News in Short
- Nvidia has released Isaac ROS 5.0 as free and open-source software.
- The update brings AI-agent capabilities to robotics development.
- Developers get reusable skills for robot setup, perception and manipulation.
- FoundationStereo can be fine-tuned for specific cameras and environments.
- FoundationPose gets an agent-ready inference library with faster object tracking.
Nvidia has launched Isaac ROS 5.0, the latest version of its open-source robotics software platform. The update brings AI-agent capabilities to the ROS ecosystem and introduces reusable workflows for robot development.
The company says the release is designed to simplify some of the more complex stages of building robotics applications. Developers can use reusable skills and documentation to build, customise and deploy robot applications more efficiently.
Isaac ROS 5.0 adds reusable AI skills
One of the main changes in Isaac ROS 5.0 is the addition of agent-ready documentation and reusable skills. These tools cover tasks such as robot setup and manipulation.
AI agents can use the documentation to work with Isaac ROS tools and workflows. Nvidia says this approach can help developers turn requirements into working applications faster.
The shift is significant because developers can use these skills across different stages of robotics development. Instead of treating every task as a separate development exercise, reusable workflows can provide a starting point for common robotics operations.
AI agents can now work with robotics perception
Isaac ROS 5.0 also introduces new tools for perception-related development. One example is the FoundationStereo fine-tuning skill.
It allows an AI agent to help adapt the stereo perception model to a developer’s cameras, environment and specific robotics application. Nvidia says this can improve perception accuracy for a particular sensor configuration.
The update also includes an agent-ready inference library for FoundationPose. The model is designed to estimate and track the position and orientation of objects. Nvidia claims the new implementation can track objects up to 5.5 times faster.
Pick-and-place becomes a reusable robotics skill
Another addition is an agent-ready pick-and-place skill. Pick-and-place is a common robotics task that involves identifying, moving and positioning objects.
By making it available as a reusable workflow, Nvidia is giving developers a ready-made building block for robotic manipulation. The company says these skills allow AI agents to work with specific robotics models and workflows rather than limiting them to individual coding tasks.
This could make the development process more modular. Developers can use existing skills while adapting them to the needs of a particular robot or application.
Support for ROS Lyrical and Ubuntu 24.04
Isaac ROS 5.0 also updates its underlying software support. The platform now supports ROS Lyrical and Ubuntu 24.04.
Nvidia worked with the Open Source Robotics Alliance to contribute a standard data-handling interface to ROS Lyrical. The interface is designed to help robotics software work across different computing hardware, including GPUs.
According to Nvidia, the interface gives developers a consistent way to accelerate robotics applications across supported hardware. This is particularly relevant as robotics workloads increasingly combine CPUs, GPUs and specialised AI hardware.
Isaac ROS 5.0 extends to Jetson hardware
The new release also connects the development workflow with Nvidia’s Jetson platform. Support extends from the Jetson Orin Nano to the higher-performance Jetson Thor.
This gives developers a path from building and testing robotics applications to running more complex workloads on compatible Jetson hardware.
Nvidia Isaac ROS 5.0 is available now as free and open-source software. With AI agents, reusable robotics skills and updated ROS support, the release brings more automation into the software side of robot development.
