Google Inc.

Oak Technology, Inc.

Lsi Corporation

Eric Jang

Sergey Levine

Sudheendra Vijayanarasimhan

Peter Pastor Sampedro

Yunfei Bai

Seyed Mohammad Khansari Zadeh

Chelsea Finn

Daniel Kappler

Matthew Bennice

Daniel Alex Lam

Brent Austin

Alexander Irpan

Nicolas Sievers

Mrinal Kalakrishnan

Paul Wohlhart

Julian Ibarz

Chen-Chi Chou

Alexander Herzog

Arup K Bhattacharya

Daniel Ho

Yu Qing Du

Mohi Khansari

Allan Zhou

Coline Manon Devin

Implementations disclosed herein relate to mitigating the reality gap through feature-level domain adaptation in training of a vision-based robotic action machine learning (ML) model. Implementations mitigate the reality gap through utilization of embedding consistency losses and/or action consistency losses during training of the action ML model.

Mitigating reality gap through feature-level domain adaptation in training of vision-based robot action model

Implementations described herein relate to training and refining robotic control policies using imitation learning techniques. A robotic control policy can be initially trained based on human demonstrations of various robotic tasks. Further, the robotic control policy can be refined based on human interventions while a robot is performing a robotic task. In some implementations, the robotic control policy may determine whether the robot will fail in performance of the robotic task, and prompt a human to intervene in performance of the robotic task. In additional or alternative implementations, a representation of the sequence of actions can be visually rendered for presentation to the human can proactively intervene in performance of the robotic task.

System(s) and method(s) of using imitation learning in training and refining robotic control policies

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

Deep machine learning methods and apparatus related to semantic robotic grasping are provided. Some implementations relate to training a training a grasp neural network, a semantic neural network, and a joint neural network of a semantic grasping model. In some of those implementations, the joint network is a deep neural network and can be trained based on both: grasp losses generated based on grasp predictions generated over a grasp neural network, and semantic losses generated based on semantic predictions generated over the semantic neural network. Some implementations are directed to utilization of the trained semantic grasping model to servo, or control, a grasping end effector of a robot to achieve a successful grasp of an object having desired semantic feature(s).

Machine learning methods and apparatus for semantic robotic grasping

Deep machine learning methods and apparatus related to manipulation of an object by an end effector of a robot. Some implementations relate to training a semantic grasping model to predict a measure that indicates whether motion data for an end effector of a robot will result in a successful grasp of an object; and to predict an additional measure that indicates whether the object has desired semantic feature(s). Some implementations are directed to utilization of the trained semantic grasping model to servo a grasping end effector of a robot to achieve a successful grasp of an object having desired semantic feature(s).

Deep machine learning methods and apparatus for robotic grasping

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training an object representation neural network. One of the methods includes obtaining training sets of images, each training set comprising: (i) a before image of a before scene of the environment, (ii) an after image of an after scene of the environment after the robot has removed a particular object, and (iii) an object image of the particular object, and training the object representation neural network on the batch of training data, comprising determining an update to the object representation parameters that encourages the vector embedding of the particular object in each training set to be closer to a difference between (i) the vector embedding of the after scene in the training set and (ii) the vector embedding of the before scene in the training set.

Self-supervised robotic object interaction

An apparatus comprising a first circuit and a second circuit. The first circuit may be configured to determine frequency of occurrence information for a range of gray levels from luminance data of an input signal. The second circuit may be configured to selectively adjust enhancement for at least one portion of the range of grey levels based upon the frequency of occurrence information.

Adaptive video enhancement gain control

A method of processing a DVD bitstream includes the steps of reading the DVD bitstream, the bitstream including a sync frame. A sync window is created, the sync window being open at least during the expected timing of a sync detection signal within the sync frame. The sync pattern is detected within the sync frame and the sync detection signal is generated only when the sync pattern has been detected and the sync window is open. A DVD sync pattern detection circuit includes a sync window generator to generate a sync window signal, a sync pattern detector, the sync pattern detector generating a sync detection signal only when both a sync pattern is detected in a DVD input stream and the sync window signal is asserted. A read channel bit counter generates a read counter signal to control the sync window generator, the read channel bit counter being reset when the sync pattern detector detects the sync pattern. The read channel bit counter controls the width of the sync window signals to allow the generation of correctly timed demodulation enable signals even when the sync frames have non-standard bit lengths. The sync windows also prevent the erroneous detection of sync patterns within the data portions of the sync frames. A pseudo sync detection signal is generated when the sync pattern is corrupted or missing from the sync frame.

Enabling accurate demodulation of a DVD bit stream using devices including a SYNC window generator controlled by a read channel bit counter

An apparatus and a method to access data from a digital recording medium configured to store data in sectors delimited by a predetermined sequence of sync patterns. The apparatus includes a sync pattern detector and a sync sequence detector to detect whether a sequence of sync patterns detected by the sync pattern detector matches the predetermined sync pattern sequence. The sync sequence detector generates a sync error signal whenever a detected sync pattern fails to match an expected sync pattern taken from the predetermined sync pattern sequence. A sync error signal counter counts the number of sync error signals generated by the sync sequence detector and a comparator compares the number of sync error signals counted by the sync error signal counter with a sync error threshold, the comparator generating a disk error signal when the threshold is exceeded. As the method and the implementing device need not rely upon the successful detection of any one sync pattern in the sector including the first sync pattern of the sector, data accesses may be carried out even though the first or a predetermined number of subsequent sync patterns in the sector may be unreadable or corrupted.

Methods and systems for accessing data from a DVD using a sync detector including a 26 state finite state machine

Growing community of inventors

Cupertino, CA, United States of America

Eric Jang