Skip to main content

ROS 2 Nodes and Topics

Why ROS 2?​

A humanoid robot is a distributed system. Its cameras, actuators, planners, and controllers are separate processes that must communicate in real time. ROS 2 (Robot Operating System 2) provides the middleware that connects them.

Think of ROS 2 as the nervous system of the robot: nodes are neurons, and topics are the axons that carry signals between them.

Nodes​

A node is a single-purpose process in the ROS 2 computation graph. Each node handles one responsibility:

import rclpy
from rclpy.node import Node

class MinimalPublisher(Node):
    def __init__(self):
        super().__init__('minimal_publisher')
        self.publisher_ = self.create_publisher(String, 'topic', 10)
        timer_period = 0.5  # seconds
        self.timer = self.create_timer(timer_period, self.timer_callback)

    def timer_callback(self):
        msg = String()
        msg.data = 'Hello from the robot nervous system'
        self.publisher_.publish(msg)

Key properties of nodes:

  • Single responsibility: one node, one job
  • Discoverable: nodes announce themselves on the ROS 2 network
  • Lifecycle-managed: nodes can be started, configured, activated, and shut down programmatically

Topics​

Topics are named buses for asynchronous message passing. A publisher sends messages to a topic; any number of subscribers can listen.

# List all active topics
ros2 topic list

# Echo messages on a topic
ros2 topic echo /camera/image_raw

The publish-subscribe pattern decouples producers from consumers. The camera node does not need to know who is reading its images.

Putting It Together​

A minimal humanoid perception pipeline:

[Camera Node] --/image_raw--> [Perception Node] --/detected_objects--> [Planner Node]

Each arrow is a topic. Each box is a node. The system scales by adding nodes without modifying existing ones.

Next Steps​

In the next section, we cover Services and Actions — synchronous request-response patterns for when publish-subscribe is not enough.