In overpopulated big cities in Japan regulation of trains delaying during rush-hours is indispensable. However, train regulation is a real-time dynamically changing complex problem. It involves not only physical factors like conflicts of platforms but human or social factors concerning commuters, crews, etc.. Therefore its automation has been impossible. Coping with these difficulties, DGCAI (Dynamic Goal centered Coordination (strategy based) AI system) was developed. It provides the knowledge representation and inference engine for 1) dynamic control of goal generation and attainment under real-time environments, and 2) easy-to-coordinate goal decomposition and attainment, and so on. DGCAI is applied to a train regulation expert system. This system is practically used in one of the biggest cities in Japan. This success is considered to prove the validation of this AI regulation system and DGCAI for solving real-time complex problems such as regulation of trains running every 2 or 3 minutes. However, the scientific validation of this system and DGCAI has many problems. For example, 1) test cases involving social or human factors are difficult to prepare, 2) the measuring for the validation is difficult to execute correctly or to announce the result officially. These problems are also discussed in this paper.
A dynamic complex AI system framework, its application, and the validation problem
LIT, Leipziger Informatik-Tage, 5 ; 145-150
1997
6 Seiten, 7 Bilder, 7 Quellen
Aufsatz (Konferenz)
Englisch
Dynamic goal centered coordination AI system: application and the validation problem
Tema Archiv | 1997
|A Framework for Verification and Validation of Complex Aerospace Systems
British Library Conference Proceedings | 2013
|A Fuzzy Framework for Dynamic Design Validation of Uncertain Structures
British Library Conference Proceedings | 2001
|Validation and extension of an MDO framework including dynamic aeroelastic analysis
Deutsches Zentrum für Luft- und Raumfahrt (DLR) | 2020
|