-
公开(公告)号:US20250153363A1
公开(公告)日:2025-05-15
申请号:US19025627
申请日:2025-01-16
Applicant: GOOGLE LLC
Inventor: Seyed Mohammad Khansari Zadeh , Eric Jang , Daniel Lam , Daniel Kappler , Matthew Bennice , Brent Austin , Yunfei Bai , Sergey Levine , Alexander Irpan , Nicolas Sievers , Chelsea Finn
Abstract: 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.
-
2.
公开(公告)号:US20210101286A1
公开(公告)日:2021-04-08
申请号:US17053335
申请日:2020-02-28
Applicant: Google LLC
Inventor: Honglak Lee , Xinchen Yan , Soeren Pirk , Yunfei Bai , Seyed Mohammad Khansari Zadeh , Yuanzheng Gong , Jasmine Hsu
Abstract: Implementations relate to training a point cloud prediction model that can be utilized to process a single-view two-and-a-half-dimensional (2.5D) observation of an object, to generate a domain-invariant three-dimensional (3D) representation of the object. Implementations additionally or alternatively relate to utilizing the domain-invariant 3D representation to train a robotic manipulation policy model using, as at least part of the input to the robotic manipulation policy model during training, the domain-invariant 3D representations of simulated objects to be manipulated. Implementations additionally or alternatively relate to utilizing the trained robotic manipulation policy model in control of a robot based on output generated by processing generated domain-invariant 3D representations utilizing the robotic manipulation policy model.
-
3.
公开(公告)号:US11872699B2
公开(公告)日:2024-01-16
申请号:US18097153
申请日:2023-01-13
Applicant: GOOGLE LLC
Inventor: Seyed Mohammad Khansari Zadeh
IPC: B25J9/16 , B25J13/08 , G05B19/42 , G05B19/423
CPC classification number: B25J9/163 , B25J13/088 , G05B19/423 , G05B2219/40465 , G05B2219/40471 , G05B2219/40474 , Y10S901/04
Abstract: Generating a robot control policy that regulates both motion control and interaction with an environment and/or includes a learned potential function and/or dissipative field. Some implementations relate to resampling temporally distributed data points to generate spatially distributed data points, and generating the control policy using the spatially distributed data points. Some implementations additionally or alternatively relate to automatically determining a potential gradient for data points, and generating the control policy using the automatically determined potential gradient. Some implementations additionally or alternatively relate to determining and assigning a prior weight to each of the data points of multiple groups, and generating the control policy using the weights. Some implementations additionally or alternatively relate to defining and using non-uniform smoothness parameters at each data point, defining and using d parameters for stiffness and/or damping at each data point, and/or obviating the need to utilize virtual data points in generating the control policy.
-
4.
公开(公告)号:US11554483B2
公开(公告)日:2023-01-17
申请号:US17094111
申请日:2020-11-10
Applicant: Google LLC
Inventor: James Davidson , Xinchen Yan , Yunfei Bai , Honglak Lee , Abhinav Gupta , Seyed Mohammad Khansari Zadeh , Arkanath Pathak , Jasmine Hsu
IPC: B25J9/16
Abstract: Deep machine learning methods and apparatus, some of which are related to determining a grasp outcome prediction for a candidate grasp pose of an end effector of a robot. Some implementations are directed to training and utilization of both a geometry network and a grasp outcome prediction network. The trained geometry network can be utilized to generate, based on two-dimensional or two-and-a-half-dimensional image(s), geometry output(s) that are: geometry-aware, and that represent (e.g., high-dimensionally) three-dimensional features captured by the image(s). In some implementations, the geometry output(s) include at least an encoding that is generated based on a trained encoding neural network trained to generate encodings that represent three-dimensional features (e.g., shape). The trained grasp outcome prediction network can be utilized to generate, based on applying the geometry output(s) and additional data as input(s) to the network, a grasp outcome prediction for a candidate grasp pose.
-
公开(公告)号:US20220331962A1
公开(公告)日:2022-10-20
申请号:US17642325
申请日:2020-09-09
Applicant: GOOGLE LLC
Inventor: Soeren Pirk , Seyed Mohammad Khansari Zadeh , Karol Hausman , Alexander Toshev
IPC: B25J9/16
Abstract: Training and/or using a machine learning model for performing robotic tasks is disclosed herein. In many implementations, an environment-conditioned action sequence prediction model is used to determine a set of actions as well as a corresponding particular order for the actions for the robot to perform to complete the task. In many implementations, each action in the set of actions has a corresponding action network used to control the robot in performing the action.
