Open Source RoboticsRobots, software & learning

SMPLSim

Software

SMPL-family humanoid construction and task environments for MuJoCo and Isaac Gym, with simple PPO training utilities. [1]

Installation & first result →Official websiteGetting started

For
Researchers extending humanoid simulation and motion-control environments [1]
Source rights
Some licenses are unresolved. Review licenses
Unknowns
operating system, compute requirements not fully established. Review gaps

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware
License
BSD-3-Clause · license not fully assessed for software · Details by part

Requirements · Research details and dates

On this page

Overview

SMPLSim builds SMPL-family humanoid models and task environments for MuJoCo and Isaac Gym. It includes simple PPO training utilities and is intended as a foundation for more involved motion-control work. This profile focuses on the MuJoCo environment entry point rather than assuming the two simulator paths share one setup. [1]

Is this for you?

Known limitations
  • The documented test sends zero actions; it does not load a trained policy.
  • Missing SMPL directory selects a bundled mean-body XML fallback.
  • Root BSD-3-Clause terms conflict with the setup.py MIT/root-license header reference.

Editorial assessment from cited documentation; no runtime testing.

[2] [3] [4] [7] · Editorial assessment

What you need

Dependencies
  • MuJoCo >=3 according to README
  • Imports include PyTorch, Gymnasium, Hydra, OmegaConf, NumPy, SciPy and media utilities
  • setup.py declares no install_requires dependency list

Editorial assessment from cited documentation; no runtime testing.

[1] [2] [3] [4] · Editorial assessment

Installation & first result

Open the upstream getting-started guide [1]

  • Check out the reviewed revision and inspect the imported dependencies.
  • Install the package into a separately resolved environment.
  • From the root run python examples/env_humanoid_test.py headless=False to inspect the environment.

Editorial assessment from cited documentation; no runtime testing.

[1] [2] [3] · Editorial assessment

Installation

An editable install exposes smpl_sim, but dependencies must be resolved separately; the package does not declare them.

Editorial assessment from cited documentation; no runtime testing.

[2] [3] [4] · Editorial assessment

Hardware, models & data

Inputs and outputs

Body configuration produces humanoid simulation models; environment actions advance simulated state.

[1] [4]

Project notes

The environment test is not policy playback

At reviewed commit b5c08720503ad5fff64050c4d289c42d947fcf8d, upstream lists python examples/env_humanoid_test.py headless=False as an evaluation command. The implementation resets an environment and repeatedly supplies zero-valued actions; it does not load a trained policy. Use it to inspect model construction and rendering, not to assess learned motion quality. No GitHub release was listed. [1] [3] [6]

Installation needs more work than the README's “pip-installable” description may suggest. setup.py declares the package but no dependency list. An editable checkout install can expose the module, but it cannot be relied on to provision the imported simulator, PyTorch, Gymnasium, Hydra/OmegaConf, scientific and media packages. Upstream calls for MuJoCo 3 or newer; this review does not establish a complete compatible environment or Python version. Resolve and record that environment before running the example from the repository root. [1] [2] [3] [4]

The default robot configuration points to data/smpl. If that directory exists, the environment constructs a model through SMPL_Robot; otherwise it prints a missing-files message and uses a bundled mean-body XML. Seeing a humanoid render therefore would not prove that custom SMPL assets were loaded. Body-model acquisition and rights must be checked separately for custom models. [4] [8]

Training commands for speed, get-up and reach tasks are separate from this zero-action example. The default configuration selects the base environment and a simple MLP learning configuration; changing the task is an explicit part of upstream's training commands. [1] [5]

What we checked

Source researched only; no installation, training, benchmark or physical hardware test performed. No affiliate links or reused media.

Not yet established

  • Requirements: operating system, host computer
  • Version-specific hardware compatibility (ecosystem tags alone do not establish it)

Updates

Upstream last updated
Nov 3, 2025 [6]
Latest release
Not established
Profile updated
Oct 17, 2026
Sources last checked
Oct 17, 2026
Next review
Nov 17, 2026
Reviewed revision
commit snapshot; no GitHub releases listed / b5c08720503ad5fff64050c4d289c42d947fcf8d · software repository

Source files and licenses

Available simulation source with useful configuration boundaries; incomplete dependency specification and conflicting file-level license notice disclosed.

LayerAvailability / license
softwareavailable · License not fully assessed
Root BSD-3-Clause conflicts with setup.py MIT/root-license header reference; file-level scope unresolved. External SMPL-family assets have separate terms. [2] [7]
documentationavailable · BSD-3-Clause
Repository materials only; external components retain their own terms. [7]

The scoped zero-action environment example loads no learned policy; custom body-model rights are separately disclosed.

The scoped environment example requires no motion training dataset.

License scope needs clarification

The root license is BSD-3-Clause, but setup.py carries a Meta copyright header referring to an MIT license in the root. That mismatch remains unresolved; the root file is not sufficient to erase file-level or body-model obligations. We inspected source without installing dependencies, running the example or training a controller. [2] [7]

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Correction guidance · Discussion is not enabled · Profile format 0.7.0-draft.5

Change history