Autonomous Decentralized Mechanism of Structure Formation Adapting to Network Conditions

We have already proposed the framework of autonomous decentralized control based on local-interaction as a novel control mechanism for communication networks. This framework is based on the relation between local interaction and the solution yielded by a partial differential equation. In this framework, the behavior of the whole system is indirectly controlled by appropriately designing the autonomous operation of the subsystems. That is, the local action rules (micro-level) are designed to produce an appropriate state of the whole system at the macro-level. In previous studies, we proposed flow control based on a diffusion equation to realize autonomous congestion avoidance in networks. This paper proposes a new autonomous decentralized structure forming method based on our framework. First, we introduce the renormalization transformation of the diffusion phenomenon for a one-dimensional network model and propose autonomous decentralized forming of a structure that has finite spatial size. In addition, to apply this method to a general network topology, we first extend the one-dimensional network model to a n-dimensional Euclid space and derive the control rule that can be applied to a general network by using the discretization method. As an application example, we demonstrate the autonomous decentralized clustering of a sensor network for a two-dimensional lattice network model and show the characteristics of the proposed method in this application.