Open Source RoboticsRobots, software & learning

RoboJuDo

Software

Humanoid policy deployment framework with MuJoCo and robot-specific adapters; the reviewed release has conflicting license labels that require clarification. [1] [7]

Installation & first result →Official websiteGetting started

For
Humanoid-control researchers integrating existing policies with simulation and robot SDKs [1]
Source rights
Some licenses are unresolved. Review licenses
Unknowns
compute requirements not fully established. Review gaps

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

Licenses & related hardware
License
not fully assessed for software, model weights, datasets, documentation · Details by part

Requirements · Research details and dates

On this page

Overview

RoboJuDo connects controller inputs, trained policies and execution environments for humanoid robots. The same framework organizes MuJoCo simulation and robot-specific SDK adapters, while configuration files select the policy and controller. Its support table is specific to robot-policy combinations; it does not mean every checkpoint works on every listed robot. [1] [4]

Is this for you?

Known limitations
  • README badge says CC-BY-NC-4.0 while LICENSE says CC-BY-4.0; rights remain unresolved.
  • Policy support varies by robot and configuration; some roadmap policies are still marked unreleased.
  • Physical deployment requires robot-specific SDK, network settings and verified stop behavior.
  • Later development integrations are outside this review of release 1.5.0.

Editorial assessment from cited documentation; no runtime testing.

[1] [4] [5] [7] [6] · Editorial assessment

What you need

Operating system
  • Ubuntu and Windows, as reported by upstream

[1]

Languages / runtimes
  • Python

[2]

Dependencies
  • Python >=3.11
  • MuJoCo and the selected mujoco_viewer module for the default simulation
  • Xbox controller for the default G1 simulation
  • Robot SDK only for physical execution; PHC and other modules depend on policy selection

[1] [2] [3] [5]

Host computer

The guide allows CPU PyTorch and robot onboard PCs; full memory and policy latency requirements remain unverified.

[1] [5]

Installation & first result

Open the upstream getting-started guide [1] · Project documentation [1]

  • Create a Python 3.11 environment and install the source editable.
  • Select required modules in submodule_cfg.yaml and run the documented module installer.
  • Use the default G1 MuJoCo simulation with the Xbox controller.
  • Match model, observation convention and motion source to the selected policy before changing configuration.

[1] [3] [4]

Installation

Editable install in Python 3.11+, followed by the selected optional modules.

[1] [2] [3]

Hardware, models & data

Inputs and outputs

Controller commands and environment observations feed a configured policy; its actions go to a simulator or a separately configured physical robot adapter.

[1]

Evaluation

Begin with the G1 simulation and inspect policy-specific observation/action configuration; no robot test was performed.

Editorial assessment from cited documentation; no runtime testing.

[1] [4] · Editorial assessment

Project notes

Start with the G1 simulation configuration

The package requires Python 3.11 or newer. The README reports Ubuntu and Windows support and offers CPU PyTorch as an option. After the editable installation, select optional modules in submodule_cfg.yaml; only mujoco_viewer is enabled by default. The optional-module configuration also lists upstream patches, so the setup needs more review than the Python requirements alone. We have not applied or tested those upstream patches. [1] [2] [3]

The documented first run is python scripts/run_pipeline.py, using the G1 MuJoCo setup and an Xbox controller. In the configuration, g1 selects a MuJoCo environment, joystick input and the Unitree policy. g1_real selects a different environment and controller, with robot network and SDK settings. Keep the initial check in the simulation configuration. [1] [8]

Match the policy's observation and motion conventions

For BeyondMimic, the policy guide distinguishes configurations with and without a state estimator, metadata-derived settings, and motion read from the ONNX model versus an external controller. Human2Humanoid and some other motion controllers require PHC-related modules. The guide still marks HugWBC as forthcoming even though other tables mention it; treat that as a release-specific coverage gap. [4] [3]

Physical execution requires the appropriate robot SDK and network interface. The Unitree guide distinguishes the Python SDK adapter from UnitreeCpp and recommends UnitreeCpp for onboard G1 compute. Upstream warns that a policy can produce violent motion when balance is lost and requires verified stop behavior. Its reported hardware tests are not tests performed by this catalog. [1] [5]

This profile covers the published 1.5.0 release's own documented configurations; later development changes are outside this review. Complete memory requirements, inference timing and physical deployment readiness remain unverified. [6]

What we checked

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

Not yet established

  • Version-specific hardware compatibility (ecosystem tags alone do not establish it)

Updates

Upstream last updated
Nov 10, 2025 [6]
Latest release
Nov 10, 2025 [6]
Profile updated
Oct 17, 2026
Sources last checked
Oct 17, 2026
Next review
Nov 17, 2026
Reviewed revision
1.5.0 / 8bffb75d757d106d336669ca5786c785fd5d222a · software repository

Source files and licenses

Source available, but README CC-BY-NC-4.0 badge conflicts with CC-BY-4.0 LICENSE. No unqualified open-source or commercial-rights claim. Third-party artifact terms unresolved.

LayerAvailability / license
softwareavailable · License not fully assessed
Conflicting CC-BY-NC-4.0 README badge and CC-BY-4.0 LICENSE text; clarification required. [1] [7]
model weightsavailable · License not fully assessed
Policy examples and model paths documented; individual model licenses not established. [4]
datasetsavailable · License not fully assessed
Example motion files referenced; dataset and motion provenance/permissions need separate review. [4]
documentationavailable · License not fully assessed
Same repository license conflict; no resolved license asserted. [1] [7]

Resolve the license conflict before relying on reuse rights

At reviewed release 1.5.0, the README badge labels the project CC-BY-NC-4.0, while the repository LICENSE text says CC-BY-4.0 and explicitly permits commercial adaptation. Those declarations conflict. This profile leaves the license unresolved rather than choosing the more permissive interpretation. Bundled policy models, motions and robot assets also need their own provenance and terms checked. [1] [7] [4]

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