No abstract

One approach to meet the requirements of next generation real-time systems is to exploit AI techniques for specifying how real-time systems should behave. For this purpose, the ARON-technique (Alternatives Regularly Qrganized and Numbered) was developed. It is based on a simple, reflectory knowledge model, which can be implemented very efficiently. Moreover, it utilizes knowledge compilation in order to improve efficiency and to eliminate data-dependant processing-times. The ARON-technique may be used further to implement machine learning. Therefore knowledge, which models a specific method of human learning, is implemented in a second system, also using the ARON-technique. This so-called "hyper-system'' performs an online-assessment of the effects the subordinate "basic-system'' has on the outer environment.If these results are not suitable, the hyper-system modifies the knowledge of the basic-system online. In this manner, self-leaming capabilities are achieved without human interaction. Real-time applicability of this learning technique is demonstrated by self-tuning knowledge based speed control of a real asynchronous motor.

No abstract

No abstract

No abstract