log in  |  register  |  feedback?  |  help  |  web accessibility
Equivariant Policy Learning for Robotic Manipulation
Dian Wang
IRB 4105 or https://umd.zoom.us/j/94340703410?pwd=rrXaGSXSpabcMTtDNmeCNf2Ih2fQYE.1
Thursday, March 13, 2025, 1:30-2:30 pm
  • You are subscribed to this talk through .
  • You are watching this talk through .
  • You are subscribed to this talk. (unsubscribe, watch)
  • You are watching this talk. (unwatch, subscribe)
  • You are not subscribed to this talk. (watch, subscribe)
Abstract

Despite recent advances in machine learning methods for robotics, existing learning-based approaches often lack sample efficiency, posing a significant challenge due to the considerable time required to collect real-robot data. In this talk, I will present our innovative methods that tackle this challenge by leveraging the inherent symmetries in the physical environment as an inductive bias in robot learning. Specifically, I will outline a comprehensive framework of equivariant policy learning—which exploits known symmetry properties to constrain and guide learning—and its application across various robotic problem settings, including reinforcement learning and behavior cloning. Our methods significantly outperform state-of-the-art baselines while achieving these results with far less data, both in simulation and in the real world. Furthermore, our approach demonstrates robustness in the presence of symmetry distortions, such as variations in camera angles, highlighting its potential for more reliable real-world deployment.

Bio
Dian Wang is a Ph.D. candidate at Northeastern University, where he is co-advised by Prof. Robert Platt and Prof. Robin Walters. Dian’s research lies at the intersection of Machine Learning and Robotics, with a focus on leveraging equivariant methods to boost robot learning efficiency and generalization. Dian has contributed to leading conferences and journals including CoRL, ICLR, RSS, IJRR, and regularly serves as reviewer in those venues. In recognition of his work, Dian was awarded the outstanding paper award finalist at CoRL 2024 and the J.P.Morgan Chase Ph.D. Fellowship in 2023.
This talk is organized by Samuel Malede Zewdu