Task Cooperation For Distributed Intelligent Systems In The Dynamic Environment

Basing on the theory of multi-agent, analyzed the problem in coordination among the distributed intelligent sub-system with heterogeneity which consist of the large complex system, and did some research in the coordination and control for such system in the dynamic environment. The main research in this paper can be classed as six parts: (1) For the interaction among agents on knowledge level, founded the uniform definition term. According to the standardization of ontolingua, we adopted the OOP language to define the term. Founded the model of KQML, which is often applied in communication of multi-agent system, and analyzed the synchronous and asynchronous communication models, also gave the principle to chose between them. (2) Basing on the game theory, gave the algorithm of forming the MAS coalition in task coordination, and analyzed the complexity and validity. The knowledge and experience holded by agent were used in the algorithm. And with the distribution, it is suit for MAS. (3) The sub-task allocation was transformed into set-covering problem, and a greedy heuristic algorithm was proposed. Then, the convergency and supremum of the algorithm were proved. At the same time, another algorithm was extended for solving the problem that sub-task can be divided, and the convergency and time complexity were also given. (4) For the synthesis of belief degree in multi-agent system, a recurrence algorithm was given to overcome the problems of positive and negative invalidations that exist in current algorithms. It has the properties of real time and rationality in solving conclusion synthesis of dynamic system. (5) The fuzzy cognitive map was used to construct the coordination model for distributed intelligent systems to coordination their decisions. Basing on this model, a novel coordination strategy was given, and the conception of stability of coordination in MAS was proposed. (6) The distributed optimized algorithm of task allocation and the proportional control were applied to the tn-level supply chain system, and proposed the control strategy of material flow. At the same time, basing on the method of ontology, the Java platform SCA was founded to simulate the supply chain system.