신경회로망을 이용한 호 접수제어기
- Alternative Title
- The Call Adimission Controller by Neural Networks
- Abstract
- B-ISDN의 전송 및 교환 기법으로 정착되고 있는 ATM망의 폭주제어를 위하여, 신경회로망을 이용한 호 접수 제어기를 설계한다. 신경회로망을 이용한 호 접수제어는 가입자가 선언한 트래픽 기술인자의 불확실성에 대한 정보를 제어기가 가지고 ?殮? 때문에, 실제 트래픽이 호 설정요구시 협약된 트래픽의 특성과 다소 상이하더라도 견고성(Robustness) 있는 트래픽 제어를 할 수 있는 장점을 가지고 있다. 본 연구에서는 신경회로망 제어기를 보다 효율적으로 학습시킬 수 있도록 두가지 학습 패턴 테이블(호 연결 요청 수락 패턴 테이블과 호 연결 요청 거절 패턴 테이블)을 이용하는 학습 알고리즘을 제안한다. 또한 모의 실험을 통하여, 제안한 신경회로망 호 접수 제어기의 성능을 분석한다.
This paper describes a call admission controller by neural network to control the traffic in ATM network for service quality. Because the information of the traffic descriptor declared by users has a characteristic of uncertainty, the design and implementation of efficient controller in the network is very difficult task. The proposed ATM call admission controller by neural network is adaptive and easy to implement. We use two learning table for efficient learning of back-propagation neural network in controller. The performance of the proposed call admission controller is evaluated by simulation, and the results are compared with those of the other control technique.
This paper describes a call admission controller by neural network to control the traffic in ATM network for service quality. Because the information of the traffic descriptor declared by users has a characteristic of uncertainty, the design and implementation of efficient controller in the network is very difficult task. The proposed ATM call admission controller by neural network is adaptive and easy to implement. We use two learning table for efficient learning of back-propagation neural network in controller. The performance of the proposed call admission controller is evaluated by simulation, and the results are compared with those of the other control technique.
- Author(s)
- 허정석; 양성룡
- Issued Date
- 1993
- Type
- Research Laboratory
- URI
- https://oak.ulsan.ac.kr/handle/2021.oak/3854
http://ulsan.dcollection.net/jsp/common/DcLoOrgPer.jsp?sItemId=000002024380
- 공개 및 라이선스
-
- 파일 목록
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.