DrakeVisualizer.jl

Interface to the RobotLocomotion Drake Visualizer tool from Julia
Popularity
18 Stars
Updated Last
1 Year Ago
Started In
April 2016

DrakeVisualizer

Build Status codecov


Note: While this package should still work, active development has been moved to MeshCat.jl, which offers more features and a much less complicated set of dependencies.


This package provides a Julia interface to the Drake Visualizer, part of the Drake project and built on top of Director, a highly customizable 3D interface for robotics visualization and interaction.

Installation

DrakeVisualizer.jl uses BinDeps.jl to try to automatically install an appropriate copy of Director for you. On Ubuntu (14.04 and higher) and macOS, this package will attempt to download the pre-built binaries of Director from http://people.csail.mit.edu/patmarion/software/director/. On other Linux platforms, it will compile Director from source. If you would like to force Director to build from source on any platform, just set the environment variable DIRECTOR_BUILD_FROM_SOURCE=1.

Dependencies

On Ubuntu (14.04 and up) and macOS, all of Director's dependencies will automatically be installed using BinDeps.jl. On other platforms, you'll need to provide them yourself. The required packages are available via apt-get as:

libqt4-dev
libqt4-opengl-dev
python-numpy
python-dev

Troubleshooting

If you have issues with the Director application itself (like your geometry not showing up), you may have an out-of-date version of the Director binaries. To clear the downloaded binaries, you can run:

julia> DrakeVisualizer.delete_director_binaries()

After which you will need to re-download the binaries with:

julia> Pkg.build("DrakeVisualizer")

Linux: Configuring Large UDP Packets

DrakeVisualizer uses LCM for communication, and LCM uses UDP under the hood. Very large LCM messages (like those created when loading a robot with lots of mesh geometries) can result in UDP packets being dropped, which will result in you not seeing anything in the visualizer. If that happens to you, you'll need to follow the instructions in this comment. Edit /etc/sysctl.conf and add:

net.core.rmem_max=2097152
net.core.rmem_default=2097152

Launching the Viewer

You can launch the viewer application with

julia> DrakeVisualizer.new_window()

which is just a convenience wrapper around a call to the drake-visualizer executable, included in the director binaries or source installation.

Usage:

This package makes use of GeometryTypes.jl to represent robot geometries and CoordinateTransformations.jl to represent spatial transformations. Check out demo.ipynb for some examples of usage.

Geometric Primitives

Geometric primitives from GeometryTypes.jl can be visualized directly:

using DrakeVisualizer
using GeometryTypes
box = HyperRectangle(Vec(0.,0,0), Vec(1.,1,1))
model = Visualizer(box)

box

sphere = HyperSphere(Point(0., 0, 0), 1.0)
model = Visualizer(sphere)

sphere

Once a Visualizer model has been created, it can be rendered at arbitrary positions and orientations:

using CoordinateTransformations
settransform!(model, Translation(1.0, 0.0, 0.0))

DrakeVisualizer can also render mesh data:

using MeshIO
using FileIO
cat = load(joinpath(Pkg.dir("GeometryTypes"), "test", "data", "cat.obj"))
Visualizer(cat)

cat mesh

And it can even generate 3D contours from functions:

# First, we'll define our function:
f = x -> sum(sin, 5 * x)

# Then we pick a region of interest in which to sample the function.
# This region starts at (-1, -1, -1) and extends to (1, 1, 1):
lower_bound = Vec(-1.,-1,-1)
upper_bound = Vec(1., 1, 1)

# contour_mesh constructs a mesh representing an approximation of
# the surface in 3D space for which f(x) = 0
mesh = contour_mesh(f, lower_bound, upper_bound)
Visualizer(mesh)

custom function contour mesh

For more visualizations, including moving and rotating visualized elements, and visualizing the level sets of functions, check out demo.ipynb.