NVIDIA Isaac ROS 5.0 Adds AI Agent Workflows and GPU Acceleration for Advanced Robotics

Developers building robots that can perceive, reason and operate in dynamic environments need new physical AI models, tools and workflows. NVIDIA Isaac ROS 5.0 brings those capabilities to the open-source Robot Operating System (ROS) ecosystem with GPU-accelerated packages, agent-enabled development tools and support for the latest ROS platform.

Announced at ROSCon in Toronto, Canada, Isaac ROS 5.0 helps developers and AI agents build, customize and deploy robotics applications faster. The release combines ROS, NVIDIA accelerated computing, physical AI models and production-ready libraries for the approximately 1.3 million ROS users worldwide.

ROS provides the open-source foundation for much of modern robotics development, including common tools, libraries and standards for building and connecting robotics applications. NVIDIA Isaac ROS extends that foundation with free, familiar open-source tools designed for high-performance robotics.

What’s new in NVIDIA Isaac ROS 5.0?

Isaac ROS 5.0 introduces AI agent workflows for robotics development, support for ROS Lyrical and Ubuntu 24.04, new perception and manipulation capabilities, and a path to deploy physical AI applications across NVIDIA Jetson platforms.

• AI agent workflows for setup, operations and robot application development
• Agent-enabled documentation and reusable Isaac ROS skills
• Support for ROS Lyrical and Ubuntu 24.04
• Faster FoundationPose inference for object recognition and tracking
• FoundationStereo fine-tuning for customized stereo perception
• Standalone Pick and Place workflows
• GPU acceleration through CUDA and NVIDIA hardware
• Support for platforms ranging from NVIDIA Jetson Orin Nano to NVIDIA Jetson Thor

AI agents are changing robotics development

AI agents are changing how software is built by automating repetitive tasks, navigating complex codebases and helping developers move from ideas to working applications more quickly. Isaac ROS 5.0 brings these capabilities to robot development.

New NVIDIA Isaac skills for setup and operations provide reusable workflows that developers and AI agents can use to complete common robot development tasks. Agent-enabled documentation helps AI agents understand Isaac ROS tools and workflows, allowing developer intent to translate into application behavior more quickly.

Some skills support more than individual coding tasks. The new FoundationStereo fine-tuning skill enables an AI agent to adapt the stereo recognition model to a developer’s camera, environment and robotics application. This can help developers achieve more accurate recognition for specific sensor configurations.

FoundationPose, NVIDIA’s foundational model for object pose estimation and tracking, now includes an agent-enabled inference library. It enables robots to recognize and track object position and orientation up to 5.5 times faster.

Pick and Place, a common workflow that connects detection, depth estimation and pose output, is also available as a standalone agent-enabled skill. This gives robot developers greater flexibility when building manipulation applications with Isaac ROS.

Support for ROS Lyrical and Ubuntu 24.04

Isaac ROS 5.0 adds support for ROS Lyrical and Ubuntu 24.04, giving developers a path to adopt the latest ROS platform while continuing to accelerate demanding robotic workloads with NVIDIA computing.

NVIDIA is also contributing to the Open Source Robotics Alliance. Its standard data processing interface for ROS Lyrical enables robot software to work efficiently across different types of computing hardware, including GPUs.

Available to the broader ROS community, the interface provides a consistent way to accelerate demanding robotics applications using CUDA and other GPU acceleration technologies.

Accelerating the open-source robotics ecosystem

The robotics ecosystem is already extending the agent approach across development workflows, hardware platforms and industrial applications.

AgenticROS is an open-source project sponsored by RealSense. It connects Isaac ROS with the open NVIDIA Nemotron model and NVIDIA NemoClaw blueprints, enabling AI agents to interact with ROS-based robots.

RealSense has also optimized AI-native 3D stereo depth cameras, including the RealSense D585 Pro, and open-source software development kits for Isaac ROS and NVIDIA Jetson Thor edge AI platforms. These tools help developers build perception, navigation and manipulation applications.

Intrinsic’s Open Machine Tending Solution is a reference application for computer numerical control machine tending. Intrinsic’s Intrinsic Core is an open-source suite of preconfigured runtime services and features designed to accelerate industrial robotics applications.

Intrinsic Core includes compatibility with NVIDIA FoundationPose. Its object registration, tracking and pose estimation capabilities help robots dynamically detect and process parts, reducing the need for rigid and expensive physical fixtures and specialized system integration.

