search

tork-a / rtmros_nextage

ROS-OpenRTM-based opensource robot controller software for dual-armed robot Nextage from Kawada Industries
http://wiki.ros.org/rtmros_nextage
27 stars 39 forks source link

eef coordinate axis may not be intuitive #161

Closed 130s closed 9 years ago

130s commented 9 years ago

Colors are RBG : XYZ.

While arms are facing down, x-axis faces up. eef_axis_nxo

I'm still new to humanoid but is this the common frame orientation? REP-0120: Coordinate Frames for Humanoids Robots doesn't specify this detail.

I feel PR2's hand is intuitive: x faces forward, z is up (and thus y is left-bound). eef_hand_pr2

k-okada commented 9 years ago

since there are so many variation on hands/gripper, so I believe it is not able to define common convention of gripper orientation. As for figures, you see joint/link coordinates on nextage and see tip link for PR2 robots, At this moment we do not define tip link in hrpsys robot.

◉ Kei Okada

On Fri, Apr 3, 2015 at 8:10 AM, Isaac I.Y. Saito notifications@github.com wrote:

Colors are RBG : XYZ.

While arms are facing down, x-axis faces up. [image: eef_axis_nxo] https://cloud.githubusercontent.com/assets/1840401/6975232/e244d400-d94e-11e4-9f3e-128e61d10ab9.png

I'm still new to humanoid but is this the common frame orientation? REP-0120: Coordinate Frames for Humanoids Robots http://www.ros.org/reps/rep-0120.html doesn't specify this detail.

PR2's hand is intuitive: x faces forward, z is up (and thus y is left-bound). [image: eef_hand_pr2] https://cloud.githubusercontent.com/assets/1840401/6975541/5284c5e2-d952-11e4-8461-3b8115c69107.png

— Reply to this email directly or view it on GitHub https://github.com/tork-a/rtmros_nextage/issues/161.

130s commented 9 years ago

since there are so many variation on hands/gripper, so I believe it is not able to define common convention of gripper orientation.

I see.

I'm sorry if I wasn't clear enough; my point was the "direction" of the axis end effector for NXO (I just referred to PR2 as an example). Since the hand looks like facing down at its "goInitial" pose, I thought the roll-axis being facing down toward the ground as well would be the most intuitive (that's the purpose I cited PR2, where the roll-axis is facing forward as the grippers "look like" facing toward the front too).

k-okada commented 9 years ago

Yes direction is also ambiguous. So each robot may have different rule. We define our local rule at JSK such as:

And even if we apply this rule to the nextage robot, we still have question that what is the natural grasping for this robot? As you said, this robot is designed to grasp the cup from top side, and obviously the nominal gripper directions is different from pr2. So we have no clear answer for that. May be I would choose z axis is eef to up, when the robot is initial pose, but you can choose any direction you want.

Please note that what you seen in rviz is joint direction and not eef direction and eef is not defined in wrl yet. In urdf if is defined as fixed link. And direction of joint is usually choose z to rotation axis, but wrl/collada/urdf is designed to choose any direction for rotation axis.

◉ Kei Okada

2015/04/04 17:00、Isaac I.Y. Saito notifications@github.com のメッセージ:

since there are so many variation on hands/gripper, so I believe it is not able to define common convention of gripper orientation.

I see.

I'm sorry if I wasn't clear enough; my point was the "direction" of the axis end effector for NXO (I just referred to PR2 as an example). Since the hand looks like facing down at its "goInitial" pose, I thought the roll-axis being facing down toward the ground as well would be the most intuitive (that's the purpose I cited PR2, where the roll-axis is facing forward as the grippers "look like" facing toward the front too).

— Reply to this email directly or view it on GitHub.

130s commented 9 years ago

Thank you for the clarification. I added a note for the sake of the programmers.