[ROS2 How-to] #2 – Create a ROS2 action server

Written by Ruben Alves


What we are going to learn

  1. – How to create a custom action message
  2. How to create an action server

List of resources used in this post

  1. Use the rosject: https://app.theconstructsim.com/#/l/4a1c58c5/
  2. The Construct: https://app.theconstructsim.com/
  3. ROS2 Courses –▸
    1. ROS2 Basics in 5 Days (Python): https://app.theconstructsim.com/#/Course/73
    2. ROS2 Basics in 5 Days (C++): https://app.theconstructsim.com/#/Course/61
    3. ROS2 Navigation training: https://www.theconstructsim.com/ros2-navigation-training/

What is a ROS2 Action

Let’s assume you wish to wash your clothing. There are two possible ways you could go about it:

  1. Go to the Laundry service provider
    1. Put your clothes to wash.
    2. Wait until the clothes are washed.
    3. Get your clothes.
  2. If you have a washing machine at home:
    1. Put your clothes to wash
    2. Instead of waiting, you can do other things and leave the watching machine doing its jobs
    3. Check once in a while if the clothes are finished
    4. Do other things.
    5. Clothes are washed.

Option 1 is a blocking activity because you have to wait (in theory not able to do anything else) for the clothes to be washed, while option 2 is non-blocking because you can do some other things while your clothes are being washed.

This non-blocking is what defines an Action. If ROS2 Services are for instant request-responses, an Action is a task that may take a lot of time to be finished, and in the meantime, a robot (or you) is free to do other things and is also able to constantly check the status of the action.

Opening the rosject

In order to learn how to create an and use an Action Server in ROS2, we need to have ROS2 installed in our system, and it is also useful to have some simulations. To make your life easier, we already prepared a rosject with a simulation for that: https://app.theconstructsim.com/#/l/4a1c58c5/.

You can download the rosject on your own computer if you want to work locally, but just by copying the rosject (clicking the link), you will have a setup already prepared for you.

After the rosject has been successfully copied to your own area, you should see a Run button. Just click that button to launch the rosject (below you have a rosject example).

Learn ROS2 Parameters - Run rosject

Learn ROS2 – Run rosject (example of the RUN button)

After pressing the Run button, you should have the rosject loaded. Let’s now head to the next section to really get some real practice.

Launching the simulation

The rosject we provided contains the packages needed to run a TurtleBot3 simulation in ROS2. The TurtleBot3 has the following sensors:

  • Lidar
  • IMU

Feel free to use this rosject to test your mobile robot programs.

The rosject is structured the following way:

  • turtlebot3_ws: this workspace contains the TurtleBot3 packages provided by ROBOTIS. Don’t modify this unless you know what you are doing and want to change something from the simulation
  • Use this workspace to develop your programs

Assuming you have opened the rosject by clicking the Run button, we can launch the simulation with:

source turtlebot3_ws/install/setup.bash
source turtlebot3_ws/install/setup.bash
export TURTLEBOT3_MODEL=burger
ros2 launch turtlebot3_gazebo turtlebot3_world.launch.py
After a few seconds, the simulation should have opened automatically:
[ROS2 How-to] #2 - Create a ROS2 action server - Simulation running

[ROS2 How-to] #2 – Create a ROS2 action server – Simulation running

In case the simulation does not pop up automatically, you can easily click the Open Gazebo button like in the example below (bear in mind that the simulation below is not the one used in this tutorial. Its main purpose is to show the Open Gazebo button):

Open Gazebo by clicking Open Gazebo

Open Gazebo by clicking Open Gazebo


Creating our ROS2 package (later used to create our Action Server)

Let’s create our ROS2 Package. For that, let’s start by opening a new terminal:

Open a new Terminal

Open a new Terminal


In the terminal that was just open, by running the “ls”  command you can see that we have at least the following folders:

ros2_ws  turtlebot3_ws


The turtlebot3_ws contains the simulation, and the ros2_ws is where we are going to place our code.

