Processing

Please wait...

PATENTSCOPE will be unavailable a few hours for maintenance reason on Saturday 31.10.2020 at 7:00 AM CET
Settings

Settings

Goto Application

1. WO2019060632 - OPTIMIZING POLICY CONTROLLERS FOR ROBOTIC AGENTS USING IMAGE EMBEDDINGS

Publication Number WO/2019/060632
Publication Date 28.03.2019
International Application No. PCT/US2018/052078
International Filing Date 20.09.2018
IPC
G06N 3/00 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
G06N 3/04 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
G06N 3/08 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
CPC
B25J 9/163
BPERFORMING OPERATIONS; TRANSPORTING
25HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
JMANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
9Programme-controlled manipulators
16Programme controls
1628characterised by the control loop
163learning, adaptive, model based, rule based expert control
B25J 9/1664
BPERFORMING OPERATIONS; TRANSPORTING
25HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
JMANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
9Programme-controlled manipulators
16Programme controls
1656characterised by programming, planning systems for manipulators
1664characterised by motion, path, trajectory planning
B25J 9/1697
BPERFORMING OPERATIONS; TRANSPORTING
25HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
JMANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
9Programme-controlled manipulators
16Programme controls
1694characterised by use of sensors other than normal servo-feedback from position, speed or acceleration sensors, perception control, multi-sensor controlled systems, sensor fusion
1697Vision controlled systems
G05B 13/0205
GPHYSICS
05CONTROLLING; REGULATING
BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
13Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
02electric
0205not using a model or a simulator of the controlled system
G05B 13/027
GPHYSICS
05CONTROLLING; REGULATING
BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
13Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
02electric
0265the criterion being a learning criterion
027using neural networks only
G06N 20/00
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
20Machine learning
Applicants
  • GOOGLE LLC [US]/[US]
Inventors
  • CHEBOTAR, Yevgen
  • SERMANET, Pierre
  • LYNCH, Harrison
Agents
  • PORTNOV, Michael
Priority Data
62/561,13320.09.2017US
Publication Language English (EN)
Filing Language English (EN)
Designated States
Title
(EN) OPTIMIZING POLICY CONTROLLERS FOR ROBOTIC AGENTS USING IMAGE EMBEDDINGS
(FR) OPTIMISATION DE DISPOSITIFS DE COMMANDE DE POLITIQUE POUR AGENTS ROBOTIQUES À L'AIDE D'INTÉGRATIONS D'IMAGE
Abstract
(EN)
There are provided systems, methods, and apparatus, for optimizing a policy controller to control a robotic agent that interacts with an environment to perform a robotic task. One of the methods includes optimizing the policy controller using a neural network that generates numeric embeddings of images of the environment and a demonstration sequence of demonstration images of another agent performing a version of the robotic task.
(FR)
L'invention concerne des systèmes, des procédés et un appareil, pour optimiser un dispositif de commande de politique destiné à commander un agent robotique qui interagit avec un environnement de manière à effectuer une tâche robotique. L'un des procédés consiste à optimiser le dispositif de commande de politique à l'aide d'un réseau neuronal qui génère des intégrations numériques d'images de l'environnement et une séquence de démonstration d'images de démonstration d'un autre agent effectuant une version de la tâche robotique.
Also published as
JP2019567999
Latest bibliographic data on file with the International Bureau