Event
Microsoft Future Leaders in Robotics and AI Seminar Series: Anushri Dixit
Friday, March 15, 2024
2:00 p.m.
Online Seminar
Robot Safety and Generalization in the Era of Foundation Models
Anushri Dixit
Postdoctoral Fellow
Princeton University
Abstract
Significant strides in AI over the past few years have enabled robotic systems to interpret and interact with the world in increasingly versatile ways. The large, often multi-modal, datasets that are used to train modern learning-based systems endow robots with capabilities like scene understanding and commonsense reasoning. However, the safe integration and reliability of these learned models for robotics applications still remains in question. Learned perception systems fail to identify objects correctly and LLM-based planners hallucinate their outputs leading to unsafe robot behavior downstream. In this talk, I will discuss a technique called conformal prediction and its usefulness for quantifying the uncertainty of such learned models. First, I will discuss a framework for rigorously quantifying the uncertainty of a pre-trained obstacle detection system in a way that provides a formal assurance on correctness and safety for planning applications. Next, I will present a LLM-based planning framework wherein given a human instruction, the robot is statistically guaranteed to complete the task while asking for human help if it is uncertain. I will provide the experimental validation of these methods on various robotic platforms for navigation and mobile manipulation tasks.
Bio
Anushri Dixit is a Postdoctoral Researcher in the Department of Mechanical & Aerospace Engineering at Princeton University. She earned her Ph.D. in Control and Dynamical Systems from California Institute of Technology in 2023 and her B.S. in Electrical Engineering from Georgia Institute of Technology in 2017. Her research focuses on motion planning and control of robots in unstructured environments while accounting for uncertainty in a principled manner. Her work on risk-aware methodologies for planning has been deployed on various robotic platforms as a part of Team CoSTAR’s effort in the DARPA Subterranean Challenge. She has received the Outstanding Student Paper Award at the Conference on Decision and Control, Best Student Paper Award at the Conference of Robot Learning, and was selected as a Rising Star in Data Science by The University of Chicago. She will start as an Assistant Professor at the University of California, Los Angeles in Mechanical and Aerospace Engineering in July 2024.
About the Seminar Series
The Future Leaders in Robotics and AI: Celebrating Diversity and Innovation Seminar Series is part of the University of Maryland and Microsoft Robotics and Diversity Initiative. This is a nationwide online seminar series for PhD students, postdoctoral researchers, or early-career professionals, especially underrepresented minorities and women. The seminar series highlights the latest research and innovation in the field of robotics and AI. The series is intended to provide exposure and mentorship opportunities to the speakers, build a network of innovators across the country, and support the speakers’ career planning.
The seminars are held once per month during the academic year. There are two speakers per seminar. Each speaker gives a 20-minute research presentation followed by a Q&A segment. Immediately after the second seminar, the speakers participate in a discussion with faculty.
