sim2real-deployment

01 前言 Front Matter

name: sim2real-deployment
license: MIT
metadata: { "disco-role": "operating" }
description: Run ASAP sim2sim playback, sim2real policy runtime, Unitree/ROS2 bridges, joystick or keyboard controls, and real-data collection safely.
disable-model-invocation: true

02 资源文件 4

03 SKILL.md 正文

Use this sub-skill when the task is to operate ASAP deployment code: MuJoCo sim2sim playback, sim2real policy runtime on a Unitree G1, ROS2/Unitree bridge setup, joystick or keyboard command input, and real-world data collection.

First Safety Gate

Real robot deployment can injure people and damage hardware. Do not start policy actions on physical hardware until a qualified human confirms the robot is mechanically configured for the intended DOF count, in low-level mode, on a clear test area or support rig, with emergency-stop procedures ready. The bundled instructions explain repository behavior; they do not replace a hardware safety review.

Route by Need

Local MuJoCo sim2sim playback: read references/workflows.md and run the simulator plus policy terminals from the repository sim2real/ directory.
Real Unitree G1 policy runtime: read references/workflows.md and references/troubleshooting.md before changing INTERFACE away from localhost.
Keyboard, joystick, or MuJoCo viewer controls: use references/workflows.md for the exact key/button maps implemented by base_policy.py, deepmimic_dec_loco.py, and unitree_sdk2py_bridge.py.
Robot/config/model path edits: use references/configuration.md before editing sim2real/config/g1_29dof_hist.yaml.
Safe readiness check: run the bundled scripts/deployment_doctor.py to check imports, config shape, assets, model paths, joystick visibility, and network-interface expectations without initializing Unitree DDS or commanding motors.
Real-data collection: use references/workflows.md for listener_deltaa.py and its recording keys.

Working Directory Convention

Deployment commands in ASAP are written to run from the repository sim2real/ directory, because config paths point to ../humanoidverse/..., policy scripts append ./rl_policy, and model examples use ./models/....

cd sim2real
python sim_env/base_sim.py --config=config/g1_29dof_hist.yaml
python rl_policy/deepmimic_dec_loco_height.py --config=config/g1_29dof_hist.yaml \
  --loco_model_path=./models/dec_loco/20250109_231507-noDR_rand_history_loco_stand_height_noise-decoupled_locomotion-g1_29dof/model_6600.onnx \
  --mimic_model_paths=./models/mimic

Boundaries and Cross-Links

For HumanoidVerse training, checkpoint export, Hydra overrides, or evaluation, use ../training-and-evaluation/SKILL.md and the root ../../SKILL.md.
For SMPL/AMASS shape fitting, motion fitting, or retargeting, use ../motion-retargeting/SKILL.md.
For base installs, simulator backends, ROS2, and optional SDK prerequisites, use root ../../references/install-and-backends.md when present.
For cross-cutting package/import/config failures, use root ../../references/troubleshooting.md when present; keep deployment-specific failures in this sub-skill's references/troubleshooting.md.

This sub-skill intentionally excludes HumanoidVerse training internals and SMPL motion retargeting details except where a trained ONNX policy or retargeted mimic model path is required for deployment.

04 同仓库更多 Skills

VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351

VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351

VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351

VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351

VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351

VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351