Kenji Horikawa, Masaki Aida and Toshiharu Sugawara, NTT, Japan
In this paper, we discuss the concept of multiagent systems as applied to high-speed network traffic control. In the high-speed domain, the propagation delay in a network causes significant problems. One of them is that any centralized control is too late to maintain normal network operations, so distributed traffic control by autonomous agents is required. The basic idea of our traffic control for high-speed networks is a combination of distributed local control and centralized global control; that is, before the global control becomes effective, all network nodes try to maintain their operations in cooperation with other nodes. The significance of propagation delay in such networks requires that the communication delay is included in the model of the node agents. This paper proposes a multiagent traffic control and illustrates the experimental results. The results show that the control based on the multiagent system can improve the efficiency of network control.