An optimum strategy for dynamic and stochastic packet routing problems by chaotic neurodynamics

The most important issue in real packet routing problem on a computer network is how to alleviate packet congestion, because it often leads to unstable and insecure communication. In order to resolve the issue, various methods have already been proposed, for example, a probabilistic routing strategy, a routing strategy using mutual connection neural networks and so on. We have also proposed a new packet routing method which involves chaotic neurodynamics to avoid the congestion. We then showed that the proposed method exhibits high performance for various structures of the computer networks. In the present paper, we evaluated the proposed method under more realistic situation: packet generating probability depends on time, and spatial structure of the computer network itself. We firstly applied the proposed method to the computer networks with the complex structures, comparing with the Dijkstra algorithm and a tabu search algorithm. We then analyzed the effectiveness of the proposed routing method, introducing the method of surrogate data, a statistical hypothesis testing which has already been used in the field of nonlinear time series analysis. As a result, the chaotic neurodynamics is the most effective way to alleviate the packet congestion in the computer network under spatio-temporal dynamic packet generation.