The Research Of Network Routing Algorithm With BP-CT Intelligence

Recently, people realize that the conventional mathematic methods can't build the precise model of complex system with the deep recognition of complex system. How to solve the complex problem through simulating human being's thinking becomes the focus. Computational intelligence provides the basis of artificial cognitive theory for human's thinking simulation and multi-agents techniques and experiment in compute provide the tools of implementation for artificial cognitive theory. In this paper, the concept of computational intelligence is introduced, and the artificial cognitive features in computational intelligence are induced. At the same time, multi-agents technique is recommended. In this paper, Q-leaning which combines the computational intelligence and multi-agents technique in engineering system and the application of Q-learning in complex network routing problem –Q-routing and AntNet routing algorithm are summarized. Then the feature of AntNet algorithm is primary analyzed and two shortages in AntNet is presented: the first, AntNet only present the probabilistic reasoning feature in computational intelligence, the complementary feature is not presented. The second, the forecasting ability with history experience and the self-adaptive ability are not presented. Then the Cross Target (CT) method presented in computational model theory in complex economics system is introduced and the consistency of CT method and Q-learning in artificial cognitive feature presentation is analyzed. So in this paper, combined the CT method and the AntNet routing algorithm, using the multi-agents technique, the artificial cognitive model--BP-CT ant routing algorithm is constructed. In this algorithm the agents have the simple artificial cognition and can interact with other agents and environment. We simulate this algorithm on OMNeT++ software platform and contrast with AntNet routing algorithm, and analyze the result of simulation. The result shows that the BP-CT ant routing algorithm in this paper is prior to AntNet in Quality of Service (QoS) such as throughput, average data packets delay, packets dropped ratio, proves the availability of the algorithm.