Abstract
A novel knowledge extraction method from autonomous behavior of multiple mobile robots by an artificial life approach is proposed in this paper. The knowledge is expressed by if-then rules and we employ a neural network for the knowledge extraction. The proposed method has the following features: 1)The structure of knowledge extraction neural network and the learning algorithm are simple; 2)Understanding and modification of the extracted knowledge are easy because weights in the knowledge extraction network directly represent the antecedents and the consequents of the if-then rules; 3)The network itself has an ability of inference using the extracted knowledge. We used a lot of autonomous mobile robots in various environments. Each robot has to avoid the obstacles to get to the goal and the local behavior is extracted and integrated in the knowledge extraction neural network as global knowledge. We confirmed the validity of the proposed method by computer simulations.
| Original language | English |
|---|---|
| Title of host publication | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
| Publisher | Springer Verlag |
| Pages | 177-186 |
| Number of pages | 10 |
| Volume | 1285 |
| ISBN (Print) | 3540633995, 9783540633990 |
| Publication status | Published - 1997 |
| Event | 1st Asia-Pacific Conference on Simulated Evolution and Learning, SEAL 1996 - Taejon, Korea, Republic of Duration: 1996 Nov 9 → 1996 Nov 12 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 1285 |
| ISSN (Print) | 03029743 |
| ISSN (Electronic) | 16113349 |
Other
| Other | 1st Asia-Pacific Conference on Simulated Evolution and Learning, SEAL 1996 |
|---|---|
| Country | Korea, Republic of |
| City | Taejon |
| Period | 96/11/9 → 96/11/12 |
Fingerprint
ASJC Scopus subject areas
- Computer Science(all)
- Theoretical Computer Science
Cite this
Knowledge extraction using neural network by an artificial life approach. / Makita, Yuji; Hagiwara, Masafumi.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1285 Springer Verlag, 1997. p. 177-186 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1285).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Knowledge extraction using neural network by an artificial life approach
AU - Makita, Yuji
AU - Hagiwara, Masafumi
PY - 1997
Y1 - 1997
N2 - A novel knowledge extraction method from autonomous behavior of multiple mobile robots by an artificial life approach is proposed in this paper. The knowledge is expressed by if-then rules and we employ a neural network for the knowledge extraction. The proposed method has the following features: 1)The structure of knowledge extraction neural network and the learning algorithm are simple; 2)Understanding and modification of the extracted knowledge are easy because weights in the knowledge extraction network directly represent the antecedents and the consequents of the if-then rules; 3)The network itself has an ability of inference using the extracted knowledge. We used a lot of autonomous mobile robots in various environments. Each robot has to avoid the obstacles to get to the goal and the local behavior is extracted and integrated in the knowledge extraction neural network as global knowledge. We confirmed the validity of the proposed method by computer simulations.
AB - A novel knowledge extraction method from autonomous behavior of multiple mobile robots by an artificial life approach is proposed in this paper. The knowledge is expressed by if-then rules and we employ a neural network for the knowledge extraction. The proposed method has the following features: 1)The structure of knowledge extraction neural network and the learning algorithm are simple; 2)Understanding and modification of the extracted knowledge are easy because weights in the knowledge extraction network directly represent the antecedents and the consequents of the if-then rules; 3)The network itself has an ability of inference using the extracted knowledge. We used a lot of autonomous mobile robots in various environments. Each robot has to avoid the obstacles to get to the goal and the local behavior is extracted and integrated in the knowledge extraction neural network as global knowledge. We confirmed the validity of the proposed method by computer simulations.
UR - http://www.scopus.com/inward/record.url?scp=84947942538&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84947942538&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:84947942538
SN - 3540633995
SN - 9783540633990
VL - 1285
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 177
EP - 186
BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PB - Springer Verlag
ER -