Speaker: Huihan Liu the University of Texas at Austin
Time: 2023-07-30 11:00-2023-07-30 12:00
Venue: Zoom link: https://us06web.zoom.us/j/7259762976?pwd=ZUovYys4UXM1cGtXdEFrbTdJaGQvZz09
With rapid advances in deep learning, we have witnessed impressive demonstrations of novel robot capabilities in research settings. However, these learning systems exhibit brittle generalization and require excessive training data for practical tasks. To harness the capabilities of state-of-the-art robot learning models while embracing their imperfections, we present Sirius, a principled framework for humans and robots to collaborate through a division of work. Our key idea is to improve robot autonomy over long-term deployments with humans in the loop. We introduce a new learning algorithm to enhance the policy's performance based on the data collected from task executions and demonstrate that Sirius consistently outperforms baselines across a range of contact-rich manipulation tasks.
Huihan Liu is a second-year Ph.D. student at the University of Texas at Austin, advised by Yuke Zhu. Her research interests lie in robot learning and human-in-the-loop learning, particularly developing algorithms that help robots learn efficiently from human feedback and building scalable, trustworthy robotic systems for deployment. Her work received the Outstanding Paper Award in Learning at ICRA 2022 and was a finalist for the Best Paper Award at RSS 2023. She earned her Bachelor of Arts degree in Computer Science from UC Berkeley.