NVIDIA Releases Isaac ROS 5.0… Accelerating AI Agent-Based Robotics Development

NVIDIA unveiled Isaac ROS 5.0 at ROSCon Toronto, introducing an agentic development workflow where humans and AI agents collaborate to build robots. This update provides NVIDIA's accelerated computing, Physical AI models, and production-ready libraries to approximately 1.3 million ROS users.

Agent-Friendly Development Environment

The newly introduced NVIDIA Isaac Skills accelerate the process of AI agents understanding Isaac ROS tools and documentation, and translating developer intent into executable applications. Specifically, the FoundationStereo fine-tuning skill allows AI agents to adapt stereo perception models to specific sensor configurations, while the FoundationPose inference library improves object position and orientation detection speed by up to 5.5x.

Platform Support and Ecosystem Expansion

Isaac ROS 5.0 supports ROS Lyrical and Ubuntu 24.04. NVIDIA has contributed standard interfaces to the Open Source Robotics Alliance to facilitate efficient data processing across various hardware. Additionally, the AgenticROS project establishes an environment where AI agents can interact with ROS-based robots.

Partner Integrations and Use Cases

Multiple partners have integrated Isaac ROS 5.0 into their solutions.

• Intrinsic: Released a reference application using FoundationPose that enables dynamic object detection and handling without physical jigs for CNC machine tending.
• Seeed Studio: Offers a Physical AI platform combining the reBot Arm and Jetson Thor for collision-aware manipulation and autonomous pick-and-place.
• Universal Robots: Embedded Isaac ROS into its AI Accelerator SDK to increase manufacturing cell flexibility, allowing adaptive handling of parts that are not precisely positioned.
• Magna: Reduces deployment risks for manufacturing and mobility automation by combining Isaac GR00T model deployment with Isaac Sim hardware-in-the-loop (HIL) testing.

Isaac ROS 5.0 supports devices from the entry-level Jetson Orin Nano to the high-performance Jetson Thor, and is distributed as open source on GitHub.