Before you continue, it is worth mentioning that in the rosject that we shared with you, the custom_interfaces package that we are going to create here already exists. We are going to create it here basically for learning purposes. You would actually not need it since the package was already created for you:

Let’s create a package named custom_interfaces with the action_msgs std_msgs rosids_default_generators  packages as dependencies:

cd ~/ros2_ws/src/

ros2 pkg create custom_interfaces2 --build-type ament_cmake --dependencies action_msgs std_msgs rosidl_default_generators

If everything went ok, you should see the following:

going to create a new package
package name: custom_interfaces
destination directory: /home/user/ros2_ws/src
package format: 3
version: 0.0.0
description: TODO: Package description
maintainer: ['user <user@todo.todo>']
licenses: ['TODO: License declaration']
build type: ament_cmake
dependencies: ['action_msgs', 'std_msgs', 'rosidl_default_generators']
creating folder ./custom_interfaces
creating ./custom_interfaces/package.xml
creating source and include folder
creating folder ./custom_interfaces/src
creating folder ./custom_interfaces/include/custom_interfaces2
creating ./custom_interfaces/CMakeLists.txt


After the package was created, let’s create a folder called action:

mkdir -p custom_interfaces/action/

and also create the action/Patrol.action file.

touch custom_interfaces/action/Patrol.action

This is the file/Interface that we will use in our Action Server for patrolling.

Let’s now open that Patrol.action file. You can open it in the Code Editor. If you do not know how to open the Code Editor, please check the image below:

Open the IDE - Code Editor

Open the IDE – Code Editor

You can now open the custom_interfaces/action/Patrol.action file and paste the following content on it:

float32 radius
bool success
float32 time_left


Now, to be able to compile our message file, we have to open the custom_interfaces/CMakeLists.txt file and paste the following content around line 14:


    DEPENDENCIES action_msgs std_msgs
In the end, the final CMakeLiss.txt file would look like the following:
cmake_minimum_required(VERSION 3.8)

  add_compile_options(-Wall -Wextra -Wpedantic)

# find dependencies
find_package(ament_cmake REQUIRED)
find_package(action_msgs REQUIRED)
find_package(std_msgs REQUIRED)
find_package(rosidl_default_generators REQUIRED)


  DEPENDENCIES action_msgs std_msgs

  find_package(ament_lint_auto REQUIRED)
  # the following line skips the linter which checks for copyrights
  # uncomment the line when a copyright and license is not present in all source files
  #set(ament_cmake_copyright_FOUND TRUE)
  # the following line skips cpplint (only works in a git repo)
  # uncomment the line when this package is not in a git repo
  #set(ament_cmake_cpplint_FOUND TRUE)



And for the file custom_interfaces/package.xml we also have to add the following code before the <export> tag:


In the end, our package.xml would look like the following:

<?xml version="1.0"?>
<?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?>
<package format="3">
  <description>TODO: Package description</description>
  <maintainer email="user@todo.todo">user</maintainer>
  <license>TODO: License declaration</license>






Please make sure you save the files with Ctrl+S after making the modifications.

Compiling our custom Action interface

Now that we defined our Custom Action interface, let’s compile it.

Let’s go to the first terminal we opened previously and run the following commands:

cd ~/ros2_ws/

colcon build

The package should have been compiled with no errors:

Starting >>> custom_interfaces
Finished <<< custom_interfaces [9.53s]

Summary: 1 package finished [9.53s]


Let’s now make sure ROS2 can find our Action interface:

source install/setup.bash

ros2 interface show custom_interfaces/action/Patrol

It should show:

float32 radius
bool success
float32 time_left


So far so good. ROS is able to find our custom interface.

The time has now come for us to create the Action Server.

