to list all possible (arm grasp) pairs for an object. This is enough to query object-to-gripper transforms, therefore to get a set of poses for the robot to try to reach.
The text was updated successfully, but these errors were encountered:
What does NOT, in general, work: going to a location described by (desig:a location (reachable-for pr2) (arm ?arm) (location '(some description of the location of an object or the pose of the object))
Why it doesn't work: unless the object is small, there may be a significant distance between its center (which is, its pose) and the pose of whatever handles the object has. Further, the way in which these handles may be graspable may have a different orientation than the object pose itself.
Example: the center of a tray is not reachable for the gripper aligned to the object orientation, but a side handle is reachable from the top with the gripper pointing down.
Apparent alternatives: (to see) seems to have been intended as this, and there may be an (obj) location designator keyword for the same purpose. I didn't get these to work however.
Suggested fix: have a way to list all possible (arm grasp) pairs for an object. This is enough to query object-to-gripper transforms, therefore to get a set of poses for the robot to try to reach.
The text was updated successfully, but these errors were encountered: