dora-openarm-mujoco

MuJoCo simulation node for the OpenArm bimanual robot, designed to run inside a dora-rs dataflow.

It replaces the physical follower arms and cameras: it accepts joint-position commands and publishes arm observations and JPEG camera frames at the same interface as the real hardware.

Installation

uv sync

Quick start

A self-contained dummy dataflow is included for testing without real hardware. It wires a dummy leader node to the MuJoCo sim and records to disk.

uv run dora build dataflow-dummy.yaml --uv
uv run dora run dataflow-dummy.yaml

Example

Dataflow configuration

Minimal (headless and position forwarding only no cameras)

- id: openarm-mujoco
  build: pip install -e .
  path: dora-openarm-mujoco
  inputs:
    position_right: leader/follower_position_right
    position_left:  leader/follower_position_left
  outputs:
    - status
    - arm_right_observation
    - arm_left_observation

Full (interactive viewer + all cameras, with contacts, can be used for vr teleoperation)

- id: openarm-mujoco
  build: pip install -e .
  path: dora-openarm-mujoco
  args: "--viewer --render --enable-collision --ctrl --keyframe home"
  inputs:
    position_right: leader/follower_position_right
    position_left:  leader/follower_position_left
  outputs:
    - status
    - arm_right_observation
    - arm_left_observation
    - camera_wrist_right
    - camera_wrist_left
    - camera_head_left
    - camera_head_right
    - camera_ceiling

Inputs

ID	Type	Description
`position_right`	`float32[8]`	Target joint positions for the right arm: joints 1–7 then the gripper. ~500 Hz.
`position_left`	`float32[8]`	Same layout for the left arm.
`pose_right`	`float32[7]`	VR controller pose `[x, y, z, qw, qx, qy, qz]`, expressed in the `--origin-frame` frame (default: the scene's `arm_origin` site). Used only with `--debug-frames`.
`pose_left`	`float32[7]`	Same for the left controller.
`button_x`	`bool[1]`	X button state. Edge-triggered: on press every scene joint on non-arm bodies (freejoint objects plus fixtures like drawers/doors) snaps back to the `--keyframe` pose; the button must be released to re-arm.

Outputs

ID	Type	Description
`arm_right_observation`	`float32[8]`	Observed joint positions, published per incoming command.
`arm_left_observation`	`float32[8]`	Same for the left arm.
`camera_wrist_right`	`uint8[N]`	JPEG frame, ~30 Hz. Requires `--render`.
`camera_wrist_left`	`uint8[N]`	JPEG frame, ~30 Hz. Requires `--render`.
`camera_head_left`	`uint8[N]`	JPEG frame, ~30 Hz. Requires `--render`.
`camera_head_right`	`uint8[N]`	JPEG frame, ~30 Hz. Requires `--render`.
`camera_ceiling`	`uint8[N]`	JPEG frame, ~30 Hz. Requires `--render`.

Camera outputs carry metadata={"encoding": "jpeg"}.

Arguments

Pass these via the args: field in the dataflow YAML, or directly on the command line.

Argument	Default	Description
`--xml PATH`	unset	MJCF scene file to load. Overrides `--scene` when set.
`--scene NAME`	`cell`	Bundled scene to load when `--xml` is not set. Choices: `cell`, `demo`, `pedestal`, `bimanual`.
`--keyframe NAME`	`home`	Keyframe in the MJCF to reset to on startup.
`--enable-collision`	off	Enable contact/collision detection. Disabled by default to avoid unexpected joint-locking during teleoperation.
`--ctrl`	off	Write incoming positions to `data.ctrl` and step the physics (`mj_step`) to simulate actuator control. The default writes directly to `data.qpos` with `mj_forward`.
`--viewer`	off	Open the interactive MuJoCo viewer window. Requires a display.
`--render`	off	Enable offscreen camera rendering and publish JPEG frames. Leave off if cameras are not needed.
`--debug-frames`	off	Draw VR controller poses as coloured arrows in the viewer. Only visible with `--viewer`.
`--origin-frame NAME`	`arm_origin`	Frame incoming `pose_right`/`pose_left` are relative to; the overlay composes poses with its live world pose and draws the reference axes there. Pass `world` for raw world-frame poses. Missing frame → warning + world fallback.
`--origin-frame-type TYPE`	`site`	MuJoCo object type of `--origin-frame`. Choices: `body`, `site`, `geom`.

License

Licensed under the Apache License 2.0. See LICENSE for details.

Code of Conduct

All participation in the OpenArm project is governed by our Code of Conduct.

Metadata

Release files for dora-openarm-mujoco 0.3.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for dora-openarm-mujoco 0.3.0
File	Size	Uploaded
dora_openarm_mujoco-0.3.0.tar.gz	17.6 kB	Jul 28, 2026	Details

Release files / dora_openarm_mujoco-0.3.0.tar.gz

Download URL	dora_openarm_mujoco-0.3.0.tar.gz
Size	17.6 kB
Tags	Source
SHA-256 checksum How to use checksums	`2c35c630999c9ecdb4a643a8d24e4d7a730774f684b59654899a6747bf27e51e`
BLAKE2b-256 checksum How to use checksums	`26b9aaf1e21568ffd128aea2025998a5b9ee6fd0c977f2478863c0edf4fa52d8`
Upload date	Jul 28, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/6.1.0 CPython/3.13.14`

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/enactic/dora-openarm-mujoco
Publishing commit: github.com/enactic/dora-openarm-mujoco/tree/2bdf35f584415568c01f8ea3c2c8523ae9b0766a

Workflow

Publishing configuration: github.com/enactic/dora-openarm-mujoco/blob/2bdf35f584415568c01f8ea3c2c8523ae9b0766a/.github/workflows/package.yaml
Publishing logs: github.com/enactic/dora-openarm-mujoco/actions/runs/30323523463/attempts/1

Artifact

Subject: dora_openarm_mujoco-0.3.0.tar.gz
SHA-256 checksum: 2c35c630999c9ecdb4a643a8d24e4d7a730774f684b59654899a6747bf27e51e Verifying attestations