1998 Fiscal Year Final Research Report Summary
Modeling and Resolution of Feature Interactions in Multimedia Networks
| Project/Area Number |
09650413
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
情報通信工学
|
|---|
| Research Institution | Hiroshima City University (1998)
Osaka University (1997) |
|---|
Principal Investigator |
KAKUDA Yoshiaki Hiroshima City University, Dept.of Computer Engineering, Professor, 情報科学部, 教授 (40233671)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
INOUE Shinji Hiroshima City University, Dept.of Computer Engineering, Research Associate, 情報科学部, 助手 (50264932)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Keywords | feature interaction / extend finite state machine / detection / resolution / normal state / abnormal state |
|---|
| Research Abstract |
This project discussed feature interactions among telecommunication services in multimedia networks and proposed detection and resolution methods for the feature interactions as follows.
1. Detection of feature interactions
In order to relax the state explosion problem, this project proposed new methods to detect feature interactions in sufficient and necessary conditions using symmetry and acyclic expansion techniques. Furtheremore, this project proposed a new method to detect feature interactions in a necessary condition using P-invariant in the Petri Net theory. The experiments to detect feature interactions among advanced telephone services showed that time to detect such feature interactions by the P-invariant based method is almost proportional to the number of users and services.
2. Resolution of feature interactions.
The project proposed new methods to resolve feature interactions during execution of telecommunication services. In these methods, agents collect information concerning telecommunication services and resolve feaure interactions named abnormal states. One method is to store normal states periodically and to revert to the stored normal state when an abnormal state is detected. The other method is to restrict executable events so that any abnormal state is not reachable. There are tradeoffs between these two methods with respect to overheads and executability.
|
