yourdfpy 0.0.50

A simpler and easier-to-use library for loading, manipulating, saving, and visualizing URDF files.

yourdfpy

Yet anOther URDF parser for Python. Yup, it's another one. Deal with it.

Yourdfpy is a simpler and easier-to-use library for loading, manipulating, validating, saving, and visualizing URDF files.

Installation

You can install yourdfpy directly from pip:

pip install yourdfpy

Visualization

Once installed, you can visualize a URDF model from the command line:

yourdfpy ./my_description/urdf/robot.urdf

You can use the following keyboard shortcuts to inspect your model:

• a: Toggle rendered XYZ/RGB axis markers (off, world frame, every frame)
• w: Toggle wireframe mode (good for looking inside meshes, off by default)
• c: Toggle back face culling (on by default but in wireframe mode it is sometimes useful to see the back sides)

But why another one?!?

Why are you wasting not only your but also our time? you might ask. Fair point. There are already urdfpy and urdf_parser_py that deal with URDFs. Unfortunately, none of these solutions allow customizable URDF parsing that is fully independent of validation and mesh loading. Dealing with filenames, outdated dependencies, open bug reports, and limited flexibility when it comes to serialization are other disadvantages. As shown in the table below, yourdfpy is the most robust one when it comes to loading URDFs in the wild.

	urdfpy	urdf_parser_py	yourdfpy
Decouple parsing from validation			:heavy_check_mark:
Decouple parsing from loading meshes		:heavy_check_mark:	:heavy_check_mark:
Visualize URDF	:heavy_check_mark:		:heavy_check_mark:
Forward Kinematics	:heavy_check_mark:		:heavy_check_mark:
Robustness test: loading 12 URDF files from here	4/12	6/12	12/12
Avg. loading time per file (w/ mesh loading)	480 ms		370 ms
(w/o mesh loading)		3.2 ms	6.2 ms
Test on 4 URDF files on which urdfpy succeeds	347.5 ms		203 ms
Test on 6 URDF files on which urdf_parser_py succeeds		2.6 ms	3.8 ms

Click to expand code listing that produces the above table entries.

robot_assets = ['robot-assets/urdfs/robots/barret_hand/bhand_model.URDF', 'robot-assets/urdfs/robots/robotiq_gripper/robotiq_arg85_description.URDF', 'robot-assets/urdfs/robots/anymal/anymal.urdf', 'robot-assets/urdfs/robots/franka_panda/panda.urdf', 'robot-assets/urdfs/robots/ginger_robot/gingerurdf.urdf', 'robot-assets/urdfs/robots/halodi/eve_r3.urdf', 'robot-assets/urdfs/robots/kinova/kinova.urdf', 'robot-assets/urdfs/robots/kuka_iiwa/model.urdf', 'robot-assets/urdfs/robots/pr2/pr2.urdf', 'robot-assets/urdfs/robots/ur10/ur10_robot.urdf', 'robot-assets/urdfs/robots/ur5/ur5_gripper.urdf', 'robot-assets/urdfs/robots/yumi/yumi.urdf']

import urdfpy
import urdf_parser_py
import yourdfpy

from functools import partial

def load_urdfs(fnames, load_fn):
    results = {fname: None for fname in fnames}
    for fname in fnames:
        try:
            x = load_fn(fname)
            results[fname] = x
        except:
            pass
    print(sum([1 for x, y in results.items() if y is not None]), "/", len(fnames))
    return results

# parsing success rate
load_urdfs(robot_assets, urdfpy.URDF.load)
load_urdfs(robot_assets, urdf_parser_py.urdf.URDF.load)
load_urdfs(robot_assets, yourdfpy.URDF.load)

# parsing times
%timeit load_urdfs(robot_assets, urdfpy.URDF.load)
%timeit load_urdfs(robot_assets, urdf_parser_py.urdf.URDF.load)
%timeit load_urdfs(robot_assets, yourdfpy.URDF.load)
%timeit load_urdfs(robot_assets, partial(yourdfpy.URDF.load, load_meshes=False, build_scene_graph=False))

# fairer comparison with yourdfpy
urdfpy_fnames = [x for x, y in load_urdfs(robot_assets, urdfpy.URDF.load).items() if y is not None]
%timeit load_urdfs(urdfpy_fnames, yourdfpy.URDF.load)

# fairer comparison with urdf_parser_py
urdfparser_fnames = [x for x, y in load_urdfs(robot_assets, urdf_parser_py.urdf.URDF.from_xml_file).items() if y is not None]
%timeit load_urdfs(urdfparser_fnames, functools.partial(yourdfpy.URDF.load, load_meshes=False, build_scene_graph=False))