Adding Dalsa camera to the URDF just for Real Robot not for the simulation

[prof-concert]

Hello All,

As there is an integration meeting between 17th-21st July 2023 we wished to integrate our Dalsa fisheye cameras to the robot for 3D Pose estimation.
For this we want following transforms to give out exact 3D Poses,

1. t_fisheye_robotBase ( transform between fisheye camera and the mobile base)
2. t_robotBase_map ( tranform between robot Base and the map)

For this I thought, it would be a good idea to add the camera to the urdf of the real robot. So that the fisheye cameras are available in the tf tree and after localization we should get the required tf's accurately.

Please let us know if further information is required. @alaurenzi @EdoardoRomiti

[ntsagarakis]

I add also @StecarGit who can provide the exact location of the base of the cameras and their mounting orientation.

[EdoardoRomiti]

Hi @prof-concert, first of all I attach here all relevant info on the cameras placement which were provided by @StecarGit:

Now what I need you to tell me, so that I can provide you the transform t_fisheye_robotBase, is the exact location where you expect the frame fisheye to be.

[EdoardoRomiti]

I attach also the datasheet of the lens:

For instance, the frame could it be placed like this? Or else where?

[prof-concert]

Hello @EdoardoRomiti

Thanks for the update, instead of the lens we want the origin of the fisheye to be located at the centre of the imaging area of the camera. As per the camera datasheet and 3D model 03-129-20166-07 GENIE NANO 1 GigE SMALL CASING CUSTOMER DRAWING.pdf could you please add the origin for fisheye at the "CENTRE OF IMAGE AREA".

[EdoardoRomiti]

Hi @prof-concert I cannot open the file you shared, it says it's corrupted. Can you share it again?

[prof-concert]

03-129-20166-07 GENIE NANO 1 GigE SMALL CASING CUSTOMER DRAWING.pdf

The link from the manufacturer https://www.teledynedalsa.com/download/27011021-0fb7-4258-81c4-079e5e997013/

[EdoardoRomiti]

I've added the two camera frames to the urdf (ADVRHumanoids/modular_hhcm@ef4f3ed).
The frames to get the t_fisheye_robotBase are called: dalsa_camera_right and dalsa_camera_left. They will be part of the TF tree.

For the t_robotBase_map instead have a look at the discussion in #10

[JakobThumm]

@prof-concert @EdoardoRomiti, the integration meeting showed that we should test the camera integration first in simulation before testing on the real robot. @prof-concert still had issues with the coordinate transforms that simulation tests could have caught. TUM will provide a human model in gazebo for that.

[EdoardoRomiti]

@JakobThumm @prof-concert Ok let me know if anything else is needed from the model side for the camera. Or if you still have issues with the TFs.

For the human model in gazebo it would be great to add it to the simulation. We can track this in #24

[prof-concert]

@EdoardoRomiti @JakobThumm I was able to correct the transforms in my code with a recorded rosbag file, recorded on 21.07.2023. But, I am not able to visualize the robot with the recorded rosbag file, @EdoardoRomiti could you tell us how could I visualize the robot model?

[prof-concert]

Created a ticket for robot visualizing in Rviz #25

[prof-concert]

@EdoardoRomiti thanks for your help in running robot visualization in Rviz. Secondly, @JakobThumm is it possible to test the safety shield with this rosbag file?