Creating our ROS2 Action Server

Let’s create a different package for the Action Server, just to keep things separated. Since we are not doing to create Interfaces in this new package, just use existing interfaces, let’s use the ament_python build type. Again, bear in mind that if you are using the rosject that we provided, the package already exists in the ~/ros2_ws/src folder:

cd ~/ros2_ws/src/
ros2 pkg create --build-type ament_python patrol_action_server --dependencies rclpy geometry_mgs custom_interfaces

The logs should be similar to the following:

going to create a new package
package name: patrol_action_server
destination directory: /home/user/ros2_ws/src
package format: 3
version: 0.0.0
description: TODO: Package description
maintainer: ['user <user@todo.todo>']
licenses: ['TODO: License declaration']
build type: ament_python
dependencies: ['rclpy', 'geometry_mgs', 'custom_interfaces']
creating folder ./patrol_action_server
creating ./patrol_action_server/package.xml
creating source folder
creating folder ./patrol_action_server/patrol_action_server2
creating ./patrol_action_server/setup.py
creating ./patrol_action_server/setup.cfg
creating folder ./patrol_action_server/resource
creating ./patrol_action_server/resource/patrol_action_server
creating ./patrol_action_server/patrol_action_server/__init__.py
creating folder ./patrol_action_server/test
creating ./patrol_action_server/test/test_copyright.py
creating ./patrol_action_server/test/test_flake8.py
creating ./patrol_action_server/test/test_pep257.py

Now that our package is created, let’s create a file patrol_action_server.py that will have the code of our Action Server:

touch patrol_action_server/patrol_action_server/patrol_action_server.py

Let’s now open that file using the Code Editor, and paste the following content to it:

#!/usr/bin/env python3
# Copyright 2019 ROBOTIS CO., LTD.
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#     http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# See the License for the specific language governing permissions and
# limitations under the License.
# Authors: Ryan Shim, Gilbert

import math
import time

import rclpy

from geometry_msgs.msg import Twist

from rclpy.action import ActionServer
from rclpy.action import CancelResponse
from rclpy.action import GoalResponse
from rclpy.callback_groups import ReentrantCallbackGroup
from rclpy.duration import Duration
from rclpy.node import Node
from rclpy.qos import QoSProfile

from rclpy.executors import MultiThreadedExecutor

from custom_interfaces.action import Patrol

class Turtlebot3PatrolServer(Node):

    def __init__(self):

        self.goal = Patrol.Goal()

        qos = QoSProfile(depth=10)

        # Initialise publishers
        self.cmd_vel_pub = self.create_publisher(Twist, 'cmd_vel', qos)

        # Initialise servers
        self._action_server = ActionServer(

        self.get_logger().info("Turtlebot3 patrol action server has been initialised.")

    def destroy(self):

    def goal_callback(self, goal_request):
        # Accepts or rejects a client request to begin an action
        self.get_logger().info('Received goal request :)')
        self.goal = goal_request
        return GoalResponse.ACCEPT

    def cancel_callback(self, goal_handle):
        # Accepts or rejects a client request to cancel an action
        self.get_logger().info('Received cancel request :(')
        return CancelResponse.ACCEPT

    async def execute_callback(self, goal_handle):
        self.get_logger().info('Executing goal...')

        radius = self.goal.radius  # unit: m
        speed = 0.5  # unit: m/s

        feedback_msg = Patrol.Feedback()
        total_driving_time = 2 * math.pi * radius / speed
        feedback_msg.time_left = total_driving_time
        last_time = self.get_clock().now()

        # Start executing an action
        while (feedback_msg.time_left > 0):
            if goal_handle.is_cancel_requested:
                self.get_logger().info('Goal canceled')
                return Patrol.Result()

            curr_time = self.get_clock().now()
            duration = Duration()
            duration = (curr_time - last_time).nanoseconds / 1e9  # unit: s

            feedback_msg.time_left = total_driving_time - duration
            self.get_logger().info('Time left until the robot stops: {0}'.format(feedback_msg.time_left))

