Robot Learning 2025: Foundational Models for Robotics and Scaling DeepRL - Robot Learning 2025: Unlocking Scalable Robot Learning in the Real World with Karl Pertsch

Abstract: Many domains of machine learning, from language modeling to computer vision, have recently undergone a shift towards generalist models, whose broad generalization abilities are fueled by large and diverse real-world training datasets and high-capacity model architectures. In robotics, however, it has been challenging to apply the same recipe: after all, we cannot easily scrape millions of hours of robot data from the internet and existing model architectures for scalable learning in vision or language modeling are not designed for the continuous control tasks we need to solve in robotics. In this lecture, I will cover some of the work my colleagues and I have been doing over the last few years to unlock scalable robot learning in the real world. Concretely, I will focus on answering the following questions: 1. How can we create large real-robot datasets? (Open X-Embodiment, DROID) 2. How can we design scalable robot policy architectures? (Vision-language-action models: RT-2, OpenVLA, pi0, pi0-FAST) 3. How can we enable these models to generalize more broadly in the future? (Embodied chain-of-thought reasoning) The result of this work is a class of “robot foundation models” that can perform various manipulation tasks out-of-the-box in unseen environments, simply by prompting them in natural language (including at the University of Montreal!), and that can be quickly adapted to new tasks. Bio: Karl Pertsch is a postdoc at UC Berkeley and Stanford, jointly advised by Sergey Levine and Chelsea Finn. He also is a member of the technical staff at Physical Intelligence. His work focuses on building generalist robot policies that can solve a wide range of physical manipulation tasks in the real world. Karl obtained his PhD from USC, advised by Joseph Lim. During his PhD he interned at MetaAI and Google Brain. His work has been awarded the Best Conference Paper Award at ICRA'24 and two Outstanding Paper Award Finalists at CoRL'24.