README
What Is This?
The demos in this package are designed to showcase how developers can manually compose ROS 2 nodes by defining them separately but combining them in varied process layouts.
These can be done while avoiding code overhauls or performance restrictions.
This package consists of the following ROS 2 nodes:
camera_nodewatermark_nodeimage_view_nodeimage_pipeline_all_in_onetwo_node_pipelinecyclic_pipelineimage_pipeline_with_two_image_view
Through the use of intra-process (as opposed to inter-process) node communication, lower latency and thus higher efficiency is observed for ROS 2 topologies that utilizes this manner of communication.
The improvements in latency reduction are even more pronounced when applied to ROS 2 systems with more complex topologies.
Build
colcon build --packages-up-to intra_process_demo
Run
1. Two Node Pipeline
Run two_node_pipeline via the commands below:
ros2 run intra_process_demo two_node_pipeline
This sets up two nodes, a ROS 2 node that publishes a string with an incremeting integer value, as well as a ROS 2 node that subscribes to print the received string.
2. Cyclic Pipeline
Run cyclic_pipeline via the commands below:
ros2 run intra_process_demo cyclic_pipeline
Similar to the previous, instead of creating a new message for each new iteration, the publisher and subscriber nodes only ever use one message instance. This is achieved by having a cycle in the graph and kickstarting the communication externally by having one of the nodes publish before spinning the executor.
3. Image Pipeline All In One
Please ensure you have a camera connected to your workstation.
image_pipeline_all_in_one consists of 3 nodes, where camera_node publishes a unique_ptr message onto the topic, \image. Subsequently, watermark_node subscribes to the previous, republishes the image after adding a watermark on the topic, \watermarked_image, and the final node, image_view_node, subscribes to this last topic in order to display it for user.
ros2 run intra_process_demo image_pipeline_all_in_one
4. Image Pipeline All Separately
Please ensure you have a camera connected to your workstation.
In direct contrast with the previous, run the following commands in separate terminals to have camera_node, watermark_node and image_view_node all in their own process, utilizing inter-process node communication.
This starts the camera_node ROS 2 node and publishes images captured from your workstation web camera onto a ROS 2 topic labelled /image.
# Open new terminal
ros2 run intra_process_demo camera_node
This starts the watermark_node ROS 2 node which subscribes to raw images from ROS 2 topic /image, overlays both process ID number and message address on top of the image visually and publishes to ROS 2 topic /watermarked_image.
# Open new terminal
ros2 run intra_process_demo watermark_node
This starts the image_view_node ROS 2 node which subscribes to /watermarked_image and displays the received images in an OpenCV GUI window.
# Open new terminal
ros2 run intra_process_demo image_view_node
5. Image Pipeline With Two Image Views
Please ensure you have a camera connected to your workstation.
Similar to the Image Pipeline All In One, running image_pipeline_with_two_image_view will display the image process through intra-process communications.
However, it now instantiates 2 image_view_node ROS 2 nodes.
# Open new terminal
ros2 run intra_process_demo image_pipeline_with_two_image_view
Verify
1. Two Node Pipeline
When executed correctly, strings should be printed in the terminal similar to what is shown below:
Published message with value: 0, and address: 0x55B68BCC6F20
Received message with value: 0, and address: 0x55B68BCC6F20
Published message with value: 1, and address: 0x55B68BCC6F20
Received message with value: 1, and address: 0x55B68BCC6F20
Published message with value: 2, and address: 0x55B68BCC6F20
Received message with value: 2, and address: 0x55B68BCC6F20
Published message with value: 3, and address: 0x55B68BCC6F20
Received message with value: 3, and address: 0x55B68BCC6F20
2. Cyclic Pipeline
When executed correctly, strings should be printed in the terminal similar to what is shown below:
Published first message with value: 42, and address: 0x555E4F029480
Received message with value: 42, and address: 0x555E4F029480
sleeping for 1 second...
done.
Incrementing and sending with value: 43, and address: 0x555E4F029480
Received message with value: 43, and address: 0x555E4F029480
sleeping for 1 second...
done.
Incrementing and sending with value: 44, and address: 0x555E4F029480
Received message with value: 44, and address: 0x555E4F029480
sleeping for 1 second...
done.
Incrementing and sending with value: 45, and address: 0x555E4F029480
Received message with value: 45, and address: 0x555E4F029480
sleeping for 1 second...
3. Image Pipeline All In One
When executed correctly, an OpenCV GUI window should appear displaying something similar to what is shown below:
Take note how the process_id and Message Pointer Address are all the same, proving that all 3 nodes are in the same process.
4. Image Pipleline All Separately
When executed correctly, an OpenCV GUI window should appear displaying something similar to what is shown below:
Notice how all the process_id and Message Pointer Address are now all different, showing that all nodes are using different processes.
5. Image Pipeline With Two Image Views
When executed correctly, 2 OpenCV GUI window should appear displaying similar to what is show below:
For more details on this implementations, please refer to the references below.