[Feature] MJCF generation at runtime #18

Ortisa · 2025-12-04T17:42:14Z

This PR allows to generate MJCF at runtime and to be able to publish that on a separate topic.

The following features are added into this PR:

Generate MJCF and publish it over a topic using argument --publish_topic
Added --scene arg to be able to switch between scene at runtime
Support for defining input tags (raw_inputs and processed_inputs) directly in the URDF
Added -a / --asset_dir argument to specify the directory to reuse already generated objects from a given asset folder, otherwise generate them.
Option to also choose whether to save the model with argument -s

saikishor · 2025-12-04T19:59:44Z

@eholum-nasa @ndunkelb-nasa when you get some time, please take a look at this PR. We will polish things by tomorrow, but good to have some initial insights 🙌🏽

Ortisa · 2025-12-09T13:47:14Z

Grabacion.de.pantalla.desde.09-12-25.14.43.48.webm

This shows the model loading from the topic and generating the model on-the-fly from URDF

Ortisa · 2025-12-09T13:48:38Z

Grabacion.de.pantalla.desde.09-12-25.14.41.22.webm

If executed without reutilizing the assets, then due to decomposition you see that it takes longer to start the simulation

eholum-nasa

FWIW I merged the trimesh changes into this branch just to check it out in jazzy any it worked! There are still functional things to worry about, like mesh colorings that bpy would just handle and trimesh doesn't seem to support. So a lot of grey in our converted MJCFs... But the changes in this particular PR seem very nice.

Added a few thoughts to the args from trying it out.

scripts/make_mjcf_from_robot_description.py

eholum-nasa · 2025-12-10T14:57:39Z

scripts/make_mjcf_from_robot_description.py

-    # Part inputs data
-    raw_inputs, processed_inputs = parse_inputs_xml(parsed_args.mujoco_inputs)
+    output_filepath = parsed_args.output
+    if not os.path.isabs(parsed_args.output) and parsed_args.publish_topic:


I'm not sure I understand this conditional? Why the check on if the topic is published?

I'll be frank, we saw that there was a minor bug on the assets directory if you use both --publish-topic and --output tags and this is just a way to fix it. We didn't spend too much time there. I can also make them mutually exclusive if needed

scripts/make_mjcf_from_robot_description.py

Co-authored-by: Erik Holum <erik.holum@nasa.gov>

saikishor · 2025-12-10T18:42:57Z

FWIW I merged the trimesh changes into this branch just to check it out in jazzy any it worked! There are still functional things to worry about, like mesh colorings that bpy would just handle and trimesh doesn't seem to support. So a lot of grey in our converted MJCFs... But the changes in this particular PR seem very nice.

@eholum-nasa I don't see the commits on the branch

ndunkelb-nasa · 2025-12-17T21:07:51Z

FWIW I merged the trimesh changes into this branch just to check it out in jazzy any it worked! There are still functional things to worry about, like mesh colorings that bpy would just handle and trimesh doesn't seem to support. So a lot of grey in our converted MJCFs... But the changes in this particular PR seem very nice.

@eholum-nasa I don't see the commits on the branch

Maybe @eholum-nasa missed the reference to this branch? Just wanted to put out there that I added some extra commits after @eholum-nasa made this comment that fixed mesh colorings (specifically from going from stl -> obj) and adding in handling daes that have textures that come from png or jpgs.

saikishor · 2025-12-17T21:19:02Z

@ndunkelb-nasa this is great. Do you want to merge it into this PR? or shall we do it in a different PR?. Let me know. Depending on it, I can merge it right now

saikishor · 2025-12-17T21:20:34Z

@ndunkelb-nasa Apart from that, I would like to hear from you what you think about the features in this PR?. Do you think it is interesting?

ndunkelb-nasa · 2025-12-17T21:23:44Z

@ndunkelb-nasa this is great. Do you want to merge it into this PR? or shall we do it in a different PR?. Let me know. Depending on it, I can merge it right now

I think I am happy to keep it separate bc there is already a bunch of focused discussion in this one. But I think after this goes in, we should certainly go ahead and do a PR for that as well. I would like someone to check my work as well!

ndunkelb-nasa · 2025-12-17T21:25:22Z

@ndunkelb-nasa Apart from that, I would like to hear from you what you think about the features in this PR?. Do you think it is interesting?

