Laura Smith

Hello! I'm a third-year PhD student in CS at UC Berkeley, where I am advised by Sergey Levine. I work on enabling robots to interact with and learn in the real world, so as to acquire human-like abilities. My research is graciously supported in part by the NSF Graduate Research Fellowship.

Email  /  CV  /  Google Scholar  /  Twitter

A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learning
Laura Smith*, Ilya Kostrikov*, Sergey Levine
[paper] [website] [code]

Can a robot learn locomotion in the wild? We found that with the right implementation, it can be a walk in the park! Recent advancements in ML algorithms and libraries can enable a real quadruped to learn to walk in only 20 minutes on several indoor and outdoor terrains.

Legged Robots that Keep on Learning: Fine-Tuning Locomotion Policies in the Real World
Laura Smith, J. Chase Kew, Xue Bin Peng, Sehoon Ha, Jie Tan, Sergey Levine
ICRA 2022.

Legged robots are physically capable of traversing a wide range of challenging environments, but designing controllers that can handle this diversity is difficult. What if instead of training controllers that are robust enough to handle any eventuality, we enable the robot to continually learn in any setting it finds itself in?

Offline Meta-Reinforcement Learning with Online Self-Supervision
Vitchyr H. Pong, Ashvin Nair, Laura Smith, Catherine Huang, Sergey Levine
ICML 2022.

Meta-RL can meta-train policies that adapt to new tasks with orders of magnitude less data than standard RL, but meta-training itself is costly and time-consuming. We propose a hybrid offline meta-RL algorithm.

PEBBLE: Feedback-Efficient Interactive Reinforcement Learning via Relabeling Experience and Unsupervised Pre-Training
Kimin Lee*, Laura Smith*, Pieter Abbeel
ICML 2021 (Long Oral).
[paper] [website] [code]

Conveying complex objectives to RL agents can often be difficult Human-in-the-loop RL methods allow practitioners to instead interactively teach agents; however, this has been hard to scale. In this work, we present an off-policy, interactive RL algorithm that capitalizes on the strengths of both feedback and off-policy learning.

AVID: Learning Multi-Stage Tasks via Pixel-Level Translation of Human Videos
Laura Smith, Nikita Dhawan, Marvin Zhang, Pieter Abbeel, Sergey Levine
RSS 2020.
[paper] [website/videos/talk] [blog]

Humans can learn from watching others, imagining how they would perform the task themselves, and then practicing on their own. Can robots do the same? We adopt a similar strategy of imagination and practice in this project to solve complex, long-horizon tasks, like operating a coffee machine or getting objects from within a closed drawer.

SOLAR: Deep Structured Latent Representations for Model-Based Reinforcement Learning
Marvin Zhang*, Sharad Vikram*, Laura Smith, Pieter Abbeel, Matthew Johnson, Sergey Levine
ICML 2019.
[paper] [website/videos/talk] [code] [blog]

Learning representations that are suitable for iterative model-based policy improvement, even when the underlying dynamical system has complex dynamics and image observations.

B-Pref: Benchmarking Preference-Based Reinforcement Learning
Kimin Lee, Laura Smith, Anca Dragan, Pieter Abbeel
NeurIPS Datasets and Benchmarks Track 2021.

B-Pref is a benchmark specially designed for preference-based RL. We simulate teachers with a wide array of irrationalities, and propose metrics not solely for performance but also for robustness to these potential irrationalities.

RLL Outreach
Email rll_outreach AT lists DOT berkeley DOT edu to arrange a visit

CS 10: The Beauty and Joy of Computing
Special topic lecture given on AI

UC Berkeley: CS Education Day 2018
Getting high school students excited about AI
CS 189: Intro to Machine Learning
Undergraduate Student Instructor, Spring 2020

CS 287: Advanced Robotics
Undergraduate Student Instructor, Fall 2019
Lecture given on Imitation Learning

CS 188: Intro to Artificial Intelligence
Undergraduate Student Instructor, Spring 2019
Undergraduate Student Instructor, Fall 2018
Student Resources

website template