Check-in opens

Introduction to NVIDIA Cosmos and Isaac GR00T as complementary foundations for physical AI: world models for scalable physical-world generation and reasoning, and foundation policies for translating perception and language into robot actions.

Embodied Intelligence — Matthew Gombolay with Zulfiqar Zaidi. Physics-aware learning, simulation, perception, and real-time control for agile robots, including autonomous wheelchair tennis, with demos and sim-to-real comparisons.

Advanced frameworks and demos for developing, simulating, and deploying physical AI, including Isaac ROS integration, core packages, and edge/Jetson use cases.

PhysCoRe: Physics-Corrected World Models for Deformable-Object Manipulation — Lu Gan with Haocheng Yin. Physics-corrected world models combine differentiable simulation, material inference, and residual dynamics correction for cloth, rope, and soft-object manipulation, with NVIDIA Warp and real-world demos.

Risk-Aware & Distributionally Robust Adaptive Autonomy — Yashwanth Nakka; student presenters TBD. Risk-aware planning, safety-critical control, uncertainty-aware estimation, and adaptive autonomy for rovers and multimodal systems in challenging real-world conditions.

Limited Availability