Robot device and robot control method

• US 20040210347A1
• Filed: 01/16/2004
• Published: 10/21/2004
• Est. Priority Date: 05/20/2002
• Status: Abandoned Application

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Sawada, Tsutomu, Fujita, Masahiro, Takagi, Tsuyoshi, Hanagata, Osamu
• Application Number:
  US10/484,058
• Publication Number:
  US 20040210347A1
• Time in Patent Office: Days
• US Class Current: 700/246
• CPC Class Codes: G06N 3/008 based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour

Abstract

The invention provides a robot apparatus and a robot controlling method wherein the robot apparatus can behave so that its user may not lose interest in the robot apparatus. A behavior management section (72) selects one of a seeking behavior and a knowledge utilizing behavior based on a behavior selection probability. When the behavior management section (72) executes the selected behavior (a), a reward (r) is provided from an environment/user (111). The behavior (72) updates a behavior value based on the reward (r) and applies the behavior value to the Boltzmann distribution to determine the behavior selection probability. Further, the behavior management section (72) varies the Boltzmann temperature based on input information. The present invention can be applied to a reinforced learning system for a robot.

Claims

1. A robot apparatus which outputs a predetermined behavior, comprising:

behavior management means for selecting a behavior which is based on knowledge relating to behaviors in the past or a behavior which is not based on the knowledge relating to the behaviors in the past; and

control means for controlling said robot apparatus;

said control means controlling said robot apparatus based on the behavior selected by said behavior management means which is based on the knowledge relating to the behaviors in the past or is not based on the knowledge relating to the behaviors in the past to output the predetermined behavior.

2. A robot apparatus according to claim 1, wherein the behavior which is based on the knowledge relating to the behaviors in the past is a knowledge utilizing behavior, and

the behavior which is not based on the knowledge relating to the behaviors in the past is a seeking behavior.

3. A robot apparatus according to claim 2, wherein said behavior management means selects one of the knowledge utilizing behavior and the seeking behavior based on a behavior selection probability.

4. A robot apparatus according to claim 3, wherein the behavior selection probability varies in response to a reward from a user for a behavior executed by said robot apparatus.

5. A robot apparatus according to claim 3, wherein the behavior selection probability varies based on the Boltzmann temperature.

6. A robot apparatus according to claim 5, wherein said robot apparatus holds a feeling parameter which determines a feeling of said robot apparatus, and

the Boltzmann temperature varies based on the feeling parameter.

7. A robot controlling method for a robot apparatus which outputs a predetermined behavior, comprising:

a behavior management step of selecting a behavior which is based on knowledge relating to behaviors in the past or a behavior which is not based on the knowledge relating to the behaviors in the past; and

a control step of controlling said robot apparatus;

the process at the control step controlling said robot apparatus based on the behavior selected by the process at the behavior management step which is based on the knowledge relating to the behaviors in the past or is not based on the knowledge relating to the behaviors in the past to output the predetermined behavior.