Subscribe Free

Join 2670+ others. No spamming.
I promise!

We are currently under high development. Follow us at github.

Looking for Python Tutorials?
Check these awesome tutorials



jdlangs / RobotOS.jl


Julia interface to ROS (Robot Operating System)



Build Status Coverage Status



This package enables interfacing Julia code with a ROS (Robot Operating System) system. It works by generating native Julia types for ROS types, the same as in C++ or Python, and then wrapping rospy through the PyCall package to get communication through topics, services, and parameters.


using RobotOS


The package will hopefully continue to undergo substantial improvement. Please feel free to submit either an issue or pull request through github if you want to fix something or suggest a needed improvment, even if it's just to add an extra sentence in this README.


Currently, Pkg.test("RobotOS") requires some bootstrapping to work properly. Before running Julia, make sure a ROS master is running and start the helper node by running the test/ file.

Usage: Type Generation

ROS types are brought into your program with the @rosimport macro which specifies a package and one or more types. The three valid syntax forms can be seen in these examples:

@rosimport std_msgs.msg.Header
@rosimport nav_msgs.srv: GetPlan
@rosimport geometry_msgs.msg: PoseStamped, Vector3

@rosimport will import the python modules for the requested type and all its dependencies but the native Julia types are not created yet since any inter-module dependencies have to be resolved first. After the final @rosimport call, initiate the type generation with:


The new types will be placed in newly created modules in Main, corresponding to the packages requested. For example, "std_msgs/Header" => std_msgs.msg.Header. After calling rostypegen() they can be interacted with just like regular modules with import and using statements bringing the generated type names into the local namespace.

using nav_msgs.msg
import geometry_msgs.msg: Pose, Vector3
p = Path()
v = Vector3(1.1,2.2,3.3)

There is one special case, where the ROS type name conflicts with a built-in Julia type name (e.g., std_msgs/Float64 or std_msgs/String). In these cases, the generated Julia type will have "Msg" appended to the name for disambiguation (e.g., std_msgs.msg.Float64Msg and std_msgs.msg.StringMsg).

An additional function, rostypereset(), resets the type generation process, possibly useful for development in the REPL. When invoked, new @rosimport calls will be needed to generate the same or different types, and previously generated modules will be overwritten after rostypegen() is called again. Keep in mind that names cannot be cleared once defined so if a module is not regenerated, the first version will remain.

Usage: ROS API

In general, the API functions provided directly match those provided in rospy, with few cosmetic differences. The rospy API functions can reviewed here:

General Functions

  • init_node(name::String; kwargs...) : Initialize node. Passes keyword arguments on to rospy directly. (Required)
  • is_shutdown() : Check for ROS shutdown state.
  • spin() : Wait for callbacks until shutdown happens.
  • logdebug,loginfo,logwarn,logerr,logfatal all work as in rospy.


Native Julia types Time and Duration are defined, both as a composite of an integral number of seconds and nanoseconds, as in rospy. Arithmetic and comparison operators are also defined. A Rate type is defined as a wrapper for the rospy Rate, which keeps loops running on a near fixed time interval. It can be constructed with a Duration object, or a floating-point value, specifying the loop rate in Hz. Other functions are:

  • get_rostime(), : Current time as Time object.
  • to_sec(time_obj), convert(Float64, time_obj) : Convert Time or Duration object to floating-point number of seconds.
  • to_nsec(time_obj) : Convert object to integral number of nanoseconds.
  • rossleep(t) with t of type Duration, Rate, Real. Also sleep(t::Duration) and sleep(t::Rate) : Sleep the amount implied by type and value of the t parameter.

Publishing Messages

Publishing messages is the same as in rospy, except use the publish method, paired with a Publisher object. For example:

using geometry_msgs.msg
pub = Publisher{PointStamped}("topic", queue_size = 10) #or...
#pub = Publisher("topic", PointStamped, queue_size = 10)
msg = PointStamped()
msg.header.stamp = now()
msg.point.x = 1.1
publish(pub, msg)

The keyword arguments in the Publisher constructor are passed directly on to rospy so anything it accepts will be valid.

Subscribing to a Topic

Subscribing to a topic is the same as in rospy. When creating a Subscriber, an optional callback_args parameter can be given to forward on whenever the callback is invoked. Note that it must be passed as a tuple, even if there is only a single argument. And again, keyword arguments are directly forwarded. An example:

using sensor_msgs.msg
cb1(msg::Imu, a::String) = println(a,": ",msg.linear_acceleration.x)
cb2(msg::Imu) = println(msg.angular_velocity.z)
sub1 = Subscriber{Imu}("topic", cb1, ("accel",), queue_size = 10) #or...
#sub1 = Subscriber("topic", Imu, cb1, ("accel",), queue_size = 10)
sub2 = Subscriber{Imu}("topic", cb2, queue_size = 10)

Using services

ROS services are fully supported, including automatic request and response type generation. For the @rosimport call, use the plain service type name. After rostypegen(), the generated .srv submodule will contain 3 types: the plain type, a request type, and a response type. For example @rosimport nav_msgs.srv.GetPlan will create GetPlan, GetPlanRequest, and GetPlanResponse. To provide the service to other nodes, you would create a Service{GetPlan} object. To call it, a ServiceProxy{GetPlan} object. The syntax exactly matches rospy to construct and use these objects. For example, if myproxy is a ServiceProxy object, it can be called with myproxy(my_request).

Parameter Server

get_param, set_param, has_param, and delete_param are all implemented in the RobotOS module with the same syntax as in rospy.

Message Constants

Message constants may be accessed using getindex syntax. For example for visualization_msgs/Marker.msg we have:

import visualization_msgs.msg: Marker
Marker[:SPHERE] == getindex(Marker, :SPHERE) == 2   # true

ROS Integration

Since Julia code needs no prior compilation, it is possible to integrate very tightly and natively with a larger ROS system. Just make sure you:

  • Keep your code inside your ROS packages as usual.
  • Ensure your .jl script is executable (e.g., chmod a+x script.jl) and has the hint to the Julia binary as the first line (#!/usr/bin/env julia).

Now your Julia code will run exactly like any python script that gets invoked through rosrun or roslaunch. And since include takes paths relative to the location of the calling file, you can bring in whatever other modules or functions reside in your package from the single executable script.

#!/usr/bin/env julia
#main.jl in thebot_pkg/src
using RobotOS

using Bot

Full example

This example demonstrates publishing a random geometry_msgs/Point message at 5 Hz. It also listens for incoming geometry_msgs/Pose2D messages and republishes them as Points.

#!/usr/bin/env julia

using RobotOS
@rosimport geometry_msgs.msg: Point, Pose2D
using geometry_msgs.msg

function callback(msg::Pose2D, pub_obj::Publisher{Point})
    pt_msg = Point(msg.x, msg.y, 0.0)
    publish(pub_obj, pt_msg)

function loop(pub_obj)
    loop_rate = Rate(5.0)
    while ! is_shutdown()
        npt = Point(rand(), rand(), 0.0)
        publish(pub_obj, npt)

function main()
    pub = Publisher{Point}("pts", queue_size=10)
    sub = Subscriber{Pose2D}("pose", callback, (pub,), queue_size=10)

if ! isinteractive()


  • 0.1 : Initial release
  • 0.2 : Changed type gen API and moved generated modules to Main
  • 0.3 : Added service type generation and API
  • 0.4 : Julia v0.4+ support only