On this page 9 sections
Isaac Gym environments (no helper planned)
Generated by MHS/registry/build_registry.py on 2026-10-20T00:00:00Z from 58 registry records and their task cards. A plan, not a conformance claim.
- bridge_status:
none - helper: none
- exemplar: none
Launch
Every driver package starts the same way; per-repo configuration (environment id, model path, port, baud rate, VISA resource) lives in the package's driver.py, never in launch flags:
mhs serve <driver-dir>A package is done when python3 -m mhs.driver check <driver-dir> --backend fake exits 0 (evidence contract_checked; a fake never yields an MHS-Device level).
Registry coverage
- Repos: 58
- Decisions: bridge-runtime 2, reference 2, skip 54
- Waves: 9: 58
- Dispatchable now: 0
- Modes: simulated 2
Applicability
Repos the catalogue files as Isaac Gym environments. No Driver helper is planned before or after the pilot, so their cards are recon-only (wave 4) or last (wave 6: GPU, Windows or an engine licence). A simulator family needs a structural marker in the code, never only a README or name match.
Capability mapping
| native call | OpenMHS target | notes |
|---|---|---|
| construct the environment or load the model | lifecycle: construction (class attributes in driver.py) | never cap.world.reset.v1 |
| reset(seed) | cap.world.reset.v1 | output: the initial observation; a seed set with cap.world.seed.v1 applies at the next reset |
| step(action) | cap.world.step.v1 {ticks, action} | the action space becomes the input schema with per-element limits |
| get_state() / set_state() | mhs.snapshot / mhs.restore (protocol methods) | only when the engine restores the complete state; cap.world.snapshot.v1 and cap.world.restore.v1 are deprecated |
| close() | lifecycle: teardown | never cap.world.restore.v1 |
Class ids come from python3 -m mhs.driver catalog search <words>. The deprecated classes cap.device.describe.v1, cap.world.state.v1, cap.world.snapshot.v1 and cap.world.restore.v1 are replaced by the protocol methods mhs.devices.get, mhs.resource.read, mhs.snapshot and mhs.restore.
Safety
The risk floor of a device is the highest catalogue risk_class of its classes (CAP-017); a driver may declare a higher risk, never a lower one. Hardware at R1 or above needs an @estop and a non-empty safe state. The risk floor of every class still comes from the catalogue.
Fidelity
Engine-specific; nothing here claims fidelity.
Pitfalls
A preview-era package pinned to old Python and CUDA versions.
First task cards
gh.leggedrobotics/legged_gym: skip, wave 9, rank 755,MHS/registry/tasks/robots/leggedrobotics__legged_gym.jsongh.lecar-lab/asap: skip, wave 9, rank 811,MHS/registry/tasks/robots/lecar-lab__asap.jsongh.clearlab-sustech/wheel-legged-gym: skip, wave 9, rank 1158,MHS/registry/tasks/robots/clearlab-sustech__wheel-legged-gym.jsongh.pku-epic/dexgraspnet: skip, wave 9, rank 1200,MHS/registry/tasks/robots/pku-epic__dexgraspnet.jsongh.lupinjia/leggedgym-ex: skip, wave 9, rank 1218,MHS/registry/tasks/robots/lupinjia__leggedgym-ex.json
Sources
- Task cards:
MHS/registry/tasks/<track>/<owner>__<repo>.json(schema/taskcard.schema.json). - Triage rules and the helper table:
MHS/registry/triage.py; curated decisions:MHS/registry/overrides/. - Driver API and gate:
MHS/sdk/docs/driver-authoring.md; exemplars:MHS/drivers/.
