Call for Actions: Action Servers and Clients

Call for Actions: Action Servers and Clients#

Warning

This topic is under construction and this might not even be its final form. Please feel free to open an issue if you spot any typos or other problems.

Added in version Jazzy: Added this section.

What about a mixture of messages and services? That is where actions come into play.

We use actions by creating an ActionServer. The ActionServer can be called by one or more ActionClients.

Similarly to a service, each action should only have a single ActionServer that will receive a goal (with a Request) and provide a Result. It will also provide Feedback through a suitable topic.

It can be argued that the main difference between a service and an action is the capability of providing feedback while the action is performed. A service, in contrast, only outputs a single, final result of the service call.

Objective#

Note

Unless otherwise explicitly stated, we will always follow the International System of Units. This means distances in meters, angles in radians, time in seconds, and so on.

We can use an action to represent the first robot-like behavior of this tutorial in an illustrative manner.

Suppose that we have a simple robot that moves in 2D space and whose orientation is not important. Let it have a current position with respect to the \(k\)-ith iteration represented by (a 2D vector)

\[\begin{split}\mathbb{R}^2 \ni \boldsymbol{p}(k) =\begin{bmatrix}x(k) \\ y(k)\end{bmatrix}\end{split}\]

and a desired position given by

\[\begin{split}\mathbb{R}^2 \ni \boldsymbol{p}_d =\begin{bmatrix}x_d \\ y_d\end{bmatrix}.\end{split}\]

Suppose that we want to design an action server that takes this robot-like object from its current position \(\boldsymbol{p}\) and moves it towards the goal \(\boldsymbol{p}_d\) with a speed \(s \in \mathbb{R}\). As feedback, it gives us the distance \(d \in \mathbb{R}\) between the current position and the desired position.

The (Euclidean) distance \(d\) will be calculated from the terms \(x\), \(x_d\), \(y\), and \(y_d\), as follows.

(1)#\[d = \sqrt{(x-x_d)^2 + (y-y_d)^2}.\]

The action server will update the position \(\boldsymbol{p}(k)\) based on a simple constant speed motion of the robot. This can be mathematically described as follows.

(2)#\[\boldsymbol{p}(k+1) = \boldsymbol{p}(k) - s\left(\frac{\boldsymbol{p} - \boldsymbol{p}_d}{d}\right)T,\]

where \(\boldsymbol{p}(k+1)\) represents the next position, \(\boldsymbol{p}(k)\) the current position, and \(T\) is the sampling time.

Create the package#

We start by creating a package to use the action we first created in The action file.

cd ~/ros2_tutorial_workspace/src
ros2 pkg create python_package_that_uses_the_actions \
--build-type ament_python \
--dependencies rclpy package_with_interfaces

Overview#

This will be the file structure for the Action tutorial. Highlighted are the main files for the ActionServer and ActionClient.

File structure

python_package_that_uses_the_actions/
|-- package.xml
|-- python_package_that_uses_the_actions
|   |-- __init__.py
|   |-- move_straight_in_2d_action_client_node.py
|   `-- move_straight_in_2d_action_server_node.py
|-- resource
|   `-- python_package_that_uses_the_actions
|-- setup.cfg
|-- setup.py
`-- test
    |-- test_copyright.py
    |-- test_flake8.py
    `-- test_pep257.py