APPLICATION OF UNCERTAIN VARIABLES AND LEARNING PROCESS TO DECISION MAKING IN A CLASS OF COMPLEX SYSTEMS

摘要

In recent years two new concepts have been introduced and developed as a tool for decision making in a class of uncertain systems (in particular, control systems) described by a relational knowledge representation: uncertain variables characterized by an expert and a learning process consisting in step by step knowledge validation and updating. The purpose of this paper is to show how these two approaches may be combined and used for decision making in a class of complex systems with a distributed knowledge. In the first part of the paper the decision problem based on the uncertain variables and the learning process are shortly described. Then the combination of these approaches in one learning system in which the expert's knowledge is modified according to current results of the learning is described and the algorithm of the decision making in the learning system is presented. In the second part it is shown how to apply these concepts in two cases of complex systems: two-level system with the distributed knowledge and a complex of parallel operations. A simple example and a result of simulations illustrate the presented approach.