Robotics companies building with Isaac ROS

Seeed Studio’s reBot Arm uses NVIDIA Isaac ROS and combines perception, spatial understanding and motion planning accelerated with NVIDIA Jetson Thor. The integration provides a practical platform for building physical AI applications, including object localization, collision-aware maneuvering and autonomous pick-and-place.

Magna uses NVIDIA Isaac ROS as a modular, GPU-accelerated foundation for robot recognition, synchronous data acquisition and NVIDIA Isaac GR00T model deployment. Combined with Isaac Sim hardware-in-the-loop testing, the platform supports intelligent automation from research to real-world manufacturing and mobility.

Prefix.dev’s Pixi package management tool connects ROS and NVIDIA CUDA, helping developers set up and share accelerated robotics workflows more easily.

As an Isaac ROS partner, Foxglove enables developers to visualize and debug live ROS applications through web and desktop tools. Its integrations support 3D topics, nvblox meshes and rosbags throughout Isaac ROS tutorials and workflows.

Flexiv integrates Isaac ROS with Rizon 4 adaptive robots, giving developers access to NVIDIA-accelerated robot capabilities and a streamlined path from testing applications in NVIDIA Isaac Sim to deploying them on physical robots.

Ekumen, a Grid Dynamics company, uses the GPU-accelerated Isaac ROS package within its existing ROS and Nav2 stack. The implementation supports high-precision docking, 3D obstacle detection, visual localization and real-time motion planning for validating Isaac Sim applications.

https://www.youtube.com/watch?v=RL4XL8EAIUw" title="Ekumen Isaac ROS demonstration

Ekumen uses isaac_ros_cumotion on the GPU to map collision-free paths for warehouse arms in about 2–5 milliseconds.

Stereolabs integrates ZED stereo cameras with NVIDIA Isaac ROS to provide GPU-accelerated perception for robotics applications. The integration simplifies the development of real-time object detection, mapping and navigation while maintaining interoperability with the broader ROS ecosystem.

From physical AI development to robot deployment

Applications built by developers and AI agents ultimately need to run on robots. NVIDIA Jetson is a scalable computing platform for running physical AI stacks at the edge with real-time performance. It integrates ROS, accelerated perception and navigation, AI models and application logic on robots.

Isaac ROS 5.0 supports computing platforms ranging from the entry-level NVIDIA Jetson Orin Nano to the high-performance NVIDIA Jetson Thor. As robotic workloads become more sophisticated, these devices provide developers with a path from development to deployment.

Robotics companies are already using this combination to bring more AI processing directly to their machines.

Mentee Robotics uses NVIDIA Isaac ROS as the perception and AI backbone for the MenteeBot humanoid. The platform allows the robot to interpret visual information and perform learned behaviors in real time. A shared software foundation across NVIDIA Jetson Orin and Jetson Thor platforms helps Mentee extend innovation from existing robots to next-generation systems.

Universal Robots has incorporated NVIDIA Isaac ROS into its AI Accelerator software development kit. The SDK helps integrators deploy advanced perception and motion capabilities without developing complex robot software from scratch. Powered by NVIDIA Jetson at the edge, the solution allows robots to adapt to less precisely positioned parts, reduce reliance on expensive fixtures and make manufacturing cells more flexible.

ROBOTIS, which builds the ROS-based TurtleBot3 for developers, integrates Isaac ROS into its AI Worker robot. GPU-accelerated object recognition enables visually guided manipulation tasks such as picking, placement and alignment.

https://www.youtube.com/watch?v=fmZdMV72IR0" title="ROBOTIS AI Worker Isaac ROS demonstration

ROBOTIS performs object manipulation tasks using NVIDIA Isaac ROS CuMotion.

Field AI’s robot foundation model can run entirely on the robot without relying on cloud connectivity. By integrating Isaac ROS on Jetson devices, the system takes advantage of GPU acceleration to improve the efficiency of the on-robot AI stack.

Noble Machines uses NVIDIA Isaac ROS on Jetson to accelerate the development of general-purpose robots for industrial applications. The company is building out-of-the-box AI and perception capabilities rather than developing them from scratch.

Build and deploy physical AI with Isaac ROS 5.0

Isaac ROS 5.0 addresses both sides of the physical AI challenge by combining an open robotics ecosystem, accelerated computing and new agent development workflows. Together, these technologies help developers build increasingly capable robotic applications and run them efficiently in the physical world.

NVIDIA Isaac ROS 5.0 is currently available as free and open-source software. Developers can learn more and get started with NVIDIA Isaac ROS through the Isaac ROS documentation and GitHub resources.