Date: August 29, 2025
Speaker: Ze Yang, University of Toronto and Waabi
Title: Toward Scalable World Modeling and Simulation for Autonomy
Abstract: The future of physical AI—from autonomous vehicles to humanoid and warehouse robots—relies on the ability to simulate the real world with realism, controllability, and scalability. In this talk, I will present our work on creating controllable digital twins of real-world environments that capture both spatial structures and dynamic behaviors and enable fine-grained manipulation of environments and physically grounded simulation of sensor data (e.g., camera and LiDAR) across diverse conditions. I will also discuss how we leverage generative models and large reconstruction models to ensure scalability and generalization. Together, these advances enable realistic and scalable world simulation systems that facilitate the development of autonomous systems, such as self-driving vehicles and embodied agents, in a safe, controlled, proactive, and cost-effective manner.