Packages
Top 24 of 58 by stars. Filter the full list.
| Package | Tier | Ring | Evidence | Stars |
|---|---|---|---|---|
| leggedrobotics/legged_gym leggedrobotics/legged_gym: legged and humanoid robots for legged and humanoid robots (robotics track). OpenMHS plan: not | Tier D | 3 | agent read | 3,137 |
| lecar-lab/asap lecar-lab/asap: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: not integrated (skip:peripheral). D | Tier D | 3 | static scan | 2,119 |
| pku-marl/dexteroushands pku-marl/dexteroushands: manipulation task environment for robot manipulation (robotics track). OpenMHS plan: not integr | Tier D | 3 | static scan | 1,109 |
| lecar-lab/human2humanoid lecar-lab/human2humanoid: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: not integrated (skip:peri | Tier D | 3 | static scan | 1,075 |
| markfzp/humanplus markfzp/humanplus: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: reference only (prior-art). Decl | Tier D | 3 | static scan | 853 |
| ntnu-arl/aerial_gym_simulator ntnu-arl/aerial_gym_simulator: UAVs for UAVs (robotics track). OpenMHS plan: not integrated (override:recon). Declared b | Tier D | 3 | static scan | 776 |
| clearlab-sustech/wheel-legged-gym clearlab-sustech/wheel-legged-gym: complete environment for legged robots (robotics track). OpenMHS plan: not integrated | Tier D | 3 | static scan | 678 |
| lecar-lab/abs lecar-lab/abs: policy/algorithm for legged robots (robotics track). OpenMHS plan: not integrated (skip:peripheral). Decl | Tier D | 3 | static scan | 632 |
| zeonsunlightyu/locomotionwithnp3o zeonsunlightyu/locomotionwithnp3o: catalog/discovery tooling (robotics track). OpenMHS plan: not integrated (override:re | Tier D | 3 | static scan | 545 |
| pku-epic/dexgraspnet pku-epic/dexgraspnet: peripheral tooling for robot manipulation (robotics track). OpenMHS plan: not integrated (skip:per | Tier D | 3 | static scan | 475 |
| lecar-lab/humanoidverse lecar-lab/humanoidverse: complete environment for humanoid robots (robotics track). OpenMHS plan: not integrated (overri | Tier D | 3 | static scan | 474 |
| lupinjia/leggedgym-ex lupinjia/leggedgym-ex: legged and humanoid robots for legged and humanoid robots (robotics track). OpenMHS plan: not int | Tier D | 3 | static scan | 374 |
| tud-amr/mppi-isaac tud-amr/mppi-isaac: catalog/discovery tooling (robotics track). OpenMHS plan: not integrated (override:recon). Language: | Tier D | 3 | static scan | 361 |
| internrobotics/physhsi internrobotics/physhsi: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: not integrated (skip:periph | Tier D | 3 | static scan | 321 |
| improbable-ai/rapid-locomotion-rl improbable-ai/rapid-locomotion-rl: policy/algorithm for legged robots (robotics track). OpenMHS plan: not integrated (sk | Tier D | 3 | static scan | 299 |
| lecar-lab/softa lecar-lab/softa: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: not integrated (skip:peripheral). | Tier D | 3 | static scan | 147 |
| nvlabs/oscar nvlabs/oscar: controller for robot manipulation (robotics track). OpenMHS plan: not integrated (override:recon). Declare | Tier D | 3 | static scan | 139 |
| fan-ziqi/rl_amp fan-ziqi/rl_amp: policy/algorithm for legged robots (robotics track). OpenMHS plan: not integrated (skip:peripheral). De | Tier D | 3 | static scan | 130 |
| hyperplane-lab/rlafford hyperplane-lab/rlafford: manipulation task environment for robot manipulation (robotics track). OpenMHS plan: not integr | Tier D | 3 | static scan | 125 |
| shengqianchen/dreamwaq_go2w shengqianchen/dreamwaq_go2w: policy/algorithm for legged robots (robotics track). OpenMHS plan: not integrated (skip:per | Tier D | 3 | static scan | 125 |
| zhengyiluo/simxr zhengyiluo/simxr: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: not integrated (skip:peripheral). | Tier D | 3 | static scan | 122 |
| humanoid-skillblender/skillblender humanoid-skillblender/skillblender: policy/algorithm for humanoid robots (robotics track). OpenMHS plan: not integrated | Tier D | 3 | static scan | 118 |
| sectionz6/unidoormanip sectionz6/unidoormanip: manipulation task environment for robot manipulation (robotics track). OpenMHS plan: not integra | Tier D | 3 | static scan | 110 |
| iamlab-cmu/isaacgym-utils iamlab-cmu/isaacgym-utils: control framework for robot arm manipulation (robotics track). OpenMHS plan: SDK bridge runti | Tier B | 3 | static scan | 100 |