Absolutely! I think this is exactly the right path and makes it more baked into the ROS workflow instead of having to have a whole new path. The main thing I am struggling with is to figure out what exactly I need to run with something like the test robot to have a reproducible example... Would you be able to add a launch arg or something in there so that it is easy to test these features? I have tried a couple of things, but because things are just topic based now, there isn't a lot of errors that get thrown, sometimes things just crash if I haven't configured everything properly (which I clearly haven't).

ndunkelb-nasa · 2025-12-17T21:28:04Z

test/launch/test_robot.launch.py


    robot_description = {"robot_description": ParameterValue(value=robot_description_content, value_type=str)}

+    converter_command = [


So what I thought I understood from the new code was that I could replace this with something like

converter_process = Node( package="mujoco_ros2_simulation", executable="make_mjcf_from_robot_description.py", output="screen", arguments=[ "--publish_topic", "mujoco_robot_description", "--scene", PathJoinSubstitution( [ FindPackageShare("mujoco_ros2_simulation"), "test_resources", "scene.xml", ] ), ], )

and I could remove the mujoco_model parameter from the urdf, and it should just work. However, when I do that, my ros2 control node crashes, but I do see the mujoco_robot_description_topic with the mujoco info.

What am I missing here?

I believe yes, something like that should already get it to work. Please give me a couple of minutes to check

I think im getting a failure on this line

Unfortunately, I cannot test it rolling with bpy

[make_mjcf_from_robot_description-5] Traceback (most recent call last): [make_mjcf_from_robot_description-5] File "/home/user/ros2_control_ws/install/mujoco_ros2_simulation/share/mujoco_ros2_simulation/scripts/make_mjcf_from_robot_description.py", line 21, in <module> [make_mjcf_from_robot_description-5] import bpy [make_mjcf_from_robot_description-5] ModuleNotFoundError: No module named 'bpy' [ERROR] [make_mjcf_from_robot_description-5]: process has died [pid 8272, exit code 1, cmd 'python3 /home/user/ros2_control_ws/install/mujoco_ros2_simulation/share/mujoco_ros2_simulation/scripts/make_mjcf_from_robot_description.py'].

Shall we merge your changes into this as well? that should make things easier

ah maybe I just don't have the inputs there

Yes, your URDF description needs to have the URDF inputs and all

alright, I actually figured it out. It was missing the inputs and the typical scene like we have there needs to get rid of this line so it doesn't try to reference a file that no longer exists

I still think this PR could benefit from a working example of this workflow, because I think this is actually what everyone should be using instead of generating separate mjcf files.

Yes, for now the idea is not to have any includes in the scene file. Just to not to overcomplicate things

saikishor · 2025-12-17T21:49:32Z

@ndunkelb-nasa If there is any other thing, let me know. I'll be happy to address them

ndunkelb-nasa · 2025-12-17T21:54:33Z

@ndunkelb-nasa If there is any other thing, let me know. I'll be happy to address them

I think the last thing that would be useful would be to add a launch arg in test_robot.launch.py that shows an example of this. I think it would be good to show users how to use this workflow, because I think it should probably be the default way to use these drivers.

saikishor · 2025-12-17T21:56:05Z

@ndunkelb-nasa If there is any other thing, let me know. I'll be happy to address them

I think the last thing that would be useful would be to add a launch arg in test_robot.launch.py that shows an example of this. I think it would be good to show users how to use this workflow, because I think it should probably be the default way to use these drivers.

Yess, I was about to do that, but the CI would fail due to the recently added tests. I will instead add that in the new PR of trimesh. Is that okay with you?

ndunkelb-nasa · 2025-12-17T21:58:27Z

@ndunkelb-nasa If there is any other thing, let me know. I'll be happy to address them

I think the last thing that would be useful would be to add a launch arg in test_robot.launch.py that shows an example of this. I think it would be good to show users how to use this workflow, because I think it should probably be the default way to use these drivers.

Yess, I was about to do that, but the CI would fail due to the recently added tests. I will instead add that in the new PR of trimesh. Is that okay with you?

Good with me! Seems like its time for us to update all of our workflows to use this as the default!

saikishor · 2025-12-17T21:59:13Z

Good with me! Seems like its time for us to update all of our workflows to use this as the default!

