BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Don't Relearn Physics: PDE-Structured Robot Motion Learning
DTSTART:20260529T110000
DTEND:20260529T120000
DTSTAMP:20260526T074434Z
UID:b3e3ea9195a7238eb6d1e297f7e4feef593d75337939a8749b1f26fe
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Ahmed Qureshi \nModern robot motion learning increasin
 gly relies on large datasets and expensive expert demonstrations to implic
 itly acquire knowledge that physics already provides in closed form. This 
 talk argues for a different approach: rather than asking neural networks t
 o rediscover what partial differential equations (PDEs) already encode\, w
 e can directly embed PDE-based priors as the structural foundation for lea
 rning 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 characterizing shortest-path and minimum-t
 ime problems — eliminating the need for expert trajectories\, graph sear
 ch\, or trial-and-error interaction. The resulting self-supervised methods
  train in minutes\, generalize across environments\, and infer motion plan
 s in near real-time speed. We demonstrate strong scalability across high-d
 imensional systems and constraint-rich manipulation tasks. We further show
  that Eikonal priors yield a novel mapping representation that encodes mot
 ion-feasible geometry directly in configuration space — outperforming oc
 cupancy grids and signed distance fields for planning purposes without add
 itional computational overhead. Finally\, we show that incorporating Eikon
 al priors into reinforcement learning significantly improves its data effi
 ciency and scalability. Together\, these results suggest that physics\, ra
 ther than data\, should be the first language of robot motion learning.\n\
 nAhmed Qureshi is an Assistant Professor in the Department of Computer Sci
 ence at Purdue University\, where he directs the Cognitive Robot Autonomy 
 and Learning (CoRAL) Lab. His research pursues a physics-first philosophy 
 for robot motion learning: rather than relying on large expert demonstrati
 ons or trial-and-error interaction\, his group develops methods that embed
  the governing laws of physics directly into learning algorithms. This app
 roach has produced self-supervised methods that train in minutes\, require
  no expert annotation\, and plan in near real-time across high-dimensional
 \, manipulation\, and unknown environments. His broader research spans sca
 lable motion planning\, dexterous manipulation\, active perception\, and m
 ulti-agent task and motion planning. Dr. Qureshi's work has been recognize
 d with spotlight and best paper awards at top academic venues. He serves a
 s an Associate Editor for IEEE Transactions on Robotics and IEEE Robotics 
 and Automation Letters\, and received the Outstanding Associate Editor Awa
 rd from RA-L in 2024. He has served on the program committees of RSS\, ICR
 A\, IROS\, and CoRL. He earned his B.S. in Electrical Engineering from NUS
 T\, M.S. from Osaka University\, and Ph.D. in Intelligent Systems\, Roboti
 cs\, and Control from UC San Diego.
LOCATION:ME A3 31 https://plan.epfl.ch/?room==ME%20A3%2031 https://purdue-
 edu.zoom.us/j/9806898618
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR