Predicting object models

Patent metadata

Field	Value
Publication number	US-2023398686-A1
Application number	US-202318114146-A
Country	US
Kind code	A1
Filing date	Feb 24, 2023
Priority date	Jun 14, 2022
Publication date	Dec 14, 2023
Grant date	—

Abstract

Apparatuses, systems, and techniques to update a machine learning model associated with an object. In at least one embodiment, the machine learning model is updated based at least in part on, for example, one or more distributions associated with the machine learning model.

First claim

What is claimed is:

1 . A system, comprising:
at least one processor; and at least one memory comprising instructions that, in response to execution by the at least one processor, cause the system to at least:
obtain a set of observed trajectories associated with an object in a scene;
cause a machine learning model to calculate a first set of values associated with the object based, at least in part, on one or more distributions encoded by the machine learning model;
cause the machine learning model to generate a set of trajectories based, at least in part, on the first set of values;
calculate a second set of values based, at least in part, on the set of observed trajectories and the set of trajectories; and
update the machine learning model based, at least in part, on the second set of values.

2 . The system of claim 1 , wherein the set of observed trajectories indicates one or more trajectories of one or more components of the object in response to a set of actions by a robot.

3 . The system of claim 1 , wherein the one or more distributions indicate one or more probabilities of one or more properties associated with the object.

4 . The system of claim 1 , the at least one memory comprising further instructions that, in response to execution by the at least one processor, cause the system to at least:
use the machine learning model to simulate one or more actions by a robot on the object based, at least in part, on the one or more distributions.

5 . The system of claim 1 , the at least one memory comprising further instructions that, in response to execution by the at least one processor, cause the system to at least:
calculate the second set of values by at least using a likelihood function to process the set of observed trajectories and the set of trajectories.

6 . The system of claim 1 , the at least one memory comprising further instructions that, in response to execution by the at least one processor, cause the system to at least:
calculate a set of actions based, at least in part, on the machine learning model; and cause a robot to perform the set of actions in connection with the object in the scene.

7 . A method, comprising:
obtaining one or more observed trajectories associated with an object; causing a machine learning model to calculate a set of values based on one or more distributions indicating probabilities of one or more properties associated with the object; causing the machine learning model to generate one or more trajectories based, at least in part, on the set of values; and updating the machine learning model based, at least in part, on the one or more observed trajectories and the one or more trajectories.

8 . The method of claim 7 , further comprising:
causing a robot to perform one or more actions in connection with the object; and obtaining the one or more observed trajectories based, at least in part, on the one or more actions.

9 . The method of claim 7 , further comprising:
causing the machine learning model to simulate one or more states of the object based, at least in part, on the set of values to generate the one or more trajectories.

10 . The method of claim 7 , wherein the machine learning model is associated with one or more probabilistic programming languages (PPLs).

11 . The method of claim 7 , wherein the one or more properties include one or more structures of the object.

12 . The method of claim 7 , further comprising:
using the machine learning model to calculate a set of responses of one or more components of the object to a set of actions of a task associated with a robot; and as a result of determining that the set of responses are in accordance with the task, causing the set of actions to be performed by the robot.

13 . A non-transitory computer-readable medium comprising instructions that, when performed by at least one processor of a computing device, cause the computing device to at least:
obtain a set of observed trajectories associated with an object and a robot; cause a machine learning model to generate a set of trajectories based, at least in part, on one or more distributions indicating probabilities of one or more properties associated with the object; process the set of observed trajectories and the set of trajectories to calculate one or more values; and update one or more parameters of the machine learning model based, at least in part, on the one or more values.

14 . The non-transitory computer-readable medium of claim 13 , comprising further instructions that when performed by at least one processor of a computing device, cause the computing device to at least:
cause the machine learning model to generate the set of trajectories based, at least in part, on a set of actions associated with the robot and the set of observed trajectories.

15 . The non-transitory computer-readable medium of claim 13 , comprising further instructions that when performed by at least one processor of a computing device, cause the computing device to at least:
calculate the one or more values by at least comparing the set of observed trajectories and the set of trajectories through one or more functions.

16 . The non-transitory computer-readable medium of claim 13 , comprising further instructions that when performed by at least one processor of a computing device, cause the computing device to at least:
cause the machine learning model to generate the set of trajectories based, at least in part, on one or more simulations of the object.

17 . The non-transitory computer-readable medium of claim 13 , comprising further instructions that when performed by at least one processor of a computing device, cause the computing device to at least:
perform one or more gradient computations to update the one or more parameters of the machine learning model based, at least in part, on the one or more values.

18 . The non-transitory computer-readable medium of claim 13 , comprising further instructions that when performed by at least one processor of a computing device, cause the computing device to at least:
cause the machine learning model to generate a second set of trajectories based, at least in part, on the one or more values; and update the machine learning model based, at least in part, on the second set of trajectories.

19 . The non-transitory computer-readable medium of claim 13 , wherein the one or more properties include one or more joint types.

20 . The non-transitory computer-readable medium of claim 13 , comprising further instructions that when performed by at least one processor of a computing device, cause the computing device to at least:
calculate one or more responses of one or more components of the object to a set of actions of a robot using the machine learning model; and determine whether to cause the robot to perform the set of actions based, at least in part, on the one or more responses.

Assignees

• Nvidia Corp

Inventors

• Tozeto Ramos Fabio
• Garg Animesh
• Jatavallabhula Krishna Murthy
• Macklin Miles

Classifications

• B25J9/1664 Primary

  characterised by motion, path, trajectory planning · CPC title

• B25J9/163

  learning, adaptive, model based, rule based expert control · CPC title

• B25J9/1671 Primary

  characterised by simulation, either to verify existing program or to create and verify new program, CAD/CAM oriented, graphic oriented programming systems · CPC title

• G05B2219/39271

  Ann artificial neural network, ffw-nn, feedforward neural network · CPC title