November 6, 2023
Conference on Robot Learning 2023
The Starling Hotel, 188 14th Street, Atlanta, Georgia 30361, USA
Main Workshop Room: Hub 1 Poster Room: Muse 1
As we increasingly rely on ML models to contend with unstructured and unpredictable environments and tasks, it is paramount that we also acknowledge the shortcomings of our models: In practice, robots often fail to meet our expectations when we deploy them in the real world, where distributions subtly shift from training data, and where we will always continue discovering rare corner cases and failure scenarios not represented at train/design time. While reliability concerns in the face of distributional shifts are well-known, a comprehensive roadmap to address these issues at all levels of a learned autonomy stack is absent. How do we unblock ourselves, and build reliable systems for the real world?
This workshop aims to bring together a diverse group of researchers and industry practitioners to chart a roadmap for 1) addressing the disruptive impact of distributional shifts and out-of-distribution (OOD) observations on the performance of robots and 2) examining opportunities to enable generalization to unseen domains. Therefore, this workshop broadly aims to address gaps between academia and practice by igniting discussions on research challenges and their synergies at all timescales crucial to improving reliability and deploying robust systems:
Identifying the right, i.e., task-relevant, definition or viewpoint: There are many ways to define what makes data out-of-distribution. Specific choices are often left implicit and vary between problem formulations and application contexts. So, what are the most task-relevant definitions for common robotics problems? How can we quantify our definitions, and how will they influence our approach?
Real-time decision-making: How can we leverage full-stack sensory information to anticipate model errors and take actions to mitigate the consequences?
Episodic interaction with an environment: Can we develop methods that mitigate or account for shifted conditions that consistently degrade a learning-enabled robot's performance?
A continual data lifecycle as we deploy, evaluate, and retrain learning-enabled robots: How can we efficiently collect data during deployment and develop learning algorithms that maximally improve system robustness and performance?
Drago joined Waymo in 2018 to lead the Research team, which focuses on pushing the state of the art in autonomous driving using machine learning, spanning multiple areas including perception, prediction, planning and simulation. Earlier in his career he spent eight years at Google; first working on 3D vision and pose estimation for StreetView, and later leading a research team which developed computer vision systems for annotating Google Photos. The team also invented popular methods such as the Inception neural network architecture, and the SSD detector, which helped win the Imagenet 2014 Classification and Detection challenges. Prior to joining Waymo, Drago led the 3D Perception team at Zoox.
Chelsea Finn is an Assistant Professor in Computer Science and Electrical Engineering at Stanford University, and the William George and Ida Mary Hoover Faculty Fellow. Her research interests lie in the capability of robots and other agents to develop broadly intelligent behavior through learning and interaction. To this end, her work has pioneered end-to-end deep learning methods for vision-based robotic manipulation, meta-learning algorithms for few-shot learning, and approaches for scaling robot learning to broad datasets. Her research has been recognized by awards such as the Sloan Fellowship, the IEEE RAS Early Academic Career Award, and the ACM doctoral dissertation award, and has been covered by various media outlets including the New York Times, Wired, and Bloomberg. Prior to Stanford, she received her Bachelor's degree in Electrical Engineering and Computer Science at MIT and her PhD in Computer Science at UC Berkeley.
Amy Zhang is an assistant professor at UT Austin in the Chandra Family Department of Electrical and Computer Engineering. Her group, Machine Intelligence through Decision-making and Interaction (MIDI) Lab, focuses on theory and algorithms for sequential decision-making problems, with an emphasis on reinforcement learning, self-supervised learning, and representation learning, with a focus on improving robustness, generalization, and sample efficiency. She did her PhD at McGill University and Mila - Quebec AI Institute, co-supervised by Joelle Pineau and Doina Precup. She also has an M.Eng. in EECS and dual B.Sci. degrees in Mathematics and EECS from MIT. She is recruiting PhD students.
Dr. Judy Hoffman is an Assistant Professor in the School of Interactive Computing at Georgia Tech, a member of the Machine Learning Center, and a Diversity and Inclusion Fellow. Her research lies at the intersection of computer vision and machine learning with specialization in domain adaptation, transfer learning, adversarial robustness, and algorithmic fairness. She has received numerous awards including NSF CAREER, Google Research Scholar Award (2022), Samsung AI Researcher of the Year Award (2021), NVIDIA female leader in computer vision award (2020), AIMiner top 100 most influential scholars in Machine Learning (2020), MIT EECS Rising Star in 2015, and the NSF Graduate Fellowship. In addition to her research, she co-founded and continues to advise for Women in Computer Vision, an organization which provides mentorship and travel support for early-career women in the computer vision community. Prior to joining Georgia Tech, she was a Research Scientist at Facebook AI Research. She received her PhD in Electrical Engineering and Computer Science from UC Berkeley in 2016 after which she completed Postdocs at Stanford University (2017) and UC Berkeley (2018).
I work on machine learning methods for autonomous vehicle planning at Waymo in California. I was fortunate to learn from Carl Rasmussen as a PhD student, and from Sergey Levine as a postdoc, focusing on probabilistic modeling for reinforcement learning. I have also enjoyed learning from Irena Koprinska, Robert Fitch, Thierry Peynot, Jeff Schneider, and Adrien Gaidon. I am passionate about autonomous vehicles, previously working at the Australian Centre for Field Robotics and UberATG (bought by Aurora), co-founded Light Blue Labs (now Wayve), managed a research team at Toyota Research Institute on prediction and planning, and organize workshops on autonomous driving.
Marco Pavone
Stanford University and NVIDIA
Anirudha Majumdar
Princeton University
UC Berkeley
Anushri Dixit
Princeton University
Ed Schmerling
Stanford University
Apoorva Sharma
NVIDIA
Sushant Veer
NVIDIA
Somrita Banerjee
Stanford University
Sampada Deglurkar
UC Berkeley
Allen Ren
Princeton University
Rohan Sinha
Stanford University
David Snyder
Princeton University
For any questions about the workshop, please email ood.workshop@gmail.com or contact Rohan Sinha at rhnsinha@stanford.edu.
We are tremendously grateful to all the reviewers of the workshop submissions, without whom we could not have released timely decisions. They are, in alphabetical order: Robert Dyro, Amine Elhafsi, Matt Foutter, Asher Hancock, Justin Lidard, Zhiting Mei, Nathaniel Simon, Alvin Sun, Sander Tonkens.
Pending, check back soon!