Tabitha Edith Lee
Date:
Speaker
Tabitha Edith Lee is a Ph.D. candidate in Robotics at Carnegie Mellon University’s Robotics Institute. She is a member of the Intelligent Autonomous Manipulation Lab and advised by Prof. Oliver Kroemer. Her thesis research investigates causal robot learning for manipulation: the interplay between robot perception and control through the lens of causality to learn and leverage the causal structure of manipulation tasks. Her research in structural sim-to-real transfer has been recognized by an Honorable Mention selection for the NCWIT Collegiate Award. She is also a Siebel Scholar in Computer Science.
Speaker Links: Google Scholar
Abstract
This talk will describe SCALE, our approach for discovering and learning a diverse set of interpretable robot skills from a limited dataset. Rather than learning a single skill that may fail to capture all of the modes in the data, we first identify the different modes via causal reasoning and learn a separate skill for each of them. Our main insight is to associate each mode with a unique set of causally relevant state space variables that are discovered by performing causal interventions in simulation. Our experiments show that our approach yields diverse manipulation skills that are compact, robust to domain shifts, and suitable for sim-to-real transfer.
Associated Papers
- (CoRL 2023) SCALE: Causal Learning and Discovery of Robot Manipulation Skills using Simulation Link