Robotic AI & Learning Lab

Welcome to the RAIL lab website! Our research focus is to enable machines to exhibit flexible and adaptable behavior, acquired autonomously through learning. To that end, we work on learning algorithms, robotics, and computer vision.


June 28, 2018 BAIR blog post on our RSS 2018 paper: One-Shot Imitation from Watching Videos
April 20, 2018 BAIR blog post on our SIGGRAPH 2018 paper: Towards a Virtual Stuntman
April 19, 2017 BAIR blog post published about one of our RSS 2018 papers: Shared Autonomy via Deep Reinforcement Learning
January 29, 2017 Seven papers accepted at the 2018 International Conference on Learning Representations (ICLR), see the publications page. Preprints for all papers coming soon!
January 12, 2017 Ten papers accepted at the 2018 International Conference on Robotics and Automation (ICRA), see the publications page. Preprints for all papers coming soon!
December 5, 2017 We showcased some of our research in a live robot demo at NIPS 2017. See here for more information.
October 20, 2017 Two more CoRL accepted papers posted: The Feeling of Success: Does Touch Sensing Help Predict Grasp Outcomes? and Self-Supervised Visual Planning with Temporal Skip Connections.
October 12, 2017 Four new papers on robotic learning posted: Using Simulation and Domain Adaptation to Improve Efficiency of Deep Robotic Grasping, Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations, Self-supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation, and Deep Object-Centric Representations for Generalizable Robot Learning
September 7, 2017 Two papers accepted at NIPS 2017: EX2: Exploration with Exemplar Models for Deep Reinforcement Learning and Interpolated Policy Gradient: Merging On-Policy and Off-Policy Gradient Estimation for Deep Reinforcement Learning
September 1, 2017 Five papers accepted at CoRL 2017: End-to-End Learning of Semantic Grasping, Learning Robotic Manipulation of Granular Media, One-Shot Visual Imitation Learning via Meta-Learning, and two more that will be posted shortly!
August 10, 2017 GPLAC: Generalizing Vision-Based Robotic Skills Using Weakly Labeled Images (accepted to ICCV 2017) and Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning released!
June 1, 2017 Four papers accepted at ICML 2017: Modular Multitask Reinforcement Learning with Policy Sketches, Reinforcement Learning with Deep Energy-Based Policies, Combining Model-Based and Model-Free Updates for Trajectory-Centric Reinforcement Learning and Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks.

Research Support

National Science Foundation, 2016 - present
Google, 2016 - present
Honda, 2017 - present
Berkeley DeepDrive (BDD), 2016 - present
Open Philanthropy Project, 2017 - present
Office of Naval Research, 2016 - present
Uber, 2017 - present
NVIDIA, 2016 - present
Siemens, 2017 - present
Amazon, 2018 - present
Berkeley AI Research (BAIR), 2016 - present