X Development LLC

Google Inc.

Qualcomm Incorporated

Paul Wohlhart

Mrinal Kalakrishnan

Kurt Konolige

Nareshkumar Rajkumar

Stefan Hinterstoisser

Seyed Mohammad Khansari Zadeh

Konstantinos Bousmalis

Sergey Levine

Yunfei Bai

Stephen James

Vincent Olivier Vanhoucke

Eric Jang

Vincent Lepetit

Clemens Arth

Julian Ibarz

Chelsea Finn

Alexander Irpan

Alexander Herzog

Daniel Kappler

Matthew Laurance Kelcey

Allan Zhou

Sergey Vladimir Levine

Techniques are disclosed that enable training a meta-learning model, for use in causing a robot to perform a task, using imitation learning as well as reinforcement learning. Some implementations relate to training the meta-learning model using imitation learning based on one or more human guided demonstrations of the task. Additional or alternative implementations relate to training the meta-learning model using reinforcement learning based on trials of the robot attempting to perform the task. Further implementations relate to using the trained meta-learning model to few shot (or one shot) learn a new task based on a human guided demonstration of the new task.

Efficient adaption of robot control policy for new task using meta-learning based on meta-imitation learning and meta-reinforcement learning

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a generator neural network to adapt input images.

Domain adaptation using simulation to simulation transfer

Methods for annotating objects within image frames are disclosed. Information is obtained that represents a camera pose relative to a scene. The camera pose includes a position and a location of the camera relative to the scene. Data is obtained that represents multiple images, including a first image and a plurality of other images, being captured from different angles by the camera relative to the scene. A 3D pose of the object of interest is identified with respect to the camera pose in at least the first image. A 3D bounding region for the object of interest in the first image is defined, which indicates a volume that includes the object of interest. A location and orientation of the object of interest is determined in the other images based on the defined 3D bounding region of the object of interest and the camera pose in the other images.

Automated data capture

Training and/or use of a machine learning model for placement of an object secured by an end effector of a robot. A trained machine learning model can be used to process: (1) a current image, captured by a vision component of a robot, that captures an end effector securing an object; (2) a candidate end effector action that defines a candidate motion of the end effector; and (3) a target placement input that indicates a target placement location for the object. Based on the processing, a prediction can be generated that indicates likelihood of successful placement of the object in the target placement location with application of the motion defined by the candidate end effector action. At many iterations, the candidate end effector action with the highest probability is selected and control commands provided to cause the end effector to move in conformance with the corresponding end effector action. When at least one release criteria is satisfied, control commands can be provided to cause the end effector to release the object, thereby leading to the object being placed in the target placement location.

Machine learning methods and apparatus for automated robotic placement of secured object in appropriate location

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training an action selection neural network that is used to control a robotic agent interacting with a real-world environment.

Using simulation and domain adaptation for robotic control

A computer-implemented method, apparatus, computer readable medium and mobile device for determining a 6DOF pose from an input image. The process of determining 6DOF pose may include processing an input image to create one or more static representations of the input image, creating a dynamic representation of the input image from an estimated 6DOF pose and a 2.5D reference map, and measuring correlation between the dynamic representation and the one or more static representations of the input image. The estimated 6DOF may be iteratively adjusted according to the measured correlation error until a final adjusted dynamic representation meets an output threshold.

Predictor-corrector based pose detection

Growing community of inventors

Sunnyvale, CA, United States of America

Paul Wohlhart