Optimization for packet routing using chaotic dynamics

We propose a new algorithm for controling packet routing using chaotic dynamics. First, we construct a packet routing method using a neural network based on information of shortest distances from a node to the other nodes, which works as the Dijkstra algorithm. We discovered that when the network structure is regular, the algorithm works well, however, when the network structure becomes irregular, the algorithm does not work well; almost all packets cannot be sent to their destinations because of congestion of packets at the nodes. To avoid this problem, we extended this model to employ chaotic dynamics. We confirm that our routing algorithm is very effective for many irregular and scale-free networks comparing with the Dijkstra algorithm