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 language | English |
|---|---|
| Pages (from-to) | 979-991 |
| Number of pages | 13 |
| Journal | Lecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science) |
| Volume | 2903 |
| Publication status | Published - 2003 Dec 1 |
Fingerprint
All Science Journal Classification (ASJC) codes
- Hardware and Architecture
- Computer Science(all)
- Theoretical Computer Science
- Engineering(all)
Cite this
}
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 journal › Conference 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 -