Document Digit closed-loop articulation limitation on Newton

[ooctipus]

Summary

• Drop newton_mjwarp from the three Digit env rows in docs/source/overview/environments.rst (Isaac-Velocity-Flat-Digit-v0, Isaac-Velocity-Rough-Digit-v0, Isaac-Tracking-LocoManip-Digit-v0).
• Add a "Closed-loop articulations on Newton (e.g. Agility Digit)" section to docs/source/refs/issues.rst describing the root cause and recommending the physx preset until the upstream fix lands.

Why

Robots with closed kinematic loops (Digit's achilles rod and toe push-rods) do not run correctly under the newton_mjwarp physics preset today. Newton's ArticulationView builds its per-link axis by walking model.joint_child over the articulation's joint range and sorted()-ing without deduplication, so a body that is the child of N joints occupies N slots on that axis. On Digit this gives view.link_count = 49 against 43 unique physical bodies; the four loop-closed bodies (left/right tarsus, left/right toe_roll) account for the six extra slots. Any code path that assumes one entry per logical body breaks (e.g. tensor-shape mismatches in feet_slide-style rewards).

The supported-environments tables previously listed newton_mjwarp for Digit, which is misleading. This PR removes that claim and documents the limitation under Known Issues.

Test plan

• ./isaaclab.sh -f (pre-commit) passes locally.
• Sphinx build renders the new Known Issues entry without warnings and the Digit rows in the environments tables show only physx.

[greptile-apps]

Greptile Summary

This is a documentation-only PR that corrects the supported-physics-preset claims for the three Agility Digit locomotion environments and adds a Known Issues entry explaining the underlying Newton ArticulationView bug.

• Removes newton_mjwarp from all six Digit environment entries (grid table + list-tables) in environments.rst, leaving only physx as the supported preset.
• Adds a new "Closed-loop articulations on Newton" section to issues.rst with a root-cause explanation (duplicate link slots for loop-closed bodies) and a physx-as-workaround recommendation.

Confidence Score: 5/5

Documentation-only change that removes misleading preset claims and adds a workaround guide; no runtime code is modified.

All three Digit environment entries are consistently updated in both the grid table and the list-tables in environments.rst, the new Known Issues section in issues.rst is accurate and well-structured, and the .skip sentinel file correctly indicates no changelog entry is needed. Nothing executable changes.

No files require special attention.

Important Files Changed

Filename	Overview
docs/source/overview/environments.rst	Removes newton_mjwarp from all six Digit environment entries (three in the grid table, three in the list-tables) so the supported presets column shows only physx.
docs/source/refs/issues.rst	Adds a new "Closed-loop articulations on Newton" section explaining the root cause (duplicate link slots in ArticulationView) and directing users to the physx preset as a workaround.
source/isaaclab_newton/changelog.d/octi-docs-digit-newton-known-limitation.skip	Empty .skip sentinel file marking that no changelog entry is required for this docs-only change.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Select Isaac Lab Environment] --> B{Robot has closed-loop\nkinematic chain?}
    B -- Yes\ne.g. Agility Digit --> C[Use physx preset]
    B -- No\ne.g. Anymal, G1, H1... --> D{Preferred preset?}
    D --> E[physx]
    D --> F[newton_mjwarp]
    C --> G[Environment runs correctly]
    E --> G
    F --> G
    subgraph Known Issue
        H[Newton ArticulationView\nwalk joint_child without dedup\nlink_count > unique bodies\ne.g. Digit: 49 slots vs 43 bodies]
    end
    B -- Yes --> H
    H --> C

Loading

_{Reviews (1): Last reviewed commit: "Document Digit closed-loop articulation ..." | Re-trigger Greptile}