From simulation to reality
The new era of robotics with NVIDIA IsaacRobotics has developed faster in the last two years than in the previous two decades. Today, we are no longer just talking about programmed machines, but about physical AI – artificial intelligence that understands the physical world, interacts with it and learns from it. At the heart of this revolution? The NVIDIA Isaac™ platform.
In the past, robots had to be laboriously programmed for each individual task. Today, NVIDIA Isaac uses foundation models to give robots a basic understanding of the world.
- Multimodal intelligence: Robots can now see, hear and understand complex voice commands in the context of their environment.
- Adaptive capabilities: Thanks to the platform, systems learn to adapt spontaneously to changing environments – whether in an unstructured warehouse or on a dynamic production line.
The milestones: GR00T and Cosmos
The milestones: GR00T and Cosmos
Two names dominate the current discussion among experts: Project GR00T and NVIDIA Cosmos.
- Project GR00T (Generalist Robot 00 Technology): This foundation model enables humanoid robots in particular to learn movements by observing humans (imitation learning). The current version, N1.6, allows for unprecedented dexterity and coordination.
- NVIDIA Cosmos: As a ‘world foundation model,’ Cosmos is used to make physically accurate predictions in simulations. It generates synthetic data that robots use to run through millions of scenarios in the virtual world before taking their first steps in reality.
To develop and operate these complex models, NVIDIA relies on a seamless hardware chain:
| Platform | Function | Objective |
| NVIDIA DGX™ | Training | Creation of massive AI models (e.g. Blackwell architecture). |
| NVIDIA OVX™ | Simulation | Operation of Isaac Sim and Omniverse for digital twins. |
| NVIDIA Thor / AGX | Deployment | The ‘edge power’ in the robot itself for real-time decisions. |
The software suite has been massively expanded to shorten development cycles:
- Isaac Lab: A specialised environment for reinforcement learning, where robots learn through high-speed trial and error.
- Isaac ROS 4.x: Offers CUDA-optimised packages for perception and mapping, which now natively support the latest transformer architectures.
"The question is no longer whether a robot can learn a complex task, but how quickly we can prepare it for this in simulation."
Overview
The NVIDIA DGX platform combines the best of NVIDIA software and infrastructure. It’s ideal for training multi-modal foundational models for robots.
NVIDIA DGX
The NVIDIA DGX platform combines the best of NVIDIA software and infrastructure. It’s ideal for training multi-modal foundational models for robots.
NVIDIA OVX
NVIDIA OVX™ systems provide industry-leading graphics and compute performance to accelerate the next generation of robotics.
NVIDIA AGX
NVIDIA AGX Systems, including NVIDIA Jetson™, offer high performance and energy efficiency, making them the leading platform for robotics. Trained, tested, and optimized robot AI models are deployed to these systems for real-world operation.
NVIDIA OSMO
NVIDIA OSMO is a cloud-native workflow orchestration platform that lets you easily scale your workloads across distributed environments—from on-premises to private and public cloud resource clusters.
The NVIDIA Isaac platform is much more than a collection of tools – it is the operating system of the autonomous future. Whether you are optimising autonomous mobile robots (AMRs) for logistics or working on the next generation of humanoid assistants, the combination of powerful hardware and cloud-native workflows such as NVIDIA OSMO makes all the difference.
At sysGen, we support you in integrating these technologies into your infrastructure. From selecting the right DGX systems for training to implementing Jetson Thor at the edge, we are your partner for the robotics of tomorrow.