orca_arm

URDF and MJCF descriptions for the OrcaArm---a bimanual OpenArm with two OrcaHand end effectors, plus every referenced mesh bundled inside the package: use it

This repository does not come with controllers, IK, motion planning, or a simulator---bring your own!

Install

git clone https://github.com/fracapuano/orca_arm.git
pip install -e .            # core: paths to URDF / MJCF / meshes
pip install -e .[viz]       # optionally adds meshcat + yourdfpy for the visualizer

What you get

The package exposes absolute paths to the bundled URDF and MJCF files of the OrcaArm:

import orca_arm

orca_arm.URDF_PATH    # path to orcabot.urdf
orca_arm.MJCF_PATH    # path to orcabot.xml (MuJoCo)

The URDF and MJCF reference meshes via paths relative to their own location, so any tool that resolves mesh paths from the URDF/MJCF file finds them without further configuration. No environment variables, no package:// resolver setup.

Visualize

You can visualize the OrcaArm with a meshcat viewer for a quick inspection:

python visualize_orcabot.py            # live FK loop with a random joint sweep - no collision detection here
python visualize_orcabot.py --idle     # static home configuration

MuJoCo

Load the bundled MJCF directly:

import mujoco
import orca_arm

model = mujoco.MjModel.from_xml_path(orca_arm.MJCF_PATH)
data = mujoco.MjData(model)

This is the entry point for any MuJoCo-based stack consuming MJCF.

Other URDF-consuming simulators

PyBullet, SAPIEN, ManiSkill, Isaac Sim, Drake, ROS — all accept the URDF path. PyBullet example:

import pybullet as p
import orca_arm

p.connect(p.DIRECT)
robot = p.loadURDF(orca_arm.URDF_PATH)

The same pattern works for any other URDF loader: hand it orca_arm.URDF_PATH.

Forward kinematics with yourdfpy

For pure kinematics — link / joint queries, forward kinematics, scene graph — without a physics simulator:

import numpy as np
import yourdfpy
import orca_arm

robot = yourdfpy.URDF.load(orca_arm.URDF_PATH)

q = np.zeros(len(robot.actuated_joint_names))
robot.update_cfg(q)

# 4x4 world transform of any link in the current configuration
T_world_link, _ = robot.scene.graph.get("<link_name>")

Layout

We are also releasing build_*.py scripts to regenerate the URDF and MJCF from the OpenArm and OrcaHand source descriptions. These are only relevant if you are updating the assets themselves; ordinary downstream use does not require running them!