/PRNewswire/ — For robots to work reliably in the real world, they must be trained in environments that accurately replicate them. That is exactly the central challenge behind Real2Sim – and the focus of XGRIDS’ appearance at GTC 2026 (March 16 to 19 in San Jose).
image
XGRIDS’ spatial intelligence solution now supports NVIDIA Omniverse NuRec for OpenUSD-based rendering and was present at several points during the event: a startup pitch, robotics demonstrations within the NVIDIA ecosystem, and a joint presentation with Amazon Web Services (AWS).
Real2Sim: From physical space to AI training
At the startup pitch, XGRIDS director Sunny Liao asked a central question:
How can training environments be built that faithfully reflect real-world conditions?
XGRIDS’ answer is a pipeline based on real data. The combination of LiDAR and computer vision for multimodal spatial perception with detailed 3D reconstruction makes it possible to transform physical environments into usable world models for simulations.
Compared to manual 3D modeling, this approach offers the following advantages:
- It reduces the cost of building high-precision environments
- It enables continuous updates as real spaces change
- It keeps the simulation closer to later use in the real world
Developers highlighted at GTC that this opens up a more practical way to train and validate robots.
Spatial intelligence for physical AI
In addition to the appearance on the Inception stage, XGRIDS solutions could be seen in several presentation areas of the GTC.
At the NVIDIA Robotics session, XGRIDS demonstrated its solution for embodied AI systems. By using spatial perception and modeling on four-legged robotic platforms, robots can continuously map and understand their environment. They use the full 3D spatial structure for path planning, behavioral decisions and task execution, instead of relying exclusively on local sensors for immediate obstacle avoidance.
This demonstrated how spatial intelligence can be integrated into embodied AI systems and how robots can use full 3D environments for path planning, decision making and task execution.
At the AWS presentation, XGRIDS presented a complete Real2Sim workflow – from capture to world model generation to training in simulation.
outlook
The long-term focus of XGRIDS remains unchanged: building a spatial intelligence infrastructure that transforms real-world environments into world models that AI systems can understand, use for inference, and use for training.
GTC 2026 marked another step in bringing this work into the physical AI ecosystem. As embodied AI systems increasingly make the leap from laboratories to warehouses, cities and construction sites, the need for accurate, scalable digital representations of environments will continue to increase.
XGRIDS is developing the sensing-to-simulation layer that makes this possible.
Media contact:
[emailprotected]
https://www.xgrids.com
Photo –https://mma.prnewswire.com/media/2938856/image.jpg
