Skip to content

updated urdf without uploading large stl files#868

Merged
JonasFrey96 merged 21 commits intomainfrom
dev/new_urdf
Jan 7, 2025
Merged

updated urdf without uploading large stl files#868
JonasFrey96 merged 21 commits intomainfrom
dev/new_urdf

Conversation

@EKrasn
Copy link
Contributor

@EKrasn EKrasn commented Dec 23, 2024

Applied changes:

  • Adapted the existing CAD model (TC ID 195276/01) with the back antenna assembly, added vertical spacers to Antcom antenna
    image
  • Exported three STL files (see drive link) with the box assembly in CPT7 CSYS, prism and prism distancer in CPT7 CSYS and back antenna assembly in robot base CSYS
  • Adapted existing .xacro file to generate a model with the new stl files
  • edited joint location of prism to be congruent with the box base joint
  • replaced stim320_imu coordinate transform to box base coordinate transform in various files

Drive Link: https://drive.google.com/drive/folders/1FAxQYGX82fU8rkGx3wlmT8toMWMfHwGG?usp=drive_link

Open TODOs:

  • import stls into blender, recolor faces and export as .dae files
  • reduce file sizes
  • correct calibration offset of real components present in the visualisation

@JonasFrey96
Copy link
Contributor

image
Added colors - What a pain file is still very big :/

@JonasFrey96
Copy link
Contributor

image
Back part is missing the offsets for the antenna.

@JonasFrey96
Copy link
Contributor

We need to still find a way to reduce the overall filesize if the mesh -> maybe one can draw a simplified model of the box in the CAD directly -> Decimation leads to weird artifacts.

@JonasFrey96
Copy link
Contributor

image
Frame is off by quite a bit for the back antenna

@JonasFrey96
Copy link
Contributor

image
Hacked things now together:
Working error hypothesis - box_base to base does not match (most likely box_base frame is wrongly selected within CAD?)

image

image

Had to move around all model_links to make it visually look correct

@fulkast
Copy link
Contributor

fulkast commented Dec 29, 2024

I tried replicating the state of the URDF locally but failed to open the URDF, as these large files are missing (I didn't find them in the Google Drive Link either).

[ERROR] [1735506748.992289975]: Could not load resource [package://box_model/urdf/box/mesh/box_colored_units.dae]: Unable to open file "package://box_model/urdf/box/mesh/box_colored_units.dae".
[ERROR] [1735506748.992610808]: Could not load resource [package://box_model/urdf/box/mesh/gps_colored_units.dae]: Unable to open file "package://box_model/urdf/box/mesh/gps_colored_units.dae".
[ERROR] [1735506748.993023876]: Could not load resource [package://box_model/urdf/box/mesh/prism_colored_units.dae]: Unable to open file "package://box_model/urdf/box/mesh/prism_colored_units.dae".

My next step was going to check whether the origin axes of box_base coincide with the CAD depiction of the CPT7 center of navigation shown here: https://docs.novatel.com/OEM7/Content/Technical_Specs_Receiver/SPANCPT7_Mechanicals.htm#Center

Top down view of the CoN of the CPT7
image

@EKrasn
Copy link
Contributor Author

EKrasn commented Jan 3, 2025

  • I checked the location of the CPT7 CSYS in the CAD compared to the location given in the technical specifications (https://docs.novatel.com/OEM7/Content/Technical_Specs_Receiver/SPANCPT7_Mechanicals.htm#Center), in the manual the location of the CoN is given in relation to the mounting holes of the device, in the CAD the location of the CSYS was based off the CSYS drawing printed onto the housing which is imprecise. This resulted in a total offset of ~0.055 mm. The location of the CPT7 CSYS has now been corrected in the CAD to match the location of the CoN is given in relation to the mounting holes of the device.
  • The only reason as to why the back assembly (exported in the robot's base frame) could be mismatched is if the robot's base frame CSYS location in the CAD does not match its location in the URDF. Is there any document we can refer to to get the absolute location of the robot's base frame to check this?
  • Added the previously missing spacers to the back antenna to the CAD
    image
  • To reduce the size of the mesh files there are two built-in options in NX we can explore:
    • "Simplify assembly" - this option creates a new file where all selected components of an assembly are united to form a single part, internal surfaces and small features like screw holes are removed. This would drastically reduce the size of the .stl as the face densities are highest near screw holes and other small features. As part of the post processing, all CSYSs necessary for the export would need to be added back in by hand
    • "Linked Exterior" - this option allows the user to manually select all faces needing to be exported and creates a copy of those within the same assembly file. The original geometry is preserved (screw holes remain intact). This option would not require the CSYSs to be added back in, but will likely be more laborious as all faces would need to be selected by hand
    • In either case coloring and other post-processing steps would need to be redone after the export

@fulkast
Copy link
Contributor

fulkast commented Jan 5, 2025

Hi all,

I just went through and updated the IMU calibrations in the calibration_latest.yaml file, and undid the box_base_to_box_base_model visual changes from 7c40008. The images below verify that the camera frames now coincide with their lenses.

image
image
image
Here in this last view, I unticked some of the camera frames to highlight the internal IMU frames. Note the positions of the ADIS and alphasense IMUs.
image

z: 0.13247608919311352
calibration_metadata:
camera: 2024-11-05-20-54-25
imu: cad
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@JonasFrey96 the calibration at the start of this issue referenced the CAD IMU calibrations, which explains why the cameras where off wrt the IMUs.

@JonasFrey96 JonasFrey96 merged commit edc351e into main Jan 7, 2025
1 check passed
@JonasFrey96 JonasFrey96 deleted the dev/new_urdf branch July 12, 2025 23:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants