#USA #San Jose #Nvidia GTC #XGRIDS #Real2Sim

XGRIDS Showcases Real2Sim at NVIDIA GTC 2026

In the heart of Silicon Valley, from March 16 to 19, 2026, NVIDIA's GTC conference brought together innovators and industry leaders focused on the future of technology. XGRIDS, a trailblazer in spatial intelligence, presented its groundbreaking Real2Sim approach to help robots train in environments that mirror the real world. This need for realistic training environments is pivotal for ensuring reliable robot performance across various applications, from warehouses to urban environments.

The Core Challenge: Real2Sim

The Real2Sim methodology addresses an essential challenge in robotics: how to create accurate training environments that reflect the complexities of the real world. During the startup pitch, XGRIDS’ CEO, Sunny Liao, emphasized the importance of environmental fidelity for effective robot learning. Leveraging advancements in laser scanning (LiDAR) and computer vision, XGRIDS creates high-fidelity 3D models of physical spaces, transforming them into usable simulation environments.

This innovative approach not only reduces the costs associated with high-fidelity environment creation but also enables continuous updates aligning with real-world changes. As a result, developers can ensure that the simulated training environments remain relevant and practical, facilitating smoother transitions from simulation to real-world application.

Enhancements in Robotic Systems

Attendees of the GTC conference were particularly intrigued by how XGRIDS is integrating spatial intelligence into embodied AI systems. Demonstrating its technology on quadruped robotic platforms, XGRIDS showcased how robots can navigate and comprehend their surroundings in real-time. By employing a complete 3D spatial mapping structure, these robots enhance their decision-making and task execution capabilities, moving beyond basic obstacle avoidance relying solely on local sensors.

The demonstrations highlighted how XGRIDS' spatial intelligence is crucial for embodying AI, facilitating a comprehensive understanding of environments and improving robots' operational effectiveness.

Moreover, during a joint presentation with Amazon Web Services (AWS), the full Real2Sim workflow was showcased, illustrating the transition from data capture to simulation training and world model generation.

Future Prospects

As XGRIDS looks to the future, its long-term vision remains clear: to build a robust spatial intelligence infrastructure that translates real-world environments into comprehensible models for AI systems. The excitement around the advancements presented at GTC 2026 marks a significant milestone in integrating these technologies into physical AI ecosystems.

As embodied AI systems prepare to move beyond lab settings into everyday environments like cities and construction sites, demand for precise and adaptable digital representations of various settings is anticipated to soar. XGRIDS is poised to lead this effort through its capture-simulation framework, essential for shaping the future of robotics and AI integration.

With the industry approaching a transformative era, technologies like Real2Sim are not just enhancing robotic capabilities but also redefining how machines comprehend and interact with the world.