Pieter Abbeel | Research UC Berkeley

Research Expertise and Interest

robotics, machine learning, artificial intelligence, Deep Learning

Research Description

Pieter Abbeel is Professor and Director of the Robot Learning Lab at UC Berkeley [2008- ], Co-Founder of covariant.ai [2017- ], Co-Founder of Gradescope [2014- ], Advisor to OpenAI, Founding Faculty Partner AI@TheHouse, Advisor to many AI/Robotics start-ups. He works in machine learning and robotics. In particular his research focuses on making robots learn from people (apprenticeship learning), how to make robots learn through their own trial and error (reinforcement learning), and how to speed up skill acquisition through learning-to-learn (meta-learning). His robots have learned advanced helicopter aerobatics, knot-tying, basic assembly, organizing laundry, locomotion, and vision-based robotic manipulation. He has won numerous awards, including best paper awards at ICML, NIPS and ICRA, early career awards from NSF, Darpa, ONR, AFOSR, Sloan, TR35, IEEE, and the Presidential Early Career Award for Scientists and Engineers (PECASE). Pieter's work is frequently featured in the popular press, including New York Times, BBC, Bloomberg, Wall Street Journal, Wired, Forbes, Tech Review, NPR.

In the News

Stock image of ones and zeros with a lock shape highlighted

UC Berkeley Joins NSF-Backed AI Institute for Cybersecurity

Five UC Berkeley faculty members will join the AI Institute for Agent-based Cyber Threat Intelligence and Operation (ACTION).

Robot dog walks across various types of terrain

Step by Step Berkeley Robots Learn to Walk on Their Own in Record Time

Using advances in machine learning, two separate teams have developed approaches to shorten in-the-field training for quadruped robots in record time.

Berkeley Robot Learning Pioneer Pieter Abbeel Wins ACM Prize in Computing

The Association for Computing Machinery (ACM), the world’s largest scientific and educational computing society, has awarded UC Berkeley professor Pieter Abbeel the 2021 ACM Prize in Computing for his foundational work in robot learning.

Learning to learn

When children play with toys, they learn about the world around them — and today’s robots aren’t all that different. At UC Berkeley’s Robot Learning Lab, groups of robots are working to master the same kinds of tasks that kids do: placing wood blocks in the correct slot of a shape-sorting cube, connecting one plastic Lego brick to another, attaching stray parts to a toy airplane.

Thumbnail of video: Blue, Berkeley's "human-safe" robot

AI-powered Berkeley robot among Popular Science’s ‘Best of What’s New'

An affordable, human-friendly robot developed by Berkeley engineers is among Popular Science’s “Best of What’s New” innovations for 2019.

Meet Blue, the low-cost, human-friendly robot designed for AI

Enter Blue, a new low-cost, human-friendly robot conceived and built by a team of researchers at the University of California, Berkeley. Blue was designed to use recent advances in artificial intelligence (AI) and deep reinforcement learning to master intricate human tasks, all while remaining affordable and safe enough that every artificial intelligence researcher — and eventually every home — could have one.

Berkeley startup to train robots like puppets

Robots today must be programmed by writing computer code, but imagine donning a VR headset and virtually guiding a robot through a task and then letting the robot take it from there.

Pieter Abbeel and BRETT, the Berkeley Robot for Elimination of Tedious Tasks.

“Deep Learning”: A Giant Step for Robots

Bakar Fellow Pieter Abbeel studies deep learning in robots. The robot BRETT (Berkeley Robot for Elimination of Tedious Tasks) has mastered a range of skills, including folding laundry, knot-tying, and basic assembly.

Pieter Abbeel, Hillel Adesnik and Sayeef Salahuddin

Three young faculty members honored by White House

Three UC Berkeley faculty members named as recipients of the Presidential Early Career Awards for Scientists and Engineers, the highest honor bestowed by the U.S. government on science and engineering professionals in the early stages of their independent research careers.

BRETT is shown here learning how to screw a cap onto a water bottle.

New ‘deep learning’ technique enables robot mastery of skills via trial and error

UC Berkeley researchers have developed algorithms that enable robots to learn motor tasks through trial and error using a process that more closely approximates the way humans learn, marking a major milestone in the field of artificial intelligence.

