Page 189 - Mechatronics for Safety, Security and Dependability in a New Era

Ch36-I044963.fm Page 173 Tuesday, August 1, 2006 3:10 PM
Page 173
1, 2006
3:10 PM
Tuesday, August
Ch36-I044963.fm
173
173
INTERACTIVE BEHAVIORAL DESIGN BETWEEN
AUTONOMOUS BEHAVIORAL CRITERIA LEARNING SYSTEM
AND HUMAN

Min An and Toshiharu Taura

Graduate School of Science and Technology, Kobe University,
1 -1, Rokkodai, Nada Kobe, 657-8501, Japan

ABSTRACT
Conventional robotic behaviors are directly programmed depending on programmer's personal
experience. On the other hand, an artist cannot easily convey their interesting behavioral patterns to
the programmers due to difficulty in expressing such behaviors. Therefore, interesting behavioral
patterns can hardly be produced at present. It is necessary to develop an effective method of designing
robotic behavior. In this study, the authors propose a method of designing robotic behavior though
interaction with a computer and establish a design system with the method. For demonstrating the
design system, we invited both engineering students and art students to use this design system and
value it in our survey. The survey results showed that the design system could not only help a user
present the behavioral pattern through an interface with the computer, but could also expand the user's
creativity from the interface with the computer.

KEYWORDS

Robotics, genetic algorithm, genetic programming, behavioral design, interactive design

INTERODUCTION

A variety of robots are created all over the world. However, there has been little research focusing on
robotic behavioral design. It is necessary to develop an effective method of designing robotic behavior.
In this study, the authors aim to establish a method of designing robotic behaviors by operating
behavioral criteria, because one of the most effective techniques in design is the operation of multiple
information or knowledge. For example, we can combine the action of moving a leg forward with the
action of rotating it at the hip into a kicking behavior. Here, the behavioral criteria of a computer
program are used to bring the behavior candidates into an optimum behavior. The behavioral criteria
measure the behavior candidates in terms of the error produced by the computer program. The closer