Self-Adjustment Mechanism Guaranteeing Asymptotic Stability of Clusters Formed by Autonomous Decentralized Mechanism

Abstract

Clustering mechanisms are crucial for scalable rout- ing in mobile ad hoc networks (MANETs). We have previously proposed an autonomous decentralized clustering mechanism that is based on differential equations. This method adopts both the diffusion effect for autonomous decentralized control, and the drift effect for balancing with the diffusion effect. In our conventional mechanism, we need to set the strength of diffu- sion effect stronger than the drift effect because fine tuning for balancing both effects is difficult. However, it results in a de- crease of the number of clusters by the diffusion effect, and it is a problem on forming sustainable cluster structures. We need the solution of this problem for clustering in MANET. In this paper, we propose a mechanism autonomously adjusting the effects of both diffusion and drift. We show that our proposed mechanism can form sustainable cluster structures reflecting the characteristic of our conventional mechanism. Finally, we eval- uate the effectiveness of the proposed mechanism for an envi- ronment in which the fusion of subnetworks occurs. From the above background, we have previously pro- posed autonomous decentralized clustering mechanism to introduce hierarchical structures in MANETs (10). In our mechanism, each node has values of a certain distribution in advance, and nodes exchange a portion of their value with the adjacent nodes by obeying a predefined action rule based only on local information. After that, our mechanism can form spatial structures in the network, then cluster structures are decided based on local maxi- mum values of the distribution. We also have verified the effectiveness of our mechanism by comparing with au- tonomous decentralized clustering mechanism based on bio-inspired approach which is well known autonomous decentralized mechanism (11), (12). In this verification,