Big NSF grant funds research into training robots to work with humans

What if robots and humans, working together, were able to perform tasks in surgery and manufacturing that neither can do alone? That’s the question driving new research by UC Berkeley robotics experts Ken Goldberg and Pieter Abbeel and colleagues from four other universities, who were awarded a $3.5 million grant from the National Science Foundation.

UC Berkeley robotics expert named among world’s top young innovators

Pieter Abbeel, a UC Berkeley, professor known for his novel work in the field of machine learning in robotics – including robots that can fold laundry – has been named to a prestigious list of 35 of the world’s top young innovators by Technology Review magazine.

Laundry duty getting you down? Robots to the rescue!

Folding laundry may seem mundane, but for a robot, identifying a 3-D object and manipulating it correctly, it’s an exercise that requires intelligence that humans may take for granted. Pieter Abbeel and his team of engineers are developing increasingly efficient strategies and algorithms to help robots fold towels, forming the foundation for the next generation.

Four young faculty members to receive $50,000 Sloan Research Fellowships

Four UC Berkeley faculty members have been awarded prestigious Sloan Research Fellowships, given annually by the Alfred P. Sloan Foundation to scientists, mathematicians and economists at an early stage of their careers.

Researchers enable a robot to fold towels

A team from Berkeley's Electrical Engineering and Computer Sciences department has figured out how to get a robot to fold previously unseen towels of different sizes. Their approach solves a key problem in robotics -- how to deal with flexible, or "deformable," objects.

Featured in the Media

Please note: The views and opinions expressed in these articles are those of the authors and do not necessarily reflect the official policy or positions of UC Berkeley.

May 6, 2020

Logistics AI Startup Covariant Reaps $40 Million in Funding Round

Wall Street Journal

Jennifer Smith

Covariant, an AI startup co-founded by electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, and building upon Berkeley research, has raised $40 million to boost hiring and adapt its AI robotics software to new industries. "What we've built is a universal brain for robotic manipulation tasks," says Covariant co-founder and CEO Peter Chen, a Berkeley alum and doctoral student in Professor Abbeel's lab. The COVID-19 pandemic has raised interest in robotics to help companies cope with dramatic shifts in consumer demand and to accommodate new limits on workplace operations, like spacing workers further apart. Stories on this topic have appeared in dozens of sources, including Venture Beat, TechCrunch, Robot Report, and Crunchbase News.

FullStory (*requires registration)

May 5, 2020

UC Berkeley researchers open-source RAD to improve any reinforcement learning algorithm

VentureBeat

Khari Johnson

A group of Berkeley researchers, including electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, has released a state-of-the-art open-sourced Reinforcement Learning with Augmented Data (RAD) module. The reinforcement learning algorithm, used in machine learning for robotics, beats prior modules, including Google AI's PlaNet, DeepMind's Dreamer, and SLAC from Berkeley and DeepMind.

FullStory

March 18, 2020

BADGR mobile robot learns to navigate on its own

The Robot Report

Eugene Demaitre

Berkeley researchers have developed a mobile robot called BADGR, which learns to navigate independently. "Most mobile robots think purely in terms of geometry; they detect where obstacles are, and plan paths around these perceived obstacles in order to reach the goal," says doctoral electrical engineering and computer sciences student Gregory Kahn, one of the study's co-authors. "This purely geometric view of the world is insufficient for many navigation problems." Kahn worked on the robot with electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, and assistant electrical engineering and computer sciences professor Sergey Levine. With just 42 hours of autonomously collected data, BADGR outperformed Simultaneous Localization and Mapping (SLAM) approaches in a test, and it had less data to work with than other navigation methods, according to the researchers. Link to video.

FullStory

February 28, 2020

Industrial robotics giant teams up with a rising A.I. startup

Fortune

Jonathan Vanian

Covariant, an AI startup co-founded by electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, has formed a new partnership with industrial robotics company ABB to build warehouse robots. ABB's forte is using AI technologies like computer vision, and it was looking with a partner that could help them make robots with grasping skills. According to Sami Atiya, ABB's president of robotics and discrete automation, Covariant was chosen because it was the only company whose software could recognize different types of items without human assistance.

