A knowledge-based dynamic scheduling decision system

There is some limitation in the application of most knowledge-based scheduling systems as below: First, the knowledge sources have their own subjective tendentiousness; second, scheduling decision making relies on partial attributes. We construct a novel multi agent system for dynamic scheduling with Contract Net Protocol as its coordination mechanism. The system acquires standardized attributes knowledge from bids and knowledge of scheduling rules from experience of experts. By means of ballot, it solves the conflicts among scheduling rules to achieve knowledge fusion. Furthermore, each rule's credibility is gotten from vote to it. Through learning scheduling history, the system acquires credibility of each resource's attributes. The attribute importance in scheduling rules is figured out according to its definition in Rough Sets, and standardized to generate attribute's weight. The system contains a Knowledge-Based Decision Making Model which is on the basis of Fuzzy Petri Net. So we reason taking attributes' credibility for tokens, attributes' weights for input intensity and scheduling rule's credibility for output intensity. As a result, decision attribute's credibility is generated. At the end, we describe the process of knowledge-based decision making and accomplish the structure design.