Adds live plots to managers

[pascal-roth]

Description

This adds a UI interface to the Managers in the ManagerBasedEnv and The MangerBasedRLEnv. Additions include:

• UI widgets for LiveLinePlot and ImagePlot
• ManagerLiveVisualizer/Cfg: Given a ManagerBase (i.e. action_manager, observation_manager, etc) and a config file this class creates the the interface between managers and the UI.
• EnvLiveVisualizer: A 'manager' of ManagerLiveVisualizer. This is added to the ManagerBasedEnv but is only called during the initialization of the managers in load_managers
• Adds get_active_iterable_terms implementation methods to ActionManager, ObservationManager, CommandsManager, CurriculumManager, RewardManager, and TerminationManager. This method exports the active term data and labels for each manager and is called by ManagerLiveVisualizer.
• Additions to BaseEnvWindow and RLEnvWindow to register ManagerLiveVisualizer UI interfaces for the chosen managers.

Screenshots

Screencast.from.09-06-2024.01.20.18.PM.webm

Implementation

Type of change

• New feature (non-breaking change which adds functionality)

Checklist

• I have run the pre-commit checks with ./isaaclab.sh --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have updated the changelog and the corresponding version in the extension's config/extension.toml file
• I have added my name to the CONTRIBUTORS.md or my name already exists there

[jsmith-bdai]

Left some initial comments, but haven't gone through it in detail.

It seems like we are defining a few pre-defined configs for these visualizers, but I wonder if there's a way we can have the user simply define the types of managers to be visualized e.g.:

class ManagerVisualizerCfg:
     to_visualize: ["actions", "observations", etc.]

We could probably make it a dictionary to enable the type of manager / specific config for that manager. Do you see this as being useful or do you see most users just wanting to visualize everything / nothing?

I haven't given it a ton of thought and the above is super rough, but we can discuss offline how this can be made a little simpler, while still enabling full flexibility. Happy to chat sometime today / tomorrow!

[jsmith-bdai]

Forgot to say, thanks for putting this up and all the reworking on the initial design! This will be super useful for debugging 👍

[pascal-roth]

@renezurbruegg, as the original author, some feedback from you would be helpful

[pascal-roth]

Looks good to me, maybe we should rename env_idx to env_ids for consistency?

Difference to the standard notation is that env_idx is typically just a single env id and env_ids a list/ sequence of ids. Thus, the difference would make sense to me. If you don't like it, I can rename it.

[pascal-roth]

We only want single floats and lists to be for the line plot, so that should be < 2 and then the image for everything that is 2D and 3D. So there is still something wrong with that check

[pascal-roth]

@Dhoeller19 do you think we can introduce a set of tests for the new class, I think we still have an issue in the dimension check.