How to add Triad/Axes to the MeshcatVisualizer

[joao-pm-santos96]

Hello!
First, thank you for your amazing work.

I have loaded the URDF of robot and would like to add a triad/visual axis system in all joints of my robot (for visualization purposes), that should move as the robot moves. Similar to this: https://images.app.goo.gl/3seYx75xiP1mSB747

I now that meshcat has a method to create a triad geometry, and it can be added. But if I do it with meshcat, I will need to update the triad pose everytime "manullay". Is there any way for this to be automatics, somethinf like the addGeometryObject() method?

Thank you.

[jcarpent]

I think @stephane-caron or @ManifoldFR will be the most skilled persons to answer this interesting question.

[stephane-caron]

But if I do it with meshcat, I will need to update the triad pose everytime "manullay".

That's how it's done indeed, for instance in these IK examples:

    viewer = viz.viewer
    meshcat_shapes.frame(viewer["base"], opacity=1.0)
    meshcat_shapes.frame(viewer["base_target"], opacity=0.5)

    while True:
        [...]
        # Update visualization frames
        viewer["base_target"].set_transform(base_target.np)
        viewer["base"].set_transform(
            configuration.get_transform_frame_to_world(base_task.body).np
        )
        [...]
        viz.display(q)
        t += dt

The frame() function reduces a tad the boilerplate for adding the triad, but I also feel this can be improved further.

For frames that are part of the model, we can accept proposals/PRs to extend the pin.visualize.MeshcatVisualizer. For instance the user could declare a set of frames to be displayed as triads, and those triads would be updated when calling MeshcatVisualizer.display. What do you think?

How would we go about target frames then? Thinking aloud: Those are usually updated manually on the user side anyway. Maybe we should encourage adding target frames to the model via fixed joints whose placements can be updated manually. I guess those would not affect FK/ID/FD computation times in any significant way, and it would be better practice to handle all frames in the model.

[stephane-caron]

But if I do it with meshcat, I will need to update the triad pose everytime "manullay".

That's how it's done indeed, for instance in these IK examples:

    viewer = viz.viewer
    meshcat_shapes.frame(viewer["base"], opacity=1.0)
    meshcat_shapes.frame(viewer["base_target"], opacity=0.5)

    while True:
        [...]
        # Update visualization frames
        viewer["base_target"].set_transform(base_target.np)
        viewer["base"].set_transform(
            configuration.get_transform_frame_to_world(base_task.body).np
        )
        [...]
        viz.display(q)
        t += dt

The frame() function reduces a tad the boilerplate for adding the triad, but I also feel this can be improved further.

For frames that are part of the model, we can accept proposals/PRs to extend the pin.visualize.MeshcatVisualizer. For instance the user could declare a set of frames to be displayed as triads, and those triads would be updated when calling MeshcatVisualizer.display. What do you think?

How would we go about target frames then? Thinking aloud: Those are usually updated manually on the user side anyway. Maybe we should encourage adding target frames to the model via fixed joints whose placements can be updated manually. I guess those would not affect FK/ID/FD computation times in any significant way, and it would be better practice to handle all frames in the model.