0
$\begingroup$

Rosanswers logo

Hi,

I wanted to use the gazebo_ros_bumpers and retrieve the force vector for each contact. However, when drawing the vectors, it appears to me they are not in a correct reference frame.

I think, according to the comments in the source file (gazebo_ros_bumper.cpp), that the contact data (force, normal, point) are all given in the "user requested" frame of reference. I am using diamondback but this seems not to have changed in trunk version of the file. see :

I can see some coordinate frame change for the data in the source code but I cannot understand why some are using RotateVector (example force and torque)

// rotate into user specified frame. frame_rot is identity if world is used.
   Vector3 force = frame_rot.RotateVector( Vector3(contact.forces[k]. body1Force.x, 

and some others are converted by RotateVectorReverse (position and normal).

// rotate normal into user specified frame. frame_rot is identity if world is used.
   Vector3 normal = frame_rot.RotateVectorReverse( Vector3(contact.normals[k].x,

with frame_rot the rotation transformation from local to global frame. To what I could read, forces and position are given by ODE in global coordinate system, so forces would need the same reverse transformation that position and normal require, but I may have missed something.

Changing both transformation to reverse seemed to show better force orientation when displayed but I cannot be totally sure as contact points are "varying a lot" due to bouncy contacts.

If any one has comments on this, it sure could help confirm the issue (or not), and I could then file a bug if required.


Originally posted by GuiHome on ROS Answers with karma: 242 on 2011-08-24

Post score: 2


Original comments

Comment by GuiHome on 2011-09-14:
As mentionned in the question, I am running diamondback. I cannot switch to electric because the Shadow Hand dos not support it yet. But I verified the trunk version at the time of the question, and the code was the same for this part. In electric the code is still the same. Will try it again soon.

Comment by hsu on 2011-09-02:
Hi, are you running on diamondback or electric?

$\endgroup$

1 Answer 1

0
$\begingroup$

Rosanswers logo

Good point. I added a test world to check simple rotation, and plotting

rxplot -p 5 "box_bumper/state/states[0]/total_wrench/force/x" "box_bumper/state/states[0]/total_wrench/force/y" "box_bumper/state/states[0]/total_wrench/force/z" "box1_bumper/state/states[0]/total_wrench/force/x" "box1_bumper/state/states[0]/total_wrench/force/y" "box1_bumper/state/states[0]/total_wrench/force/z"

to found that RotateVectorReverse() appears to be the correct call to make in all instances. Please test and let me know if you conclude similarly? Thanks.


Originally posted by hsu with karma: 5780 on 2011-09-15

This answer was ACCEPTED on the original site

Post score: 0


Original comments

Comment by hsu on 2011-09-25:
great! thanks for testing. I'll make a release soon.

Comment by GuiHome on 2011-09-25:
Test world is perfect. Without modif, rotating box1 shows wrong force on x & y. With Reverse everywhere, we get correct indication on each axis (positive if considered axis is upward, negative if downward). Other box keeps force positive on Z since bumper is on world frame. My patch was OK ! Thanks.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.