Home
Center for Autonomy
Enabling high-impact research in autonomous system design
About the Center for Autonomy
Located at the University of Illinois at Urbana-Champaign, the Center for Autonomy will enable high-impact research and develop new educational programs for students and professionals. The Center will play an important role in designing innovative systems that can function autonomously, or without human intervention, in a safe and reliable way.
From self-driving cars to intelligent robotic assistants to remote surgical systems, autonomous technology will revolutionize the way we live, work, and play. In order to enable this revolution, however, advancements in foundational research and workforce development must first take place to provide assured and certified-safe performance.
Application areas include:
UPCOMING SEMINAR
Don’t Relearn Physics: Learning Robot Motion with PDE Structure
Friday,April 3, 2026
2:00pm Central Time
Speaker: Ahmed Qureshi (Purdue)
In Person Event, CSL Studio 1232
Modern robot motion learning increasingly relies on large datasets and expensive expert demonstrations to implicitly acquire knowledge that physics already provides in closed form. This talk presents an alternative perspective: rather than asking neural networks to rediscover what partial differential equations (PDEs) already encode, we embed PDE-based priors directly as the structural foundation for learning robot motion. Concretely, we formulate motion planning as learning a continuous value function governed by the Eikonal PDE, a special case of Hamilton–Jacobi equations that characterize shortest path and minimum time problems. This formulation eliminates the need for expert trajectories, graph search, or trial and error interaction. The resulting self-supervised methods train in minutes, generalize across environments, and infer motion plans in near real time. The talk will demonstrate scalability across high dimensional robotic systems and constraint rich manipulation tasks. We further show that Eikonal priors enable a new mapping representation that encodes motion feasible geometry directly in configuration space, outperforming occupancy grids and signed distance fields for planning without additional computational overhead. Finally, we show how incorporating Eikonal priors into reinforcement learning significantly improves its data efficiency and scalability. Together, these results suggest that physics, rather than data, should be the first language of robot motion learning.