news    views    podcast    learn    |    about    contribute     republish     events



The Berkeley Artificial Intelligence Research (BAIR) Lab brings together UC Berkeley researchers across the areas of computer vision, machine learning, natural language processing, planning, and robotics. BAIR includes over two dozen faculty and more than a hundred graduate students pursuing research on fundamental advances in the above areas as well as cross-cutting themes including multi-modal deep learning, human-compatible AI, and connecting AI with other scientific disciplines and the humanities. The BAIR Blog provides an accessible, general-audience medium for BAIR researchers to communicate research findings, perspectives on the field, and various updates. Posts are written by students, post-docs, and faculty in BAIR, and are intended to provide relevant and timely discussion of research findings and results, both to experts and the general audience.

by   -   December 31, 2017

By Ivan Evtimov, Kevin Eykholt, Earlence Fernandes, and Bo Li based on recent research by Ivan Evtimov, Kevin Eykholt, Earlence Fernandes, Tadayoshi Kohno, Bo Li, Atul Prakash, Amir Rahmati, Dawn Song, and Florian Tramèr.

Deep neural networks (DNNs) have enabled great progress in a variety of application areas, including image processing, text analysis, and speech recognition. DNNs are also being incorporated as an important component in many cyber-physical systems. For instance, the vision system of a self-driving car can take advantage of DNNs to better recognize pedestrians, vehicles, and road signs. However, recent research has shown that DNNs are vulnerable to adversarial examples: Adding carefully crafted adversarial perturbations to the inputs can mislead the target DNN into mislabeling them during run time. Such adversarial examples raise security and safety concerns when applying DNNs in the real world. For example, adversarially perturbed inputs could mislead the perceptual systems of an autonomous vehicle into misclassifying road signs, with potentially catastrophic consequences.

by   -   December 24, 2017

By Carlos Florensa

Reinforcement Learning (RL) is a powerful technique capable of solving complex tasks such as locomotion, Atari games, racing games, and robotic manipulation tasks, all through training an agent to optimize behaviors over a reward function. There are many tasks, however, for which it is hard to design a reward function that is both easy to train and that yields the desired behavior once optimized.

by   -   December 18, 2017

By Changliu Liu, Masayoshi Tomizuka

Democratization of Robots in Factories

In modern factories, human workers and robots are two major workforces. For safety concerns, the two are normally separated with robots confined in metal cages, which limits the productivity as well as the flexibility of production lines. In recent years, attention has been directed to remove the cages so that human workers and robots may collaborate to create a human-robot co-existing factory.

by   -   December 7, 2017

By Sylvia Herbert, David Fridovich-Keil, and Claire Tomlin

The Problem: Fast and Safe Motion Planning

Real time autonomous motion planning and navigation is hard, especially when we care about safety. This becomes even more difficult when we have systems with complicated dynamics, external disturbances (like wind), and a priori unknown environments. Our goal in this work is to “robustify” existing real-time motion planners to guarantee safety during navigation of dynamic systems.

by   -   December 1, 2017

By Anusha Nagabandi and Gregory Kahn

Enabling robots to act autonomously in the real-world is difficult. Really, really difficult. Even with expensive robots and teams of world-class researchers, robots still have difficulty autonomously navigating and interacting in complex, unstructured environments.

by   -   November 22, 2017
Toyota HSR Trained with DART to Make a Bed.

By Michael Laskey, Jonathan Lee, and Ken Goldberg

In Imitation Learning (IL), also known as Learning from Demonstration (LfD), a robot learns a control policy from analyzing demonstrations of the policy performed by an algorithmic or human supervisor. For example, to teach a robot make a bed, a human would tele-operate a robot to perform the task to provide examples. The robot then learns a control policy, mapping from images/states to actions which we hope will generalize to states that were not encountered during training.