Static Apple pick & place: Update docs and add data collection instructions

[pulkitg01]

Summary

Add data collection guidelines, clarify teleop and evaluation docs

Detailed description

• Reason for the change:

  • Users need clearer prerequisites for IsaacTeleop hardware/system support.
  • The static apple workflow needed more explicit guidance for collecting clean, high-quality demonstrations and evaluating policy success reliably.

• What changed:

  • Added IsaacTeleop supported-device and system-requirements links to teleoperation docs across the example workflows.
  • Expanded static apple data collection guidance with warm-up, smoothness, fixed-body motion, grasp diversity, clean-success criteria, trajectory length, replay validation, and demo-count guidance.
  • Updated the suggested apple-to-plate collection sequence to first move the right hand aside, then approach/grasp/move/place the apple.
  • Clarified evaluation guidance to recommend complete-episode evaluation with --num_episodes 100 or --num_episodes 1000, plus approximate --num_steps equivalents.

• Impact:

  • Improves documentation clarity for teleop setup, demonstration quality, and success-rate evaluation.
  • Docs-only change; no runtime behavior changes.

[isaaclab-review-bot]

Thank you for improving the documentation, @pulkitg01!

Summary

This PR adds valuable clarity to the teleoperation and evaluation workflows across multiple example pages. The changes are well-structured and address real user needs.

What looks good

✅ Hardware prerequisites: Adding links to IsaacTeleop supported devices and system requirements across all teleoperation docs is helpful — users will appreciate knowing prerequisites upfront.

✅ Data collection protocol: The expanded guidance in the static apple workflow covers important best practices (warm-up runs, smoothness, grasp diversity, trajectory length, replay validation). This kind of explicit protocol helps users collect higher-quality demonstrations.

✅ Suggested task sequence: The updated step-by-step sequence (move right arm aside → approach → grasp → lift → place) is clearer and more actionable than the previous version.

✅ Evaluation guidance: Clarifying that --num_episodes 100 or --num_episodes 1000 is preferred over short --num_steps runs, along with the step-to-episode math, helps users understand how to get representative success rates.

✅ External links verified: Both IsaacTeleop documentation links resolve correctly.

Minor suggestions (non-blocking)

1. Typo in commit message: The first commit message has "requriement" instead of "requirement" — not a blocker since it's just commit history, but worth noting.

2. Consider cross-referencing: The detailed data collection protocol in static_apple/step_2_teleoperation.rst could be useful for other workflows too. Consider whether a shared include or cross-reference might reduce duplication in the future.

Verdict

Documentation-only change with accurate content and working links. No issues found.

🟢 LGTM

---

Update (aad96f5): Updated scene screenshot (docs/images/static_apple_scene.png) to highlight the XR tab — helpful visual improvement for users following the setup instructions. ✅ Still LGTM.

---

Update (28dc765): Nice refinements in this push:

• IsaacTeleop requirements links: Updated across 5 docs to point to the more specific #teleoperation-with-isaac-sim-and-isaac-lab anchor — helps users land directly on the relevant section.
• Replay clarification: Important addition explaining that replay_demos.py replays actions (not exact trajectory playback), and that open-loop replay can miss/drop objects due to physics differences. This manages expectations well.
• Episode length source: Evaluation guidance now points to the actual config (episode_length_s=6.0) — cleaner than leaving it as a magic number.

All documentation refinements, no code changes. ✅ Still LGTM.

[greptile-apps]

Greptile Summary

This is a documentation-only PR that improves the static apple pick-and-place workflow docs and adds consistent IsaacTeleop prerequisite notes across four teleoperation guides.

• Added a uniform .. note:: block linking to IsaacTeleop supported-device and system-requirements pages in all four step_2_teleoperation.rst files (locomanipulation, sequential_static_manipulation, static_apple, static_manipulation).
• Expanded the static apple step_2_teleoperation.rst with a detailed data-collection protocol covering warm-up, smoothness, body motion, grasp diversity, clean successes, trajectory length, and replay validation, plus an updated suggested sequence that clears the right arm first before the apple approach.
• Updated step_4_evaluation.rst to recommend episode-based evaluation (--num_episodes 100/1000) with correct --num_steps equivalents derived from the task's 6-second/50 Hz episode configuration; also updated the scene image.

Confidence Score: 5/5

Documentation-only change with no runtime behavior modifications; safe to merge.

All six changed files are documentation (RST) and an LFS image pointer. The new content is internally consistent: the recording command already carries --num_success_steps 10 (matching the important block's guidance), and the evaluation arithmetic (300 steps/episode × 100/1000 episodes) is correct. No code paths are affected.

No files require special attention.

Important Files Changed

Filename	Overview
docs/pages/example_workflows/static_apple/step_2_teleoperation.rst	Largest change in the PR: adds IsaacTeleop prerequisite note, a detailed data-collection .. important:: block, and a revised 5-step suggested sequence; content is accurate and internally consistent with the recording command shown in the file.
docs/pages/example_workflows/static_apple/step_4_evaluation.rst	Replaces vague "~1000 policy steps" guidance with episode-based recommendations; the step-count arithmetic (300 steps/ep × 100/1000 episodes) is correct.
docs/pages/example_workflows/locomanipulation/step_2_teleoperation.rst	Adds the same 7-line IsaacTeleop prerequisite note as the other teleoperation docs; no other changes.
docs/pages/example_workflows/sequential_static_manipulation/step_2_teleoperation.rst	Adds the same 7-line IsaacTeleop prerequisite note; no other changes.
docs/pages/example_workflows/static_manipulation/step_2_teleoperation.rst	Adds the same 7-line IsaacTeleop prerequisite note; no other changes.
docs/images/static_apple_scene.png	LFS pointer update replacing the scene screenshot; no code impact.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[User reads teleoperation doc] --> B[New: Check IsaacTeleop\nsupported devices & system requirements]
    B --> C[Start CloudXR Runtime\n& connect headset]
    C --> D[Run recording script\n--num_demos 10 tutorial\nor --num_demos 400 full]
    D --> E{New: Data collection\nprotocol}
    E --> E1[Warm-up: 5 practice runs]
    E --> E2[Smooth motions, fixed torso]
    E --> E3[Diverse grasps\ntop-down & side]
    E --> E4[Clean successes only\n200-400 timestep episodes]
    E1 & E2 & E3 & E4 --> F[New: Suggested sequence\n1. Move right arm aside\n2. Approach apple with left arm\n3. Grasp firmly\n4. Lift straight up\n5. Lower & release onto plate]
    F --> G[Replay HDF5 for\naction-level sanity check]
    G --> H{Ready for training?}
    H -- Yes --> I[Convert to LeRobot\n& policy post-training]
    H -- No --> D
    I --> J[Evaluate policy]
    J --> K[New: --num_episodes 100\nquick estimate\nor --num_episodes 1000\nstrong estimate]

Loading

_{Reviews (4): Last reviewed commit: "Add reason for num episodes" | Re-trigger Greptile}