I'm glad. Shall we wait for @eholum-nasa's approval too? or we proceed with merge?

ndunkelb-nasa · 2025-12-17T22:00:16Z

Good with me! Seems like its time for us to update all of our workflows to use this as the default!

I'm glad. Shall we wait for @eholum-nasa's approval too? or we proceed with merge?

I imagine he is good to go, but I think he will probably see this early tomorrow morning, so I think we might as well wait for him.

saikishor · 2025-12-17T22:01:30Z

@ndunkelb-nasa, what's the best media to reach you for some important discussions?

ndunkelb-nasa · 2025-12-17T22:11:27Z

@ndunkelb-nasa, what's the best media to reach you for some important discussions?

I'm not sure I would have a good way at the moment. What would be normal for y'all? Discord?

saikishor · 2025-12-17T22:13:38Z

I'm not sure I would have a good way at the moment. What would be normal for y'all? Discord?

We are connected via slack channel of ros-controls or I also use Discord with same username as github

Ortisa and others added 20 commits December 4, 2025 18:09

added topic publisher

38617b8

restored the order of the instructions in main

30cfe0c

fixed mujoco model publish

dfa82e4

Added arguments

9e77ada

Improved arguments management in main

836a447

changed to general topic name

35b481d

Added logic to reuse existing mesh

2f8f6ae

Optimized using previous generated .obj

3062d5f

Removed unnecessary variable

cfd3d65

Using existing global variable

c478723

added arg error management

469dda5

Managed reusing second level generated folder

c1396a4

removed temporal folder

8dc4c79

Managed threshold check for decomposed meshes

6b3e0f2

improved args and functions in main

373fa1f

fixed asset mesh path

afdba2e

improved identation

01593e6

added scene

0bf8cce

improved logic and structure for obj meshes

7bf23e3

fixed the comments

2b0c230

saikishor assigned saikishor and Ortisa Dec 4, 2025

saikishor and others added 6 commits December 4, 2025 22:44

apply pre-commit changes

1ee8b51

minor typos

b65d1c3

change metatdata name and added check for paths

a878972

Managed scene creation for different cases

68b8994

Added comments

ee31f78

Add pre-commit fix

be9f283

saikishor requested a review from eholum-nasa December 5, 2025 20:05

Merge branch 'main' into feature_mjcf_generator_runtime

69a77dc

eholum-nasa reviewed Dec 10, 2025

View reviewed changes

saikishor and others added 2 commits December 10, 2025 19:40

Apply suggestions from code review

1003cba

Co-authored-by: Erik Holum <erik.holum@nasa.gov>

Use try_shutdown

a92298f

saikishor added 2 commits December 10, 2025 19:43

Merge branch 'main' into feature_mjcf_generator_runtime

ff6a118

Merge branch 'main' into feature_mjcf_generator_runtime

0de172d

ndunkelb-nasa reviewed Dec 17, 2025

View reviewed changes

ndunkelb-nasa approved these changes Dec 17, 2025

View reviewed changes

eholum-nasa approved these changes Dec 18, 2025

View reviewed changes

saikishor enabled auto-merge (squash) December 18, 2025 16:12

saikishor disabled auto-merge December 18, 2025 16:13

saikishor merged commit 3c19c9b into ros-controls:main Dec 18, 2025
5 checks passed

saikishor deleted the feature_mjcf_generator_runtime branch December 18, 2025 16:13


		robot_description = {"robot_description": ParameterValue(value=robot_description_content, value_type=str)}

		converter_command = [

[Feature] MJCF generation at runtime #18

[Feature] MJCF generation at runtime #18

Uh oh!

Conversation

Ortisa commented Dec 4, 2025

Uh oh!

saikishor commented Dec 4, 2025

Uh oh!

Ortisa commented Dec 9, 2025

Uh oh!

Ortisa commented Dec 9, 2025

Uh oh!

eholum-nasa left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

saikishor Dec 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

saikishor commented Dec 10, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ndunkelb-nasa Dec 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

ndunkelb-nasa commented Dec 17, 2025

Uh oh!

saikishor commented Dec 17, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

saikishor Dec 10, 2025 •

edited

Loading

ndunkelb-nasa Dec 17, 2025 •

edited

Loading