-
公开(公告)号:US12083678B2
公开(公告)日:2024-09-10
申请号:US17422260
申请日:2020-01-23
Applicant: Google LLC
Inventor: Mrinal Kalakrishnan , Yunfei Bai , Paul Wohlhart , Eric Jang , Chelsea Finn , Seyed Mohammad Khansari Zadeh , Sergey Levine , Allan Zhou , Alexander Herzog , Daniel Kappler
IPC: B25J9/16
CPC classification number: B25J9/163 , G05B2219/40116 , G05B2219/40499
Abstract: 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.
-
公开(公告)号:US20210334599A1
公开(公告)日:2021-10-28
申请号:US17279924
申请日:2019-09-27
Applicant: Google LLC
Inventor: Soeren Pirk , Yunfei Bai , Pierre Sermanet , Seyed Mohammad Khansari Zadeh , Harrison Lynch
Abstract: Training a machine learning model (e.g., a neural network model such as a convolutional neural network (CNN) model) so that, when trained, the model can be utilized in processing vision data (e.g., from a vision component of a robot), that captures an object, to generate a rich object-centric embedding for the vision data. The generated embedding can enable differentiation of even subtle variations of attributes of the object captured by the vision data.
-
8.
公开(公告)号:US20210053217A1
公开(公告)日:2021-02-25
申请号:US17094111
申请日:2020-11-10
Applicant: Google LLC
Inventor: James Davidson , Xinchen Yan , Yunfei Bai , Honglak Lee , Abhinav Gupta , Seyed Mohammad Khansari Zadeh , Arkanath Pathak , Jasmine Hsu
IPC: B25J9/16
Abstract: Deep machine learning methods and apparatus, some of which are related to determining a grasp outcome prediction for a candidate grasp pose of an end effector of a robot. Some implementations are directed to training and utilization of both a geometry network and a grasp outcome prediction network. The trained geometry network can be utilized to generate, based on two-dimensional or two-and-a-half-dimensional image(s), geometry output(s) that are: geometry-aware, and that represent (e.g., high-dimensionally) three-dimensional features captured by the image(s). In some implementations, the geometry output(s) include at least an encoding that is generated based on a trained encoding neural network trained to generate encodings that represent three-dimensional features (e.g., shape). The trained grasp outcome prediction network can be utilized to generate, based on applying the geometry output(s) and additional data as input(s) to the network, a grasp outcome prediction for a candidate grasp pose.
-
9.
公开(公告)号:US20200094405A1
公开(公告)日:2020-03-26
申请号:US16617169
申请日:2018-06-18
Applicant: Google LLC
Inventor: James Davidson , Xinchen Yan , Yunfei Bai , Honglak Lee , Abhinav Gupta , Seyed Mohammad Khansari Zadeh , Arkanath Pathak , Jasmine Hsu
IPC: B25J9/16
Abstract: Deep machine learning methods and apparatus, some of which are related to determining a grasp outcome prediction for a candidate grasp pose of an end effector of a robot. Some implementations are directed to training and utilization of both a geometry network and a grasp outcome prediction network. The trained geometry network can be utilized to generate, based on two-dimensional or two-and-a-half-dimensional image(s), geometry output(s) that are: geometry-aware, and that represent (e.g., high-dimensionally) three-dimensional features captured by the image(s). In some implementations, the geometry output(s) include at least an encoding that is generated based on a trained encoding neural network trained to generate encodings that represent three-dimensional features (e.g., shape). The trained grasp outcome prediction network can be utilized to generate, based on applying the geometry output(s) and additional data as input(s) to the network, a grasp outcome prediction for a candidate grasp pose.
-
10.
公开(公告)号:US12226920B2
公开(公告)日:2025-02-18
申请号:US18233251
申请日:2023-08-11
Applicant: GOOGLE LLC
Inventor: Seyed Mohammad Khansari Zadeh , Eric Jang , Daniel Lam , Daniel Kappler , Matthew Bennice , Brent Austin , Yunfei Bai , Sergey Levine , Alexander Irpan , Nicolas Sievers , Chelsea Finn
Abstract: 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.
-
-
-
-
-
-
-
-
-