            # Give vel_cmd to Turtlebot3
            twist = Twist()
            twist = self.drive_circle(radius, speed)

            # Process rate
            time.sleep(0.010)  # unit: s

        # When the action is completed
        twist = Twist()

        result = Patrol.Result()
        result.success = True
        self.get_logger().info('Returning result: {0}'.format(result.success))

        return result

    def drive_circle(self, radius, velocity):
        self.twist = Twist()
        self.linear_velocity = velocity  # unit: m/s
        self.angular_velocity = self.linear_velocity / radius  # unit: rad/s

        self.twist.linear.x = self.linear_velocity
        self.twist.angular.z = self.angular_velocity

        return self.twist

def main(args=None):

    patrol_action_server = Turtlebot3PatrolServer()

    # Use a MultiThreadedExecutor to enable processing goals concurrently
    executor = MultiThreadedExecutor()

    rclpy.spin(patrol_action_server, executor=executor)


if __name__ == '__main__':

The code used above is just an adaptation of a code already provided by ROBOTIS.
Before we compile our code, we also have to open the patrol_action_server/setup.py file and modify the entry_points section to define our executable called patrol_action_server_exe in the following way:
        'console_scripts': [
            'patrol_action_server_exe = patrol_action_server.patrol_action_server:main',


In the end, the complete ~/ros2_ws/src/patrol_action_server/setup.py would be as follows:

from setuptools import setup

package_name = 'patrol_action_server'

            ['resource/' + package_name]),
        ('share/' + package_name, ['package.xml']),
    description='TODO: Package description',
    license='TODO: License declaration',
        'console_scripts': [
            'patrol_action_server_exe = patrol_action_server.patrol_action_server:main',


Compiling our ROS2 Action Server

With everything in place, we compile our package just as before:

cd ~/ros2_ws/

colcon build
Starting >>> custom_interfaces
Finished <<< custom_interfaces [9.53s]
Starting >>> patrol_action_server
Finished <<< patrol_action_server [5.33s]

Summary: 2 packages finished [15.2s]
We can now run our server with the following commands:
source install/setup.bash

ros2 run patrol_action_server patrol_action_server_exe
The server should start with no problems:
[INFO] [1651528559.914166370] [turtlebot3_patrol_server]: Turtlebot3 patrol action server has been initialised

Calling our ROS2 Action Server

Ok, if you did not kill the Action Server launched in the previous section, please open a second terminal that we will use to call the Action Server.

With “ros2 node list” we should be able to find our node running:

ros2 node list

And with ros2 action list, we should be able to see the /patrol action:
ros2 action list

We can now call our Action Server. If you remember when we created the Patrol.action, we defined a radius. Let’s them call the Action Server passing a radius of 0.5. The robot will be basically rotating:
ros2 action send_goal --feedback /patrol custom_interfaces/action/Patrol radius:\ 0.5\
You should now see the feedback sent by the action server:
Waiting for an action server to become available...
Sending goal:
     radius: 0.5

Goal accepted with ID: dd32bc835d7a4ef5ae854d0bfb4b119f

    time_left: 6.2831525802612305

    time_left: 6.271763801574707

    time_left: 6.260392665863037

    time_left: 6.2484917640686035

    time_left: 6.237414836883545

    time_left: 6.2265496253967285

    time_left: 6.215761661529541

^CCanceling goal...
time_left: 5.634908676147461
Goal canceled.

Remember that you can easily cancel the call to the action server by pressing CTRL+C.

If you look at the simulation after sending a goal to the Action Server, you should see the robot spinning around 0.5 meters.

Congratulations. You now know how to create a ROS2 Action Server from scratch. If you want more details about the code of the Action Server, please check the video in the next section.

Youtube video

So this is the post for today. Remember that we have the live version of this post on YouTube. If you liked the content, please consider subscribing to our youtube channel. We are publishing new content ~every day.

Keep pushing your ROS Learning.

Related Courses & Training

If you want to learn more about ROS and ROS2, we recommend the following courses:

Check Out These Related Posts


Pin It on Pinterest

Share This