ROS node

A ROS node is a single process or component in a ROS system that performs one task, such as reading a camera, estimating pose, or controlling a motor, and communicates with other nodes through topics, services, actions, and parameters. A robot's software is a graph of many nodes, and recording the messages they publish is what produces a rosbag or MCAP log.

What is a ROS node?

ROS 2 breaks robot software into small programs called nodes. One node might publish camera images, another might detect objects in them, and a third might plan a path. Each node does one job, and together they form the ROS graph, a network of nodes passing messages.
Nodes are written with ROS client libraries, mainly rclcpp for C++ and rclpy for Python. A node can run as its own process, or several nodes can share one process through composition, which cuts the cost of passing large messages such as images.

Key takeaways

  • A ROS node is one program, or component, that does one job in a robot's software.
  • Nodes talk through topics, services, actions, and parameters.
  • The messages nodes publish on topics are what a rosbag or MCAP recording captures.

How it works

Nodes communicate in four main ways:
  • Topics are named streams, such as /camera/image or /joint_states. A node publishes messages on a topic, and any number of nodes subscribe to it.
  • Services are request and response calls for quick, one-off tasks.
  • Actions handle longer tasks, such as moving an arm to a pose, with progress feedback and the option to cancel.
  • Parameters are settings a node exposes, such as a camera's frame rate.
In ROS 2, nodes find each other automatically through the middleware, which is DDS by default, so there is no ROS 1 style master process. Each message has a defined type, such as sensor_msgs/msg/Image, which is also the schema a recording stores alongside the data. The ros2 node list and ros2 node info commands show which nodes are running and what they publish and subscribe to.
Each topic a node publishes becomes a channel in the recorded log.

See which streams a robot log recorded, and which ones went quiet

A robot log is only as reliable as the nodes that wrote it. Watch for:
  • topics that stop publishing partway through a run
  • streams recorded at a lower rate than expected
  • timestamps from nodes whose clocks were not synchronized
In Voxel51, DROID holds 1,000 real-world manipulation episodes as MCAP, with three synced cameras and robot state on one timeline.

Robotics Droid

1,000 episodes

Real-world robot manipulation episodes from DROID, stored as MCAP with three synced cameras and fields for task prompt, success, and number of grasps.

Try this: Open an episode, open the Topics tab, and list which streams were recorded. On a ROS robot, each one would come from a node.

Explore DROID in Voxel51

Free with a work or .edu email.

Go deeper

Why it matters

A robot log is a record of what its nodes said. When you record with rosbag2 into MCAP, each topic becomes a channel, and many problems in the data trace back to a node: a camera node that dropped frames, two nodes on machines with clocks out of sync, or a node that published on the wrong topic. Knowing which node produced a stream is often the fastest way to explain why it looks wrong.
A robot log is a record of what its nodes said.

Frequently asked questions

What is the difference between a ROS node and a topic?

A node is a program. A topic is a named channel that nodes publish messages to and subscribe to. One node can publish and subscribe to many topics.

How many nodes does a robot run?

It depends on the robot. A simple arm might run a handful, while a mobile robot with navigation, perception, and manipulation can run dozens or more.

Do ROS 2 nodes need a master?

No. ROS 1 used a central master to connect nodes. ROS 2 nodes discover each other automatically through the middleware, which is DDS by default.
black and white photo of Jesse Mostipak
Director of Growth
Jesse Mostipak is the Director of Growth at Voxel51, where the work is helping humans find and trust what the brand knows, and teaching the Google knowledge graph and the LLMs answering on their behalf to do the same. That question, how knowledge gets built inside a system, is one Jesse has been chasing for years. Earlier versions of it ran through a New York City high school science classroom, data science and machine learning, and developer relations at Kaggle, Posit (formerly RStudio), and Baseten. The answer doesn't change much depending on whether the learner is a teenager, a software engineer, or a knowledge graph. Jesse holds a Master's in Education from CUNY Hunter College. LinkedIn
See all articles by Jesse Mostipak
Last updated November 19, 2026

Building visual or physical AI?

Let's talk.