BN+BN

Behavior network with Bayesian network for intelligent agent

Kyung Joong Kim, Sung Bae Cho

Research output: Contribution to journalConference article

1 Citation (Scopus)

Abstract

In the philosophy of behavior-based robotics, design of complex behavior needs the interaction of basic behaviors that are easily implemented. Action selection mechanism selects the most appropriate behavior among them to achieve goals of robot. Usually, robot might have one or more goals that conflict each other and needs a mechanism to coordinate them. Bayesian network represents the dependencies among variables with directed acyclic graph and infers posterior probability using prior knowledge. This paper proposes a method to improve behavior network, action selection mechanism that uses the graph of behaviors, goals and sensors with activation spreading, using goal inference mechanism of Bayesian network learned automatically. Experimental results on Khepera mobile robot show that the proposed method can generate more appropriate behaviors.

Original languageEnglish
Pages (from-to)979-991
Number of pages13
JournalLecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science)
Volume2903
Publication statusPublished - 2003 Dec 1

Fingerprint

Intelligent agents
Intelligent Agents
Bayesian networks
Bayesian Networks
Robots
Mobile robots
Robotics
Chemical activation
Sensors
Robot
Directed Acyclic Graph
Posterior Probability
Prior Knowledge
Mobile Robot
Activation
Sensor
Experimental Results
Graph in graph theory
Interaction

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Computer Science(all)
  • Theoretical Computer Science
  • Engineering(all)

Cite this

@article{c1dd23c9c5e44a109e4a262f497c8517,
title = "BN+BN: Behavior network with Bayesian network for intelligent agent",
abstract = "In the philosophy of behavior-based robotics, design of complex behavior needs the interaction of basic behaviors that are easily implemented. Action selection mechanism selects the most appropriate behavior among them to achieve goals of robot. Usually, robot might have one or more goals that conflict each other and needs a mechanism to coordinate them. Bayesian network represents the dependencies among variables with directed acyclic graph and infers posterior probability using prior knowledge. This paper proposes a method to improve behavior network, action selection mechanism that uses the graph of behaviors, goals and sensors with activation spreading, using goal inference mechanism of Bayesian network learned automatically. Experimental results on Khepera mobile robot show that the proposed method can generate more appropriate behaviors.",
author = "Kim, {Kyung Joong} and Cho, {Sung Bae}",
year = "2003",
month = "12",
day = "1",
language = "English",
volume = "2903",
pages = "979--991",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

BN+BN : Behavior network with Bayesian network for intelligent agent. / Kim, Kyung Joong; Cho, Sung Bae.

In: Lecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science), Vol. 2903, 01.12.2003, p. 979-991.

Research output: Contribution to journalConference article

TY - JOUR

T1 - BN+BN

T2 - Behavior network with Bayesian network for intelligent agent

AU - Kim, Kyung Joong

AU - Cho, Sung Bae

PY - 2003/12/1

Y1 - 2003/12/1

N2 - In the philosophy of behavior-based robotics, design of complex behavior needs the interaction of basic behaviors that are easily implemented. Action selection mechanism selects the most appropriate behavior among them to achieve goals of robot. Usually, robot might have one or more goals that conflict each other and needs a mechanism to coordinate them. Bayesian network represents the dependencies among variables with directed acyclic graph and infers posterior probability using prior knowledge. This paper proposes a method to improve behavior network, action selection mechanism that uses the graph of behaviors, goals and sensors with activation spreading, using goal inference mechanism of Bayesian network learned automatically. Experimental results on Khepera mobile robot show that the proposed method can generate more appropriate behaviors.

AB - In the philosophy of behavior-based robotics, design of complex behavior needs the interaction of basic behaviors that are easily implemented. Action selection mechanism selects the most appropriate behavior among them to achieve goals of robot. Usually, robot might have one or more goals that conflict each other and needs a mechanism to coordinate them. Bayesian network represents the dependencies among variables with directed acyclic graph and infers posterior probability using prior knowledge. This paper proposes a method to improve behavior network, action selection mechanism that uses the graph of behaviors, goals and sensors with activation spreading, using goal inference mechanism of Bayesian network learned automatically. Experimental results on Khepera mobile robot show that the proposed method can generate more appropriate behaviors.

UR - http://www.scopus.com/inward/record.url?scp=7444239161&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=7444239161&partnerID=8YFLogxK

M3 - Conference article

VL - 2903

SP - 979

EP - 991

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

ER -