Robot system and driving method

Patent metadata

Field	Value
Publication number	US-11654553-B2
Application number	US-202016799955-A
Country	US
Kind code	B2
Filing date	Feb 25, 2020
Priority date	Sep 4, 2019
Publication date	May 23, 2023
Grant date	May 23, 2023

Abstract

A robot system according to an embodiment includes one or more processors. The processors acquire first input data predetermined as data affecting an operation of a robot. The processors calculate a calculation cost of inference processing using a machine learning model for inferring control data used for controlling the robot, on the basis of the first input data. The processors infer the control data by the machine learning model set according to the calculation cost. The processors control the robot using the inferred control data.

First claim

What is claimed is:

1. A robot system, comprising:
one or more processors configured to
acquire first input data predetermined as data affecting an operation of a robot;
calculate, based on the first input data, a calculation cost of inference processing required for a machine learning model to infer control data used for controlling the robot;
infer the control data by the machine learning model, which is set according to the calculated calculation cost; and
control the robot using the inferred control data,

wherein the one or more processors are further configured to calculate the calculation cost using one of a previously obtained function that inputs the first input data and outputs the calculation cost, or a distribution indicating a relationship between a plurality of calculation costs and a plurality of pieces of first input data.

2. The robot system according to claim 1 , wherein the one or more processors are further configured to:
determine a control parameter of the machine learning model for executing inference according to the calculated calculation cost;
change the machine learning model according to the determined control parameter; and
infer the control data using the changed machine learning model.

3. The robot system according to claim 1 ,
wherein the robot is a picking robot that grips a target object, and
the first input data is at least one of a number of target objects, a number of types of target objects, a weight of a target object, and a success rate of gripping the target object.

4. The robot system according to claim 1 ,
wherein the robot is a moving robot, and
the first input data is at least one of a width of a route that the robot moves, a number of objects around the robot, a number of objects mounted on the robot, a number of types of objects mounted on the robot, and a weight of an object mounted on the robot.

5. The robot system according to claim 1 , wherein the machine learning model is a model that is set such that an inference accuracy is decreased as the calculated calculation cost is decreased.

6. The robot system according to claim 1 , wherein the one or more processors are further configured to learn the machine learning model using a loss function according to the calculated calculation cost.

7. The robot system according to claim 1 , wherein the one or more processors are further configured to infer the control data by inputting second input data, different from the first input data, to the machine learning model.

8. A driving method, comprising:
acquiring first input data predetermined as data affecting an operation of a robot;
calculating, based on the first input data, a calculation cost of inference processing required for a machine learning model to infer control data used for controlling the robot;
inferring the control data by the machine learning model set according to the calculation cost; and
controlling the robot using the inferred control data,
wherein the calculation step comprises calculating the calculation cost by using one of a previously obtained function that inputs the first input data and outputs the calculation cost, or a distribution indicating a relationship between a plurality of calculation costs and a plurality of pieces of first input data.

9. The robot system according to claim 1 , wherein for the machine learning model, the one or more processors are configured to set a control parameter according to the calculated calculation cost, the control parameter being a size of the machine learning model, a channel width, or a step width.

Assignees

• Toshiba Kk

Inventors

• Nitta Shuhei
• Yaguchi Atsushi
• Sakata Yukinobu
• Tanizawa Akiyuki
• Takeyama Yasutoyo
• Watanabe Tomoki

Classifications

• G06N3/0495

  Quantised networks; Sparse networks; Compressed networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/082

  modifying the architecture, e.g. adding, deleting or silencing nodes or connections · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title