FullStory

January 29, 2020

A Warehouse Robot Learns to Sort Out the Tricky Stuff

New York Times

Adam Satariano, Cade Metz

A new warehouse robot built by Covariant.AI is demonstrating that it's capable of sorting small electrical parts quickly and accurately, and, according to these reporters, represents a "major advance in artificial intelligence and the ability of machines to perform human labor." Says one warehouse automation expert: "I've worked in the logistics industry for more than 16 years and I've never seen anything like this." Covariant was co-founded by electrical engineering and computer sciences professor Pieter Abbeel, director of the Berkeley Robot Learning Lab and co-director of the Berkeley Artificial Intelligence Research (BAIR) lab, and it builds upon research he and his colleagues conducted at Berkeley. He says that humans will continue to work alongside robots like his in warehouses, but that the job market would surely change as machine learning improves. "If this happens 50 years from now, there is plenty of time for the educational system to catch up to the job market," he says.

FullStory (*requires registration)

December 5, 2019

The 100 greatest innovations of 2019

Popular Science

BLUE, a relatively inexpensive robotic arm developed in the lab of electrical engineering and computer sciences professor Pieter Abbeel and produced by the spinoff Berkeley Open Robotics, received a nod in Popular Science's Best of What's New list of top innovations for 2019. Currently only available to campus labs, it's expected to hit the market in a few years, "where it might someday have as many uses as smartphones do today," the editors say. The robot's machine-learning algorithms train it to do tasks that could include folding laundry, cleaning a bathroom, or unloading a dishwasher. According to the listing: "What the bulky, low-powered arm lacks in precision, it makes up for it with uncanny, humanlike perception: It uses visual and tactile sensors to judge distances and apply gentle pressure through clamplike grippers." For more on this, see our press release at Berkeley News.

FullStory

April 10, 2019

Blue robot is made for AI -- on the cheap

New Atlas

Ben Coxworth

Blue, a new low-cost robot that learns by trial and error and has been designed to safely interact with humans, has been introduced by a team of roboticists at Berkeley. Now in commercial development by a spinoff startup called Berkeley Open Arms, it is expected to cost an astonishingly low $5,000. Led by electrical engineering and computer sciences professor Pieter Abbeel, co-developers include postdoctoral research fellow Stephen McKinley, and graduate student David Gealy. Eventually, the team expects that Blue robots will help people with all kinds of repetitive but light tasks that require manipulative dexterity, such as folding laundry, washing dishes, and picking up messes. Explaining why the team deliberately tried to develop a robot that is "weak," Gealy says: "Essentially, we can get more out of a weaker robot. And a weaker robot is just safer. The strongest robot is most dangerous. We wanted to design the weakest robot that could still do really useful stuff." Link to video. For more on this, see our press release at Berkeley News. Stories on this topic appeared in dozens of sources, including Science & Enterprise, Millennium Post, Odborne Casopisy (Czech Republic), FARS News Agency (Iran), and Outlook India.

FullStory

April 9, 2019

Meet Blue: The Cheap and Manipulative (In A Good Way) Robot

Wired

A team of Berkeley roboticists is introducing Blue, a new low-cost robot that could someday work alongside people in their homes and workplaces. Prior industrial robots have been unsafe to work next to humans, and have traditionally been caged, but Blue is more sensitive, responding to human contact by stopping its movements so as not to hurt anyone. And since it's made largely by 3D-printed parts, it's significantly cheaper than other robots. "This robot is designed for the assumption that in the future, robots will be controlled much more intelligently by AI systems that use visual feedback, that use force feedback, much like how humans control their own arms," says electrical engineering and computer sciences professor Pieter Abbeel, the project leader. Eventually, the team expects that new editions of Blue will be helping people with all kinds of repetitive tasks that require manipulative dexterity, such as folding laundry, washing dishes, and picking up messes. For more on this, see our press release at Berkeley News. Other stories have appeared in dozens of sources, including MIT Technology Review, IEEE Spectrum, and The Verge.

FullStory